diff --git a/2002.12229/main_diagram/main_diagram.drawio b/2002.12229/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..479b16c5d10934d09a63dda50d2a16de17b44254 --- /dev/null +++ b/2002.12229/main_diagram/main_diagram.drawio @@ -0,0 +1 @@ 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 \ No newline at end of file diff --git a/2002.12229/main_diagram/main_diagram.pdf b/2002.12229/main_diagram/main_diagram.pdf new file mode 100644 index 0000000000000000000000000000000000000000..de15c1e1a1d3937ab6dc933ea2f6fef5a508fc04 Binary files /dev/null and b/2002.12229/main_diagram/main_diagram.pdf differ diff --git a/2002.12229/paper_text/intro_method.md b/2002.12229/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..4b5c272594da8c2c4c4cdc6692920a9f0c044255 --- /dev/null +++ b/2002.12229/paper_text/intro_method.md @@ -0,0 +1,67 @@ +# Introduction + +In this section, we introduce more detailed background knowledge. + +Let $\mathbf{x}$ be a high-dimensional continuous variable. We suppose that $\mathbf{x}$ is drawn from $p^*(\mathbf{x})$, which is the true data distribution. Given a collected dataset $\mathcal{D} = \{\mathbf{x}_1, \mathbf{x}_2,...,\mathbf{x}_D\}$, we are interested in approximating $p^*(\mathbf{x})$ with a model $p_{\theta}(\mathbf{x})$. We optimize $\theta$ by minimizing the negative log-likelihood $$\begin{equation} +\mathcal{L}(\mathcal{D}) = \sum_{i=1}^{D} - \log p_{\theta}(\mathbf{x}_i). +\label{eq:likelihood} +\end{equation}$$ + +For some settings, variable $\tilde{\mathbf{x}}$ is discrete, e.g., image pixel values are often integers. In these cases, we dequantize $\tilde{\mathbf{x}}$ by adding continuous noise $\bm{\mu}$ to it, resulting in a continuous variable $\mathbf{x} = \tilde{\mathbf{x}} + \bm{\mu}$. As shown by @ho2019flow, the log-likelihood of $\tilde{\textbf{x}}$ is lower-bounded by the log-likelihood of $\mathbf{x}$. + +Normalizing flows enable computation of $p_{\theta}(\mathbf{x})$, even though it is usually intractable for many other model families. A normalizing flow [@rezende2015variational] is composed of a series of invertible functions $\mathbf{f} = \mathbf{f}_1 \circ \mathbf{f}_2 \circ ... \circ \mathbf{f}_K$, which transform $\mathbf{x}$ to a latent code $\mathbf{z}$ drawn from a simple distribution. Therefore, with the *change of variables* formula, we can rewrite $\log p_{\theta}(\mathbf{x})$ to be $$\begin{equation} +\log p_{\theta}(\mathbf{x}) = \log p_{Z}(\mathbf{z}) + \sum_{i=1}^{K} \log \left|\det \left(\frac{\partial \mathbf{f}_i}{\partial \mathbf{r}_{i-1}}\right)\right|, +\label{eq:relikelihood} +\end{equation}$$ where $\mathbf{r}_i = \mathbf{f}_i(\mathbf{r}_{i-1})$, $\mathbf{r}_{0} = \mathbf{x}$, and $\mathbf{r}_{K}=\mathbf{z}$. + +Emerging convolutions [@hoogeboom2019emerging] combine two autoregressive convolutions [@germain2015made; @kingma2016improved]. Formally, $$\begin{eqnarray*} + \mathbf{M}'_1 = \mathbf{M}_1 \odot \mathbf{A}_1, ~~~~~~~~ + \mathbf{M}'_2 = \mathbf{M}_2 \odot \mathbf{A}_2, ~~~~~~~~ + \mathbf{y} = \mathbf{M}'_2 \star (\mathbf{M}'_1 \star \mathbf{x}), +\end{eqnarray*}$$ where $\mathbf{M}_1, \mathbf{M}_2$ are convolutional kernels whose size is $c \times c \times d \times d$, and $\mathbf{A}_1, \mathbf{A}_2$ are binary masks. The symbol $\star$ represents the convolution operator.[^2] An emerging convolutional layer has the same receptive fields as standard convolutional layers, which can capture correlations between a target pixel and its neighbor pixels. However, like other autoregressive convolutions, computing the inverse of an emerging convolution requires sequentially traversing each dimension of input, so its computation is not parallelizable and is a computational bottleneck when the input is high-dimensional. + +Periodic convolutions [@hoogeboom2019emerging; @Finzi2019Invertible] use discrete Fourier transformations to transform both the input and the kernel to Fourier domain. A periodic convolution is computed as $$\begin{equation*} +\mathbf{y}_{u,:,:} = \sum_{v} \mathcal{F}^{-1}(\mathcal{F}(\mathbf{M}^{(p)}_{u,v,:,:})\odot \mathcal{F}(\mathbf{x}_{v,:,:})), +\end{equation*}$$ where $\mathcal{F}$ is a discrete Fourier transformation, and $\mathbf{M}^{(p)}$ is the convolution kernel whose size is $c \times c \times d \times d$. The computational complexity of periodic convolutions is $\mathcal{O}(c^2hw\log(hw) +c^3hw)$. In our experiments, we found that the Fourier transformation requires a large amount of memory. These two problems impact the efficiency of both training and sampling when the input is high-dimensional. + +Memory-Efficient Woodbury transformations can effectively reduce the space complexity. The main idea is to perform spatial transformations along the height and width axes separately, i.e., a height transformation and a width transformation. The transformations are: $$\begin{eqnarray} +\mathbf{x}_c &=& (\mathbf{I}^{(c)} + \mathbf{U}^{(c)}\mathbf{V}^{(c)}) \mathbf{x}, \nonumber\\ +\mathbf{x}_w &=& \text{reshape}(\mathbf{x}_c, (ch, w)), \nonumber\\ +\mathbf{x}_w &=& \mathbf{x}_c (\mathbf{I}^{(w)} + \mathbf{U}^{(w)}\mathbf{V}^{(w)}), \nonumber\\ +\mathbf{x}_h &=& \text{reshape}(\mathbf{x}_w, (cw, h)), \nonumber \\ +\mathbf{y} &=& \mathbf{x}_h(\mathbf{I}^{(h)} + \mathbf{U}^{(h)}\mathbf{V}^{(h)}), \nonumber\\ +\mathbf{y} &=& \text{reshape}(\mathbf{y}, (c, hw)), +\label{eq:me-w} +\end{eqnarray}$$ where $\text{reshape}(\mathbf{x}, (n,m))$ reshapes $\mathbf{x}$ to be an $n \times m$ matrix. Matrices $\mathbf{I}^{(w)}$ and $\mathbf{I}^{(h)}$ are $w$- and $h$-dimensional identity matrices, respectively. Matrices $\mathbf{U}^{(w)}, \mathbf{V}^{(w)}, \mathbf{U}^{(h)}$, and $\mathbf{V}^{(h)}$ are $w \times d_w$, $d_w \times w$, $w \times d_w$, and $d_w \times w$ matrices, respectively, where $d_w$ and $d_h$ are constant latent dimensions. + +Using the Woodbury matrix identity and the Sylvester's determinant identity, we can compute the inverse and Jacobian determinant: $$\begin{eqnarray} +\mathbf{y} &=& \text{reshape}(\mathbf{y}, (cw, h)), \nonumber\\ +\mathbf{x}_h &=& \mathbf{y}(\mathbf{I}^{(h)} - \mathbf{U}^{(h)}(\mathbf{I}^{(d_h)} + \mathbf{V}^{(h)}\mathbf{U}^{(h)})^{-1}\mathbf{V}^{(h)}), \nonumber\\ +\mathbf{x}_w &=& \text{reshape}(\mathbf{x}_h, (ch, w)), \nonumber\\ +\mathbf{x}_w &=& \mathbf{x}_w(\mathbf{I}^{(w)} - \mathbf{U}^{(w)}(\mathbf{I}^{(d_w)} + \mathbf{V}^{(w)}\mathbf{U}^{(w)})^{-1}\mathbf{V}^{(w)}), \nonumber \\ +\mathbf{x}_c &=& \text{reshape}(\mathbf{x}_w, (c, hw)), \nonumber\\ +\mathbf{x} &=& (\mathbf{I}^{(c)} - \mathbf{U}^{(c)}(\mathbf{I}^{(d_c)} + \mathbf{V}^{(c)}\mathbf{U}^{(c)})^{-1}\mathbf{V}^{(c)})\mathbf{x}_c, +\end{eqnarray}$$ $$\begin{eqnarray} +\log \left| \det(\frac{\partial \mathbf{y}}{\partial \mathbf{x}}) \right| &= hw \log\left|\det(\mathbf{I}^{(d_c)}+\mathbf{V}^{(c)}\mathbf{U}^{(c)})\right| + ch\log\left|\det(\mathbf{I}^{(d_w)}+\mathbf{V}^{(w)}\mathbf{U}^{(w)})\right| \nonumber\\ +&+ cw\log\left|\det\left(\mathbf{I}^{(d_h)}+\mathbf{V}^{(h)}\mathbf{U}^{(h)}\right)\right|, +\end{eqnarray}$$ where $\mathbf{I}^{(d_w)}$ and $\mathbf{I}^{(d_h)}$ are $d_w$- and $d_h$-dimensional identity matrices, respectively. The Jacobian of the $\text{reshape}()$ is an identity matrix, so its log-determinant is $0$. + +We call Equation [\[eq:me-w\]](#eq:me-w){reference-type="ref" reference="eq:me-w"} the memory-efficient Woodbury transformation because it reduces space complexity from $\mathcal{O}(c+hw)$ to $\mathcal{O}(c+h+w)$. This method is effective when $h$ and $w$ are large. To analyze its complexity, we let all latent dimensions be less than $d$ as before. The complexity of forward transformation is $\mathcal{O}(dchw)$; the complexity of computing the determinant is $\mathcal{O}(d(c+h+w)+d^3)$; and the complexity of computing the inverse is $\mathcal{O}(dchw + d^2(c+ch+cw)+d^3)$. The same as Woodbury transformations, when the input is high dimensional, we can omit $d$. Therefore, the computational complexities of the memory-efficient Woodbury transformation are also linear with the input size. + +We list the complexities of different methods in Table [2](#tab:complexity){reference-type="ref" reference="tab:complexity"}. We can see that the computational complexities of Woodbury transformations are comparable to other methods, and maybe smaller when the input is high-dimensional, i.e., the $c,h,w$ are big. + +:::: center +::: {#tab:complexity} + Method Forward Backward + ---------------------------- ---------------------------------- ---------------------------------- + 1x1 convolution $\mathcal{O}(c^2hw+c^3)$ $\mathcal{O}(c^2hw)$ + Periodic conolution $\mathcal{O}(chw\log(hw)+c^3hw)$ $\mathcal{O}(chw\log(hw)+c^2hw)$ + Emerging convolution $\mathcal{O}(c^2hw)$ $\mathcal{O}(c^2hw)$ + ME-Woodbury transformation $\mathcal{O}(dchw)$ $\mathcal{O}(dchw)$ + Woodbury transformation $\mathcal{O}(dchw)$ $\mathcal{O}(dchw)$ + + : Comparisons of computational complexities. +::: +:::: + +In this section, we present additional details about our experiments to aid reproducibility. diff --git a/2004.00048/main_diagram/main_diagram.drawio b/2004.00048/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..e4c5a3add874c3efd6f5341f507edf70c4f84a42 --- /dev/null +++ b/2004.00048/main_diagram/main_diagram.drawio @@ -0,0 +1 @@ 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 \ No newline at end of file diff --git a/2004.00048/main_diagram/main_diagram.pdf b/2004.00048/main_diagram/main_diagram.pdf new file mode 100644 index 0000000000000000000000000000000000000000..2a8a583e665a4f53e2e07f0308512399bee9a7de Binary files /dev/null and b/2004.00048/main_diagram/main_diagram.pdf differ diff --git a/2004.00048/paper_text/intro_method.md b/2004.00048/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..8e6e923539cd72d4a00e61cac3b36532854ef8d2 --- /dev/null +++ b/2004.00048/paper_text/intro_method.md @@ -0,0 +1,98 @@ +# Introduction + +We recall the single agent fully-observable RL setting [@sutton1998introduction], where the environment is typically formulated as a Markov decision process (MDP). At every time step, $t=1, 2, \dots$, the agent observes the environment's state $s_t \in \mathcal{S}$, and uses it to select an action $a_t \in \mathcal{A}$. As a consequence, the agent receives a reward $r_t\in\mathcal{R}\subset \mathbb{R}$ and the environment transitions to the state $s_{t+1}$. The tuple ($s_{t+1}, r_t$) is sampled from the static probability distribution $p: \mathcal{S} \times \mathcal{A} \to \mathcal{P}(\mathcal{S}\times \mathcal{R})$ whilst the actions $a_t$ are sampled from the parametric policy function $\pi_\theta: \mathcal{S} \to \mathcal{P}(\mathcal{A})$: $$\begin{equation} + s_{t+1}, r_t \sim p(s_{t+1}, r_{t}|s_t, a_t), \quad a_t \sim \pi_\theta(a_t|s_t) +\end{equation}$$ The goal of the agent is to find the optimal policy parameters $\theta^*$ that maximise the expected return $\Bar{R}=\mathbb{E}[\sum_{t=0}^\infty\gamma^t r_t]$, where $\gamma$ is the discount factor. In the more general framework, the state is only partially observable, meaning that the agent can not directly observe the state but instead it observes $o_t \in \mathcal{O}$ which is typically given by a function of the state. In this situation, the environment is modelled by a partial observable Markov decision process (POMDP) and the policy usually incorporates past history $h_t=a_0o_0r_0, \dots, a_{t-1}o_{t-1}r_{t-1}$. + +The action-value function $Q^\pi$ gives the estimated return when the agent has the state history $h_t$, executes action $a_t$ and follows the policy $\pi$ on the future time steps. It can be recursively defined by $Q^{\pi}(h_t, a_t) = \mathbb{E}_{s_{t+1},r_{t} \sim p}\big[r_{t}+\gamma \mathbb{E}_{a_{t+1} \sim \pi}[Q^\pi(h_{t+1}, a_{t+1})]\big]$. Q-learning and Deep Q-Networks (DQN) [@mnih2015human] are popular methods for obtaining the optimal action value function $Q^*$. Once we have $Q^*$, the optimal policy is also available as $\pi^*=\arg\max_{a_t} Q^*(h_t,a_t)$. In DQN, the action-value function is approximated by a deep NN with parameters $\theta$. $Q_\theta^*$ is found by minimising the loss function: $$\begin{equation} + \mathcal{L}_t(\theta) = \mathbb{E}_{h_t,a_t,r_t,h_{t+1}}[(y_t-Q_\theta^\pi(h_t, a_t))^2],\quad \text{where } y_t=r_t+\gamma \max_{a'}Q_\theta^\pi(a_{t+1}, h_{t+1}) +\end{equation}$$ Where $\pi$ is the $\epsilon$-greedy policy which takes action $\arg\max_{a_t} Q^\pi(a_t, h_t)$ with probability $1-\epsilon$, and takes a random action with probability $\epsilon$. + +In this work, we consider the MARL setting where the underlying environment is modelled by a partially observable Markov game [@hansen2004dynamic]. In this setting, the environment is populated by multiple agents which have individual observations and rewards and act according to individual policies. Their goal is to maximise their own expected return. + +Our work builds on VDN [@sunehag2017value], which was designed to address the binary cooperative MARL setting. In this setting, the agents are grouped into teams and all the agents within a team receive the same reward. VDN's main assumption is that the joint action-value function of the whole team of cooperative agents can be additively decomposed into the action-value functions across the members of the team. $$\begin{equation} + \label{eq:vdn_team} + Q^\mathcal{T}((h_t^1, h_t^2, \dots, h_t^{|\mathcal{T}|}), (a_t^1, a_t^2, \dots, a_t^{|\mathcal{T}|})) \approx \sum_{i \in \mathcal{T}}\tilde{Q}^i(h_t^i,a_t^i) +\end{equation}$$ Where $\mathcal{T}$ is the set of agents belonging to the team, and $\tilde{Q}^i(h_t^i,a_t^i)$ is the value function of agent $i$ which depends solely on its partial observation of the environment and its action at time $t$. $\tilde{Q}^i$ are trained by back-propagating gradients from the Q-learning rule through the summation. $$\begin{equation} + g_i = \nabla \theta_i(y^\mathcal{T}_t-\sum_{i \in \mathcal{T}}\tilde{Q}(h^i_t,a^i_t|\theta_i))^2, \qquad + y^\mathcal{T}_t = r^\mathcal{T}_t+\gamma \sum_{i \in \mathcal{T}}\max_{a^i_{t+1}}\tilde{Q}(h^i_{t+1},a^i_{t+1}|\theta_i) +\end{equation}$$ Where $\theta_i$ are the parameters of $\tilde{Q}^i$, $g_i$ is its gradient and $r^\mathcal{T}_t$ is the reward for the team $\mathcal{T}$ at the time instant $t$. Note that even though the training process is centralised, the learned agents can be deployed independently, since each agent acting greedily with respect to its own $\tilde{Q}^i$ will also maximise its team value function $\arg\max_{a^i_t} Q^{\mathcal{T}}_t(\dots) \approx \arg\max_{a^i_t} \tilde{Q}^i(h^i_t, a^i_t)$. + +In this section, we propose a reward function that enables RL algorithms to search for policies with increasingly evolutionary success. We call this reward the evolutionary reward because it is always aligned with the fitness function (see section [1.2](#sec:twoloops){reference-type="ref" reference="sec:twoloops"} for a definition of this alignment), and therefore, we don't need to go through the expensive process of aligning the agents' reward functions through evolution. We also propose a specific RL algorithm that is particularly suited to maximise the evolutionary reward in open-ended evolutionary environments (however other RL algorithms could also be used). + +The evolutionary reward of an agent is proportional to the number of copies its replicators have in the world's population. Maximising this reward leads to the maximisation of the survival and reproduction success of the replicators an agent carries. We start by defining the kinship function between a pair of agents $i$ and $j$, who carry $N$ replicators represented by the integer vectors $\bm{g}^i$ and $\bm{g}^j$ (we chose to use $\bm{g}$ for genome, which in biology is the set of genes (replicators) an agent carries): $$\begin{equation} + \label{def:kinship} + %\begin{split} + %k\colon \mathbb{Z}^N\times \mathbb{Z}^N \to [0, 1]&,\quad (\bm{g}^i, \bm{g}^j) \mapsto k(\bm{g}^i, \bm{g}^j)\\ + k\colon \mathbb{Z}^N\times \mathbb{Z}^N \to [0, 1], \qquad + k(\bm{g}^i, \bm{g}^j) = \frac{1}{N}\sum_{p=1}^N \delta_{g^i_p, g^j_p} + %\end{split} +\end{equation}$$ Where $\delta_{g^i_p, g^j_p}$ is the Kronecker delta which is one if $g^i_p=g^j_p$ and zero otherwise . When agent $i$ is alive at time $t+1$, it receives the reward: $$\begin{equation} + \label{eq:reward} + r_t^i = \sum_{j \in \mathcal{A}_{t+1}}{k(\bm{g}^i, \bm{g}^j)} +\end{equation}$$ Where $\mathcal{A}_{t+1}$ is the set of agents alive at the instant $t+1$. Note that since agent $i$ is alive at $t+1$, $\mathcal{A}_{t+1}$ includes agent $i$. $T^i-1$ is the last time step that agent $i$ is alive and so, at this instant, the agent receives its final reward which is proportional to the discounted sum of the number of times its genes will be present on other agents after its death: $$\begin{equation} + \label{eq:final_reward} + r_{T^i-1}^i = \sum_{t=T^i}^\infty \gamma^{t-T^i}\sum_{j \in \mathcal{A}_t}{k(\bm{g}^i, \bm{g}^j)} +\end{equation}$$ + +With this reward function, the agents are incentivised to maximise the survival and replication success of the replicators they carry. In the agent-centered view, the agents are incentivised to survive and replicate, but also to help their family (kin) survive and replicate; and to make sure that when they die their family is in a good position to carry on surviving and replicating. + +The discount factor, $\gamma$, needs to be in the interval $[0, 1[$ to ensure the final reward remains bounded. Due to the exponential discounting we can compute the final reward up to an error of $\epsilon$ by summing over a finite period of time denoted by the effective horizon ($h_e$). To see how to compute the $h_e$ for a given environment and $\epsilon$ see the appendix [10.1](#sec:he){reference-type="ref" reference="sec:he"}. We can now use Q-learning to train agents with this evolutionary reward. However, in the next section we introduce a more practical algorithm that allows us to estimate the final reward without having to simulate the environment forward for $h_e$ iterations. + +We propose Evolutionary Value-Decomposition Networks (E-VDN) as an extension of VDN (explained in detail in section [3.0.0.4](#sec:vdn){reference-type="ref" reference="sec:vdn"}) from the binary cooperative setting with static teams to the continuous cooperative setting with dynamic families. E-VDN helps us reduce the variance of the value estimation and allows us to estimate the final evolutionary reward without having to simulate the environment forward for $h_e$ iterations. + +Within a team, each agent fully cooperates with all the other members of the team, and it does not cooperate at all with any agent outside of the team. Moreover, if $a$ and $b$ are members of the same team and $c$ is a member of $a$'s team then $c$ and $b$ are also in the same team. Within a family, the degrees of cooperation amongst its members depends on their kinship degree (which can be any real number from 0 to 1). Also, if $a$ and $b$ are members of the same family and $c$ is part of $a$'s family, $c$ is not necessarily part of $b$'s family. + +Each agent $i$ sees the members of its family from an unique perspective, based on the kinship degree it shares with them. In E-VDN, each agent $i$ has a joint action-value function, $Q^i$. E-VDN assumes $Q^i$ can be composed by averaging the action-value functions across the members of $i$'s family weighted by their kinship with agent $i$ (this extends VDN's assumption, given by [\[eq:vdn_team\]](#eq:vdn_team){reference-type="eqref" reference="eq:vdn_team"}, to the continuous cooperative setting): $$\begin{equation} + \label{eq:maeq} + Q^i((h_t^1, h_t^2, \dots, h_t^{|\mathcal{A}_t|}), (a_t^1, a_t^2, \dots, a_t^{|\mathcal{A}_t|})) \approx \frac{1}{n^i_t}\sum_{j \in \mathcal{A}_t}k(\bm{g}^i,\bm{g}^j)\tilde{Q}^j(h_t^j,a_t^j) +\end{equation}$$ Where $n^i_t$ is a normalisation coefficient defined as $n^i_t=\sum_{j \in \mathcal{A}_t}k(\bm{g}^i,\bm{g}^j)$. Composing $Q^i$ with an average, instead of a sum, is necessary as E-VDN allows the number of value functions contributing to the composition to vary as the family gets bigger or smaller (agents born and die). This averaging allows us to incorporate the local observations of each family member and reduce variance in the value estimation. + +More importantly, E-VDN allows us to deal with the difficulty of estimating the terminal reward [\[eq:final_reward\]](#eq:final_reward){reference-type="eqref" reference="eq:final_reward"} in a particularly convenient way. As is clear from its definition [\[eq:final_reward\]](#eq:final_reward){reference-type="eqref" reference="eq:final_reward"}, the terminal reward is the expected sum (over time) of kinship that agent $i$ has with other agents $j$ after its death. The key idea is to note that this value ($r^i_{T^i-1}$) can be approximated by the Q-value of other agents $j$ that are close to (have high kinship with) agent $i$: $$\begin{equation} + \hat{r}^i_{T^i-1} = \left\{ + \begin{array}{ll} + \frac{1}{n^i_{T^i}}\sum_{j\in \mathcal{A}_{T^i}}k(\bm{g}^i,\bm{g}^j)\tilde{Q}^j_{T^i}(\dots)\approx Q^i_{T^i}(\dots) & \mbox{if } n^i_{T^i} > 0 \\ + 0 & \mbox{if } n^i_{T^i} = 0 + \end{array} + \right. \label{eq:final_rew_estimate} +\end{equation}$$ The final reward is zero if, and only if, at the time of its death the agent has no surviving family. + +Each $\tilde{Q}^i$ is trained by back-propagating gradients from the Q-learning rule: $$\begin{equation} + \label{eq:gradient} + g^i_t = \nabla \bm{\theta_i}(y^i_t-\frac{1}{n^i_t}\sum_{j \in \mathcal{A}_t}k(\bm{g}^i,\bm{g}^j)\tilde{Q}^j(h^j_t, a^j_t|\tilde{\theta}_j))^2 \approx \nabla \bm{\theta_i}(y^i_t-Q^i_t(\dots|\bm{\theta_i}))^2 +\end{equation}$$ Where $\bm{\theta}_i$ is the concatenation of all the parameters $\tilde{\theta}_j$, used in each $\tilde{Q}^j$, contributing to the estimation of $Q^i$; i.e. $\bm{\theta}_i \coloneqq \{ \tilde{\theta}_j \}_{j~\text{s.t.}~k(\bm{g}^i, \bm{g}^j) > 0}$. Note that $\tilde{Q}^i$ are neural networks with parameters $\tilde{\theta}_i$ and $Q^i$ is simply the average stated in [\[eq:maeq\]](#eq:maeq){reference-type="eqref" reference="eq:maeq"}. + +The learning targets $y^i_t$ are given by: $$\begin{align} + y^i_t &= \left\{ + \begin{array}{ll} + r^i_t+\gamma \max_{\bm{a}_{t+1}}Q^i_{t+1}(\dots)) & \mbox{if } t < T^i-1 \\ + \hat{r}^i_{T^i-1} & \mbox{if } t = T^i-1 + \end{array} + \right. \label{eq:learning_targets} +\end{align}$$ $r^i_t$ is the evolutionary reward [\[eq:reward\]](#eq:reward){reference-type="eqref" reference="eq:reward"} and $\hat{r}^i_{T^i-1}$ is the estimate of the final evolutionary reward [\[eq:final_rew_estimate\]](#eq:final_rew_estimate){reference-type="eqref" reference="eq:final_rew_estimate"}. We don't use a replay buffer in our training (which is commonly used in DQN) due to the non-stationary of multi-agent environments (more about this in the appendix [10.2](#sec:replay_buffer){reference-type="ref" reference="sec:replay_buffer"}). + +Since the joint action-value $Q^i$ increases monotonically with increasing $\tilde{Q}^i$, an agent acting greedily with respect to its action-value function will also act greedily in respect to its family action-value function: $\arg\max_{a^i_t}Q^i_t(\dots) \approx \arg\max_{a^i_t}\tilde{Q}^i(h^i_t, a^i_t)$. + +# Method + +We want to find the number of iterations ($h_e$) that guarantee an error between the estimate of the final reward and the actual final reward to be less or equal than a given $\epsilon$, $|r^i_{T^i-1}-\hat{r}^i_{T^i-1}| \leq \epsilon$.\ +Remember that the final reward is given by: $$r^i_{T^i-1} = \sum_{t=T^i}^\infty \gamma^{t-T^i}\sum_{j \in \mathcal{A}_t}{k(\bm{g}^i, \bm{g}^j)}=\sum_{t'=0}^\infty \gamma^{t'}k^i_{t'}$$ + +Where $t'=t-T^i$ and $k^i_{t'} = \sum_{j\in \mathcal{A}_{t'}}k(\bm{g}^i, \bm{g}^j)$. The estimate of the final reward is computed with the following finite sum $\hat{r}^i_t=\sum_{t'=0}^{h_e-1} \gamma^{t'}k^i_{t'}$.\ +Note that $k^i_t$ is always positive so the error $r^i_{T^i-1}-\hat{r}^i_{T^i-1}$ is always positive as well. To find the $h_e$ that guarantees an error smaller or equal to epsilon we define $r_b$ as the upper bound of $k^i_t$ and ensure that the worst possible error is smaller or equal to epsilon: $$\begin{align} + \sum_{t'=0}^{\infty} \gamma^{t'}r_b-\sum_{t'=0}^{h_e-1} \gamma^{t'}r_b &\leq \epsilon\\ + \frac{r_b}{1-\gamma}-r_b\frac{1-\gamma^{h_e}}{1-\gamma} &\leq \epsilon\\ + \frac{r_b\gamma^{h_e}}{1-\gamma} &\leq \epsilon\\ + h_e\log{\gamma} &\leq \log{\frac{\epsilon(1-\gamma)}{r_b}}\\ + h_e &\leq \frac{\log{\frac{\epsilon(1-\gamma)}{r_b}}}{\log{\gamma}} +\end{align}$$ + +We go from (1) to (2) by using the known convergences of geometric series: $\sum_{k=0}^\infty ar^k = \frac{a}{1-r}$ and $\sum_{k=0}^{n-1} ar^k = a\frac{1-r^n}{1-r}$ for $r < 1$. Since $h_e$ needs to be a positive integer we take the ceil $h_e = \ceil*{\frac{\log{\frac{\epsilon(1-\gamma)}{r_b}}}{\log{\gamma}}}$ and note that this equation is only valid when $\frac{\epsilon(1-\gamma)}{r_b}<1$. For example, an environment that has the capacity to feed at most 100 agents has an $r_b=100$ (which is the best possible reward when the kinship between every agent is 1). If we use $\epsilon=0.1$ and $\gamma=0.9$ then $h_e=88$. + +When using Q-learning methods with DQN, as we are, it's common practice to use a replay buffer. The replay buffer stores the experiences ($s_t, a_t, r_t, s_{t+1}$) for multiple time steps $t$. When training, the algorithm randomly samples experiences from the replay buffer. This breaks the auto-correlation between the consecutive examples and makes the algorithm more stable and sample efficient. However, for non-stationary environments, past experiences might be outdated. For this reason, we don't use a replay buffer. Instead, we break the auto-correlations by collecting experiences from many independent environments being sampled in parallel. After a batch of experiences is used we discard them. In our experiments, we simulated 400 environments in parallel and collected one experience step from each agent at each environment to form a training batch. + +In some situations, we used a denser version of the evolutionary reward to speed up the training process. We call it the *sugary* reward, $r_t^{\prime i} = \sum_{j \in \mathcal{A}_t}{k(\bm{g}^i, \bm{g}^j)}f_t^j$ where $f_t^j$ is the food collected by agent $j$ at the time instant $t$. In these simple environments, the *sugary* and the evolutionary reward are almost equivalent since a family with more members will be able to collect more food and vice-versa. However, the *sugary* reward contains more immediate information whilst the evolutionary reward has a lag between good (bad) actions and high (low) rewards; a family that is not doing a good job at collecting food will take a while to see some of its members die from starvation. Nonetheless, the evolutionary reward is more correct since it describes exactly what we want to maximise. Note that this reward was not used to produce the results when comparing E-VDN with CMA-ES. + +When using the standard evolutionary reward to evolve the larger NNs, the same four eras, that were observed with the *sugary* reward, emerge. However, their progression is not as linear. In this case, the families take longer to learn and sometimes one family evolves much faster than the others. When this happens, the families left behind eventually catch up with the most developed ones. The behaviour of the emerging families successfully interferes with the developed ones creating a temporary disruption in the environment which disrupts its macro-statistics. Two disruptions were observed in one of our simulations and we named them the First and the Second Family Wars (fig. [5](#fig:standard_history){reference-type="ref" reference="fig:standard_history"}). + +![Macro-statistics when evolving bacteria using the standard evolutionary reward. I: learning to survive (blue line), II: learning to reproduce (orange), III: learning to detect kin (green), IV: learning to self-sacrifice (red). The red bands correspond to the First and Second Family Wars.](appendix/gene_history_1.png){#fig:standard_history} diff --git a/2102.05714/main_diagram/main_diagram.drawio b/2102.05714/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..c3773b90e58064b1ba1e50d5e13a927c853b1cde --- /dev/null +++ b/2102.05714/main_diagram/main_diagram.drawio @@ -0,0 +1 @@ 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 \ No newline at end of file diff --git a/2102.05714/paper_text/intro_method.md b/2102.05714/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..ca7415e96e19382de9c658ae59c21bf65ab3e368 --- /dev/null +++ b/2102.05714/paper_text/intro_method.md @@ -0,0 +1,57 @@ +# Introduction + +Deep reinforcement learning has been successful in a series of control problems, such as Atari 2600 video games [@mnih2013playing] and MuJoCo environments [@lillicrap2015continuous]. However, the advances of deep RL relies on a large amount of interactions with the environment. In addition, the policy tends to specialize to the training domain and fails to generalize to new domains even when these two domains are similar. It has been shown that slight visual changes on pixel-based observations from Atari games could cause the well trained policy totally break down [@gamrian2019transfer]. These two limitations make deep reinforcement learning algorithms inefficient when applied to sets of tasks. As a result, efficient domain adaptation approaches are important for the applicability of Deep RL. + +Although state-of-the-art methods have demonstrated compelling performance in domain adaptation in RL, these approaches all have their limitations. Domain randomization [@tobin2017domain; @andrychowicz2020learning; @slaoui2020robust] relies on the availability of multiple source domains for training and cannot be applied in one-to-many generalization scenarios. Image-to-image translation approaches [@pan2017virtual; @tzeng2020adapting; @gamrian2019transfer] need a computationally expensive generator model for image translation. The extra burden on computation brought by the generator model is impractical for real-time applications such as autonomous driving. Other approaches utilize the latent embedding of encoder-decoder models to extract internal state representation for better generalization [@higgins2017darla]. However, domain-specific variations are also compressed into the latent embedding which could be problematic for zero-shot policy transfer. + +To solve the problem of domain adaptation across related RL tasks and avoid limitations of prior methods, we propose to learn a latent unified state representation (LUSR) for different domains and then train RL agents in the source domain based on that. After the RL training, zero-shot policy transfer is evaluated in target domains. To learn LUSR, we split the latent state representation into domain-general embedding which contains information existing in all domains and domain-specific embedding that compress domain specific information. LUSR is composed of domain-general embedding only and thus is able to ignore domain-specific variations and generalize across domains. + +To empirically justify our approach, we conducted experiments in two car driving tasks with different visual complexity. We first applied our approach in CarRacing games with analysis of final domain adaptation performance, domain adaptation performance across the training period, generalization to totally unseen domains and policy explanation with saliency maps. Then we evaluated our approach in autonomous driving tasks in CARLA simulator [@Dosovitskiy17] with more challenging and realistic visual observations. + +In comparison with other approaches, LUSR does not need RL training in multiple source domains like domain randomization and thus is applicable to a wider range of tasks. In addition, LUSR does not need computationally expensive generator models and can achieve better training efficiency compared with image-to-image translation approaches that operate in pixel-space. Finally, in contrast with other approaches that use latent state representation, LUSR filters out the factors of variation across domains and ensures the latent state representation is unified across all domains. + +# Method + +Our approach focuses on learning a latent unified state representation (LUSR) for states from different domains in RL. In this section, we first introduce the definition of LUSR and then introduce how to learn it. + +We first introduce two notions for state space in RL which are the agent's raw observation state space $S^o$ and the agent's internal latent state space $S^z$. Raw observation states $s^o$ consists of a grid of pixels while each unit in the internal latent state $s^z$ represents a high level semantic feature. A mapping function $\mathcal{F}: S^o \to S^z$ maps the observation state to the corresponding internal latent state. In our work, high level semantic features in $S^z$ are further divided into domain-specific ones (such as weather conditions in the driving task) and domain-general ones (such as vehicle dynamics). Here we denote $S^z = (\widehat{S^z}, \overline{S^z})$ where $\widehat{S^z}$ represents domain-specific features and $\overline{S^z}$ represents domain-general features. For state representation in source and target domains, this is summarized as + +$$\begin{equation} +\begin{aligned} + S^o_S & \neq S^o_T \\ + S^z_S = (\widehat{S^z_S}, \overline{S^z_S}); & \quad S^z_T = (\widehat{S^z_T}, \overline{S^z_T}) \\ + \overline{S^z_S} = \overline{S^z_T}; & \quad \widehat{S^z_S} \neq \widehat{S^z_T} +\end{aligned} +\end{equation}$$ + +In our setting of domain adaptation, the transition function $T$ and reward function $R$ only depend on $\overline{S^z}$ which is consistent across domains. Here we define the reward and transition function that take $s^o$ as input as $R^o$ and $T^o$ while the reward and transition function that take $s^z$ as input as $R^z$ and $T^z$. Then, we have + +$$\begin{equation} +\begin{aligned} + {T^o_S} \neq {T^o_T}; &\quad {R^o_S} \neq {R^o_T} \\ + {T^z_S} = T(\overline{S^z_S}) & = T(\overline{S^z_T}) = {T^z_T} \\ + {R^z_S} = R(\overline{S^z_S}) & = R(\overline{S^z_T}) = {R^z_T} \\ +\end{aligned} +\end{equation}$$ + +Since $\overline{S^z}$ is consistent across domains and the reward structure ($T$ and $R$) depend only on this representation (not on $\widehat{S^z}$), the RL agent taking $\overline{S^z}$ as input will be able to be trained successfully and the trained agent also has the capability to adapt from the source domain to target domains. As a result, the goal of our approach is learning the mapping function $\mathcal{F}: S^o \to \overline{S^z}$ that maps raw observation states to the latent unified state representation which we call LUSR. + +In this work, we choose to learn the mapping function $\mathcal{F}: S^o \to \overline{S^z}$ via Cycle-Consistent VAE [@jha2018disentangling] which is a non-adversarial approach to disentangle domain-general and domain-specific factors of variation. Similar to VAE [@kingma2013auto], Cycle-Consistent VAE is also composed of an encoder and a decoder. However, the output from the encoder is split into domain-general and domain-specific embeddings. To learn the mapping function $\mathcal{F}$, a number of random observation states from a set of pre-defined domains are first collected and then used as input for Cycle-Consistent VAE model training. Once the model is trained, the encoder is able to map observation states $s^o$ from any domain in the domain set to a latent state representation composed of $\overline{s^z}$ and $\widehat{s^z}$. As a result, we use the trained encoder as our mapping function $\mathcal{F}$ and keep only domain-general representation as LUSR. + +Cycle-Consistent VAE is based on the idea of cycle consistency whose intuition is that two well trained forward and reverse transformations composed together *in any order* should approximate an identity function. For example, in the VAE, the encoder is a forward transformation that converts an input image to a latent vector while the decoder is the reverse transformation that converts the latent vector back to a reconstructed image. Here we define the forward cycle as: $Dec(Enc(s^o)) = s^{o}\prime$ and the reverse cycle as $Enc(Dec(\widehat{s^z}, \overline{s^z})) =(\widehat{s^z \prime}, \overline{s^z \prime})$. As indicated by the cycle consistency, $s^{o}\prime$ should be close to $s^{o}$ and also $(\widehat{s^z \prime}, \overline{s^z \prime})$ should be close to $(\widehat{s^z}, \overline{s^z})$. + +In the forward cycle of Cycle-Consistent VAE, for two observation states $s^o_1$, $s^o_2$ from the same domain, $Enc(s^o_1) = \widehat{s^z_1}, \overline{s^z_1}$ and $Enc(s^o_2) = \widehat{s^z_2}, \overline{s^z_2}$. Since both originate from the same domain and $\widehat{s^z}$ contains only domain-specific information, swapping $\widehat{s^z_1}$ and $\widehat{s^z_2}$ should have no effect on the reconstruction loss which means we should get $Dec(\widehat{s^z_2}, \overline{s^z_1}) \approx s^o_1$ and $Dec(\widehat{s^z_1}, \overline{s^z_2}) \approx s^o_2$. This operation ensures that domain-specific information and domain-general information are compressed into $\widehat{s^z}$ and $\overline{s^z}$ separately. + +In the reverse cycle, a randomly sampled $\overline{s^z}$ is passed through the decoder in combination with two domain-specific embeddings $\widehat{s^z_1}$ and $\widehat{s^z_2}$ to obtain two reconstructed images $s^o_1\prime$ and $s^o_2\prime$. Since both $s^o_1\prime$ and $s^o_2\prime$ are generated based on the same $\overline{s^z}$, their corresponding domain-general latent embedding $\overline{s^z_1}\prime$ and $\overline{s^z_2}\prime$ should also be the same. + +As a result, the objective for Cycle-Consistent VAE to minimize is $$\begin{equation} +\mathcal{L}_{cyclic} = \mathcal{L}_{forward} + \mathcal{L}_{reverse} +\end{equation}$$ where $$\begin{equation*} +\begin{aligned} +% \mathcal{L}_{forward} = & -\mathbb{E}_{q_\phi(\overline{s^z}, \widehat{s^z} \mid s^o)} [\log p_\theta(s^o \vert \overline{s^z}, \widehat{s^z\ast})] \\ & + KL(q_\phi(s^z \vert s^o) || p(s^z)) \\ +\mathcal{L}_{forward} = & -\mathbb{E}_{q_\phi(\overline{s^z}, \widehat{s^z} \mid s^o)} [\log p_\theta(s^o \vert \overline{s^z}, \widehat{s^z\ast})] \\ & + KL(q_\phi(\overline{s^z} \vert s^o) || p(\overline{s^z})) \\ +\mathcal{L}_{reverse} = & \mathbb{E}_{\overline{s^z} \sim p(\overline{s^z})} [|| \overline{ q_{\phi}} (p_{\theta}(\overline{s^z}, \widehat{s^z_1})) - \overline{q_{\phi}} (p_{\theta}(\overline{s^z}, \widehat{s^z_2})) ||_1] +\end{aligned} +\end{equation*}$$ + +$\mathcal{L}_{forward}$ here is a modified variational upper-bound and $\mathcal{L}_{reverse}$ is the loss for cycle consistency. $q_\phi$ and $p_\theta$ are parameterized functions of the encoder and decoder. We define $\overline{q_\phi}$ as $q_\phi$ that only keeps the domain general embedding as output. The latent embedding $s^z$ is composed of $\overline{s^z}$ and $\widehat{s^z}$ which are domain-general and domain-specific latent embeddings corresponding to observation state $s^o$. $\widehat{s^z\ast}$ represents any random domain-specific embedding from the same domain while $\widehat{s^z_1}$ and $\widehat{s^z_2}$ are two different domain-specific embeddings. diff --git a/2104.07186/main_diagram/main_diagram.drawio b/2104.07186/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..60dd57c66265018c90c5758aae6fe0c7cb008278 --- /dev/null +++ b/2104.07186/main_diagram/main_diagram.drawio @@ -0,0 +1 @@ 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 \ No newline at end of file diff --git a/2104.07186/main_diagram/main_diagram.pdf b/2104.07186/main_diagram/main_diagram.pdf new file mode 100644 index 0000000000000000000000000000000000000000..b8c428a643e87caa7b5aba46df334b42eabb1ea0 Binary files /dev/null and b/2104.07186/main_diagram/main_diagram.pdf differ diff --git a/2104.07186/paper_text/intro_method.md b/2104.07186/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..382cb6639949ce6d515a2286075a0e37db9e6455 --- /dev/null +++ b/2104.07186/paper_text/intro_method.md @@ -0,0 +1,128 @@ +# Introduction + +Widely used, bag-of-words (BOW) information retrieval (IR) systems such as BM25 rely on exact lexical match [2](#page-0-1) between query and document terms. Recent study in neural IR takes a different approach and compute soft matching between all query and document terms to model complex matching. + +The shift to soft matching in neural IR models attempts to address *vocabulary mismatch* problems, that query and the relevant documents use different terms, e.g. cat v.s. kitty, for the same concept [\(Huang et al.,](#page-9-0) [2013;](#page-9-0) [Guo et al.,](#page-9-1) [2016;](#page-9-1) [Xiong](#page-10-0) [et al.,](#page-10-0) [2017\)](#page-10-0). Later introduction of contextualized representations [\(Peters et al.,](#page-10-1) [2018\)](#page-10-1) from deep language models (LM) further address *semantic mismatch*, that the same term can refer to different concepts, e.g., bank of river vs. bank in finance. Fine-tuned deep LM rerankers produce token representations based on context and achieve state-of- + +the-art in text ranking with huge performance leap [\(Nogueira and Cho,](#page-10-2) [2019;](#page-10-2) [Dai and Callan,](#page-9-2) [2019b\)](#page-9-2). + +Though the idea of soft matching all tokens is carried through the development of neural IR models, seeing the success brought by deep LMs, we take a step back and ask: how much gain can we get if we introduce contextualized representations back to lexical exact match systems? In other words, can we build a system that still performs exact querydocument token matching but compute matching signals with contextualized token representations instead of heuristics? This may seem a constraint on the model, but exact lexical match produce more explainable and controlled patterns than soft matching. It also allows search to focus on only the subset of documents that have overlapping terms with query, which can be done efficiently with inverted list index. Meanwhile, using dense contextualized token representations enables the model to handle semantic mismatch, which has been a long-standing problem in classic lexical systems. + +To answer the question, we propose a new lexical matching scheme that uses vector similarities between query-document overlapping term contextualized representations to replace heuristic scoring used in classical systems. We present COntextualized Inverted List (COIL), a new exact lexical match retrieval architecture armed with deep LM representations. COIL processes documents with deep LM offline and produces representations for each document token. The representations are grouped by their surface tokens into inverted lists. At search time, we build representation vectors for query tokens and perform contextualized exact match: use each query token to look up *its own* inverted list and compute vector similarity with document vectors stored in the inverted list as matching scores. COIL enables efficient search with rich-in-semantic matching between query and document. + +Our contributions include 1) introduce a novel + +1Our code is available at [https://github.com/](https://github.com/luyug/COIL) [luyug/COIL](https://github.com/luyug/COIL). + +2Exact match up to morphological changes. + +retrieval architecture, *contextualized inverted lists* (COIL) that brings semantic matching into lexical IR systems, 2) show matching signals induced from exact lexical match can capture complicated matching patterns, 3) demonstrate COIL significantly outperform classical and deep LM augmented lexical retrievers as well as state-of-theart dense retrievers on two retrieval tasks. + +# Method + +In this section, we first provide some preliminaries on exact lexical match systems. Then we discuss COIL's contextualized exact match design and how its search index is organized. We also give a comparison between COIL and other popular retrievers. + +![](_page_2_Figure_7.jpeg) + +Figure 3: COIL's index and retrieval architecture. COIL-tok relies on the exact token matching (lower). COIL-full includes in addition CLS matching (upper). + +Classic lexical retrieval system relies on *overlapping* query document terms under morphological generalization like stemming, in other words, *exact lexical match*, to score query document pair. A scoring function is defined as a sum of matched term scores. The scores are usually based on statistics like term frequency (*tf*). Generally, we can write, + + +$$s = \sum_{t \in q \cap d} \sigma_t(h_q(q, t), h_d(d, t)) \tag{1}$$ + +where for each overlapping term t between query q and document d, functions $h_q$ and $h_d$ extract term + +information and a term scoring function $\sigma_t$ combines them. A popular example is BM25, which computes, + +$$s_{\text{BM25}} = \sum_{t \in q \cap d} idf(t) h_q^{\text{BM25}}(q, t) h_d^{\text{BM25}}(d, t)$$ + +$$h_q^{\text{BM25}}(q, t) = \frac{t f_{t,q} (1 + k_2)}{t f_{t,q} + k_2}$$ + +$$h_d^{\text{BM25}}(d, t) = \frac{t f_{t,d} (1 + k_1)}{t f_{t,d} + k_1 (1 - b + b \frac{|d|}{\text{avad}})}$$ +(2) + +where $tf_{t,d}$ refers to term frequency of term t in document d, $tf_{t,q}$ refers to the term frequency in query, idf(t) is inverse document frequency, and b, $k_1$ , $k_2$ are hyper-parameters. + +One key advantage of exact lexical match systems lies in efficiency. With summation over exact matches, scoring of each query term only goes to documents that contain matching terms. This can be done efficiently using *inverted list* indexing (Figure 2). The inverted list maps back from a term to a list of documents where the term occurs. To compute Equation 1, the retriever only needs to traverse the subset of documents in query terms' inverted lists instead of going over the entire document collection. + +While recent neural IR research mainly focuses on breaking the exact match bottleneck with soft matching of text, we hypothesize that exact match itself can be improved by replacing semantic independent frequency-based scoring with semantic rich scoring. In the rest of this section, we show how to modify the exact lexical match framework with contextualized term representations to build effective and efficient retrieval systems. + +Instead of term frequency, we desire to encode the semantics of terms to facilitate more effective matching. Inspired by recent advancements in deep LM, we encode *both* query and document tokens into contextualized vector representations and carry out matching between exact lexical matched tokens. Figure 1d illustrates the scoring model of COIL. + +In this work, we use a Transformer language $model^3$ as the contextualization function. We encode a query q with the language model (LM) and represent its i-th token by projecting the corresponding output: + +$$\boldsymbol{v}_{i}^{q} = \boldsymbol{W}_{tok} LM(q, i) + \boldsymbol{b}_{tok} \tag{3}$$ + +where $W_{tok}^{n_t \times n_{lm}}$ is a matrix that maps the LM's $n_{lm}$ dimension output into a vector of lower dimension $n_t$ . We down project the vectors as we hypothesize that it suffices to use lower dimension token vectors. We confirm this in section 5. Similarly, we encode a document d's j-th token $d_j$ with: + +$$\boldsymbol{v}_{i}^{d} = \boldsymbol{W}_{tok} LM(d, j) + \boldsymbol{b}_{tok}$$ + (4) + +We then define the *contextualized exact lexical match scoring function* between query document based on vector similarities between exact matched query document token pairs: + + +$$s_{\text{tok}}(q, d) = \sum_{q_i \in q \cap d} \max_{d_j = q_i} (\boldsymbol{v}_i^{q \mathsf{T}} \boldsymbol{v}_j^d)$$ + (5) + +Note that, importantly, the summation goes through only overlapping terms, $q_i \in q \cap d$ . For each query token $q_i$ , we finds all *same* tokens $d_j$ in the document, computes their similarity with $q_i$ using the *contextualized* token vectors. The maximum similarities are picked for query token $q_i$ . Max operator is adopted to capture the most important signal (Kim, 2014). This fits in the general lexical match formulation, with $h_q$ giving representation for $q_i$ , $h_t$ giving representations for all $d_j = q_i$ , and $\sigma_t$ compute dot similarities between query vector with document vectors and max pool the scores. + +As with classic lexical systems, $s_{tok}$ defined in Equation 5 does not take into account similarities between lexical-different terms, thus faces vocabulary mismatch. Many popular LMs (Devlin et al., 2019; Yang et al., 2019; Liu et al., 2019) use a special CLS token to aggregate sequence representation. We project the CLS vectos with $\boldsymbol{W}_{cls}^{n_c \times n_{lm}}$ to represent the entire query or document, + +$$egin{aligned} oldsymbol{v}_{cls}^q &= oldsymbol{W}_{cls} \mathrm{LM}(q, \mathrm{CLS}) + oldsymbol{b}_{cls} \ oldsymbol{v}_{cls}^d &= oldsymbol{W}_{cls} \mathrm{LM}(d, \mathrm{CLS}) + oldsymbol{b}_{cls} \end{aligned} \tag{6}$$ + +The similarity between $v_{cls}^q$ and $v_{cls}^d$ provides highlevel semantic matching and mitigates the issue of vocabulary mismatch. The full form of COIL is: + +$$s_{\text{full}}(q, d) = s_{\text{tok}}(q, d) + \boldsymbol{v}_{cls}^{q} \boldsymbol{\tau} \boldsymbol{v}_{cls}^{d}$$ + (7) + +In the rest of the paper, we refer to systems with CLS matching **COIL-full** and without **COIL-tok**. + +COIL's scoring model (Figure 1d) is fully differentiable. Following earlier work (Karpukhin et al., 2020), we train COIL with negative log likelihood defined over query q, a positive document $d^+$ and a + +&lt;sup>3We used the base, uncased variant of BERT. + +set of negative documents $\{d_1^-, d_2^-, ... d_l^-..\}$ as loss. + +$$\mathcal{L} = -\log \frac{\exp(s(q, d^{+}))}{\exp(s(q, d^{+})) + \sum_{l} \exp(s(q, d_{l}^{-}))}$$ +(8) + +Following Karpukhin et al. (2020), we use in batch negatives and hard negatives generated by BM25. Details are discussed in implementation, section 4. + +COIL pre-computes the document representations and builds up a search index, which is illustrated in Figure 3. Documents in the collection are encoded offline into token and CLS vectors. Formally, for a unique token t in the vocabulary V, we collect its contextualized vectors from all of its mentions from documents in collection C, building token t's contextualized inverted list: + +$$I^{t} = \{ \mathbf{v}_{j}^{d} \mid d_{j} = t, d \in C \}, \tag{9}$$ + +where $\boldsymbol{v}_j^d$ is the BERT-based token encoding defined in Equation 4. We define search index to store inverted lists for all tokens in vocabulary, $\mathbb{I} = \{I^t \mid t \in V\}$ . For COIL-full, we also build an index for the CLS token $I^{cls} = \{\boldsymbol{v}_{cls}^d \mid d \in C\}$ . + +As shown in Figure 3, in this work we implement COIL's by stacking vectors in each inverted list $I^t$ into a matrix $M^{n_t \times |I^k|}$ , so that similarity computation that traverses an inverted list and computes vector dot product can be done efficiently as one matrix-vector product with optimized BLAS (Blackford et al., 2002) routines on CPU or GPU. All $v_{cls}^d$ vectors can also be organized in a similar fashion into matrix $M_{cls}$ and queried with matrix product. The matrix implementation here is an exhaustive approach that involves all vectors in an inverted list. As a collection of dense vectors, it is also possible to organize each inverted list as an approximate search index (Johnson et al., 2017; Guo et al., 2019) to further speed up search. + +When a query q comes in, we encode every of its token into vectors $\boldsymbol{v}_i^q$ . The vectors are sent to the subset of COIL inverted lists that corresponds query tokens $\mathbb{J}=\{I^t\mid t\in q\}$ . where the matrix product described above is carried out. This is efficient as $|\mathbb{J}|<<|\mathbb{I}|$ , having only a small subset of all inverted lists to be involved in search. For COIL-full, we also use encoded CLS vectors $\boldsymbol{v}_{cls}^q$ to query the CLS index to get the CLS matching scores. The scoring over different inverted lists can + +serve in parallel. The scores are then combined by Equation 5 to rank the documents. + +Readers can find detailed illustration figures in the Appendix A, for index building and querying, Figure 4 and Figure 5, respectively. + +Deep LM based Lexical Index Models like DeepCT (Dai and Callan, 2019a, 2020) and DocT5Query (Nogueira and Lin, 2019) alter $tf_{t,d}$ in documents with deep LM BERT or T5. This is similar to a COIL-tok with token dimension $n_t = 1$ . A single degree of freedom however measures more of a term *importance* than *semantic agreement*. + +**Dense Retriever** Dense retrievers (Figure 1b) are equivalent to COIL-full's CLS matching. COIL makes up for the lost token-level interactions in dense retriever with exact matching signals. + +**ColBERT** ColBERT (Figure 1c) computes relevance by soft matching *all* query and document term's contextualized vectors. + +$$s(q,d) = \sum_{q_i \in [cls:a:exv]} \max_{d_j \in [cls;d]} (\boldsymbol{v}_i^{q\mathsf{T}} \boldsymbol{v}_j^d) \qquad (10)$$ + +where interactions happen among query q, document d, cls and set of query expansion tokens exp. The all-to-all match contrasts COIL that only uses exact match. It requires a dense retrieval over all document tokens' representations as opposed to COIL which only considers query's overlapping tokens, and are therefore much more computationally expensive than COIL. + +Datasets We experiment with two large scale ad hoc retrieval benchmarks from the TREC 2019 Deep Learning (DL) shared task: MSMARCO passage (8M English passages of average length around 60 tokens) and MSMARCO document (3M English documents of average length around 900 tokens)4. For each, we train models with the MSMARCO Train queries, and record results on MSMARCO Dev queries and TREC DL 2019 test queries. We report mainly full-corpus retrieval results but also include the rerank task on MSMARCO Dev queries where we use neural scores to reorder BM25 retrieval results provided by MSMARO organizers. Official metrics include + +&lt;sup>4Both datasets can be downloaded from https://microsoft.github.io/msmarco/ + +MRR@1K and NDCG@10 on test and MRR@10 on MSMARCO Dev. We also report recall for the dev queries following prior work [\(Dai and Callan,](#page-9-8) [2019a;](#page-9-8) [Nogueira and Lin,](#page-10-10) [2019\)](#page-10-10). + +Compared Systems Baselines include 1) traditional exact match system BM25, 2) deep LM augmented BM25 systems DeepCT [\(Dai and Callan,](#page-9-8) [2019a\)](#page-9-8) and DocT5Query [\(Nogueira and Lin,](#page-10-10) [2019\)](#page-10-10), 3) dense retrievers, and 4) soft all-to-all retriever ColBERT. For DeepCT and DocT5Query, we use the rankings provided by the authors. For dense retrievers, we report two dense retrievers trained with BM25 negatives or with mixed BM25 and random negatives, published in [Xiong et al.](#page-10-11) [\(2020\)](#page-10-11). However since these systems use a robust version of BERT, RoBERTa [\(Liu et al.,](#page-10-14) [2019\)](#page-10-14) as the LM and train document retriever also on MSMARCO passage set, we in addition reproduce a third dense retriever, that uses the exact same training setup as COIL. All dense retrievers use 768 dimension embedding. For ColBERT, we report its published results (available only on passage collection). BERT reranker is added in the rerank task. + +We include 2 COIL systems: 1) COIL-tok, the exact token match only system, and 2) COLL-full, the model with both token match and CLS match. + +Implementation We build our models with Pytorch [\(Paszke et al.,](#page-10-15) [2019\)](#page-10-15) based on huggingface transformers [\(Wolf et al.,](#page-10-16) [2019\)](#page-10-16). COIL's LM is based on BERT's base variant. COIL systems use token dimension nt = 32 and COIL-full use CLS dimension nc = 768 as default, leading to 110M parameters. We add a Layer Normalization to CLS vector when useful. All models are trained for 5 epochs with AdamW optimizer, a learning rate of 3e-6, 0.1 warm-up ratio, and linear learning rate decay, which takes around 12 hours. Hard negatives are sampled from top 1000 BM25 results. Each query uses 1 positive and 7 hard negatives; each batch uses 8 queries on MSMARCO passage and 4 on MSMARCO document. Documents are truncated to the first 512 tokens to fit in BERT. We conduct validation on randomly selected 512 queries from corresponding train set. Latency numbers are measured on dual Xeon E5-2630 v3 for CPU and RTX 2080 ti for GPU. We implement COIL's inverted lists as matrices as described in [subsection 3.3,](#page-4-2) using NumPy [\(Harris et al.,](#page-9-19) [2020\)](#page-9-19) on CPU and Pytorch on GPU. We perform a) a set of matrix products to compute token similarities + +over contextualized inverted lists, b) scatter to map token scores back to documents, and c) sort to rank the documents. Illustration can be found in the appendix, [Figure 5.](#page-12-0) diff --git a/2106.01342/main_diagram/main_diagram.drawio b/2106.01342/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..ab9f17cbcd6cde422618a4825f4f2c753bc65071 --- /dev/null +++ b/2106.01342/main_diagram/main_diagram.drawio @@ -0,0 +1 @@ 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 \ No newline at end of file diff --git a/2106.01342/main_diagram/main_diagram.pdf b/2106.01342/main_diagram/main_diagram.pdf new file mode 100644 index 0000000000000000000000000000000000000000..9a828cd9f33e3d73b228d60a90ccc60cbfb7e62b Binary files /dev/null and b/2106.01342/main_diagram/main_diagram.pdf differ diff --git a/2106.01342/paper_text/intro_method.md b/2106.01342/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..ec2abf116b71e00876efdb5513ea1fc1bddc7250 --- /dev/null +++ b/2106.01342/paper_text/intro_method.md @@ -0,0 +1,72 @@ +# Introduction + +While machine learning for image and language processing has seen major advances over the last decade, many critical industries, including financial services, health care, and logistics, rely heavily on data in structured table format. Tabular data is unique in several ways that have prevented it from benefiting from the impressive success of deep learning in vision and language. First, tabular data often contain heterogeneous features that represent a mixture of continuous, categorical, and ordinal values, and these values can be independent or correlated. Second, there is no inherent positional information in tabular data, meaning that the order of columns is arbitrary. This differs from text, where tokens are always discrete, and ordering impacts semantic meaning. It also differs from images, where pixels are typically continuous, and nearby pixels are correlated. Tabular models must handle features from multiple discrete and continuous distributions, and they must discover correlations without relying on the positional information. Sufficiently powerful deep learning systems for tabular data have the potential to improve performance beyond what is achieved by classical methods, like linear classifiers and random forests. Furthermore, without performant deep learning models for tabular data, we lack the ability to exploit compositionality, end-to-end multi-task models, fusion with multiple modalities (e.g. image and text), and representation learning. + +We introduce SAINT, the Self-Attention and Intersample Attention Transformer, a specialized architecture for learning with tabular data. SAINT leverages several mechanisms to overcome the difficulties of training on tabular data. SAINT projects all features -- categorical and continuous -- into a combined dense vector space. These projected values are passed as tokens into a transformer encoder which uses attention in the following two ways. First, there is "self-attention," which attends to individual features within each data sample. Second, we propose a novel "intersample attention," which enhances the classification of a row (i.e., a data sample) by relating it to other rows in the table. Intersample attention is akin to a nearest-neighbor classification, where the distance metric is learned end-to-end rather than fixed. In addition to this hybrid attention mechanism, we also leverage self-supervised contrastive pre-training to boost performance for semi-supervised problems. + +We provide comparisons of SAINT to a wide variety of deep tabular architectures and commonly used tree-based methods using a diverse battery of datasets. We observe that SAINT, on average, outperforms all other methods on supervised and semi-supervised tasks. More importantly, SAINT often out-performs boosted trees (including XGBoost [@chen2016xgboost], CatBoost [@dorogush2018catboost], and LightGBM [@ke2017lightgbm]), which have long been an industry favorite for complex tabular datasets. Finally, we visualize the attention matrices produced by our models to gain insights into how they behave. + +# Method + +SAINT is inspired by the transformer encoder of @vaswani2017attention, designed for natural language, where the model takes in a sequence of feature embeddings and outputs contextual representations of the same dimension. A graphical overview of SAINT is presented in Figure [1](#fig:saint_arch_training){reference-type="ref" reference="fig:saint_arch_training"}(a). + +SAINT is composed of a stack of $L$ identical stages. Each stage consists of one self-attention transformer block and one intersample attention transformer block. The self-attention transformer block is identical to the encoder from [@vaswani2017attention]. It has a multi-head self-attention layer (MSA) (with $h$ heads), followed by two fully-connected feed-forward (FF) layers with a GELU non-linearity [@hendrycks2016gaussian]. Each layer has a skip connection [@he2016deep] and layer normalization (LN) [@ba2016layer]. The intersample attention transformer block is similar to the self-attention transformer block, except that the self-attention layer is replaced by an intersample attention layer (MISA). The details of the intersample attention layer are presented in the following subsection. + +The SAINT pipeline, with a single stage ($L=1$) and a batch of $b$ inputs, is described by the following equations. We denote multi-head self-attention by MSA, multi-head intersample attention by MISA, feed-forward layers by FF, and layer norm by LN: $$\begin{align} + \mathbf{z_i^{(1)}} &= \operatorname{LN}(\operatorname{MSA}(\mathbf{E}(\mathbf{x_i}))) + \mathbf{E}(\mathbf{x_i}) & \mathbf{z_i^{(2)}} &= \operatorname{LN}(\operatorname{FF_1}(\mathbf{z_i^{(1)}})) +\mathbf{z_i^{(1)}} + \label{eq:sa_block_apply} \\ + \mathbf{z_i^{(3)}} &= \operatorname{LN}(\operatorname{MISA}(\{\mathbf{z_i^{(2)}}\}_{i=1}^b)) +\mathbf{z_i^{(2)}} & \mathbf{r_i}&= \operatorname{LN}(\operatorname{FF_2}(\mathbf{z_i^{(3)}})) +\mathbf{z_i^{(3)}} + \label{eq:isa_block_apply} + % \mbf{y} &= \op{LN}(\mbf{z}_L^0) \label{eq:final_rep} +\end{align}$$ + +where $\mathbf{r_i}$ is SAINT's contextual representation output corresponding to data point $\mathbf{x_i}$. This contextual embedding can be used in downstream tasks such as self-supervision or classification. + +
+ +
Intersample attention on a batch of 3 points. In this plot, d is the size of value vectors vi. See Section 3.2 for details.
+
+ +We introduce intersample attention (a type of row attention) where the attention is computed across different data points (rows of a tabular data matrix) in a given batch rather than just the features of a single data point. Specifically, we concatenated the embeddings of each feature for a single data point, then compute attention over samples (rather than features). This enables us to improve the representation of a given point by inspecting other points. When a feature is missing or noisy in one row, intersample attention enables SAINT to borrow the corresponding features from other similar data samples in the batch. + +An illustration of how intersample attention is performed in a single head is shown in Figure  [2](#fig:ISA){reference-type="ref" reference="fig:ISA"} and the pseudo-code is presented in Algorithm [\[alg:MISA\]](#alg:MISA){reference-type="ref" reference="alg:MISA"}. Unlike the row attention used in [@ho2019axial; @child2019generating; @rao2021msa; @iida2021tabbie], intersample attention allows all features from different samples to communicate with each other. In our experiments, we show that this ability boosts performance appreciably. In the multi-head case, instead of projecting $q,k,v$ to a given dimension $d$, we project them to $d/h$ where $h$ is the number of heads. Then we concatenate all the updated value vectors, $v_i$, to get back a vector of length $d$. + +::: algorithm +``` {.python language="python"} + +# W_q, W_k, W_v are weight matrices of dimension dxd + # mm: matrix-matrix multiplication + def self_attention(x): + # x is bxnxd + q, k, v = mm(W_q,x), mm(W_k,x), mm(W_v,x) #q,k,v are bxnxd + attn = softmax(mm(q,np.transpose(k, (0, 2, 1)))/sqrt(d)) # bxnxn + out = mm(attn, v) #out is bxnxd + return out + + def intersample_attention(x): + # x is bxnxd + b,n,d = x.shape # as mentioned above + x = reshape(x, (1,b,n*d)) # reshape x to 1xbx(n*d) + x = self_attention(x) # the output x is 1xbx(n*d) + out = reshape(x,(b,n,d)) # out is bxnxd + return out +``` +::: + +Contrastive learning, in which models are pre-trained to be invariant to reordering, cropping, or other label-preserving "views" of the data  [@chen2020simple; @he2020momentum; @pathak2016context; @grill2020bootstrap; @vincent2008extracting], is a powerful tool in the vision and language domains that has never (to our knowledge) been applied to tabular data. We present a contrastive pipeline for tabular data, a visual description of which is shown in Figure [1](#fig:saint_arch_training){reference-type="ref" reference="fig:saint_arch_training"}. Existing self-supervised objectives for tabular data include denoising [@vincent2008extracting], a variation of which was used by VIME [@yoon2020vime], masking, and replaced token detection as used by TabTransformer [@huang2020tabtransformer]. We find that, while these methods are effective, superior results are achieved by contrastive learning. + +**Generating augmentations** Standard contrastive methods in vision craft different "views" of images using crops and flips. It is difficult to craft invariance transforms for tabular data. The authors of VIME [@yoon2020vime] use mixup in the non-embedded space as a data augmentation method, but this is limited to continuous data. We instead use CutMix [@yun2019cutmix] to augment samples in the input space and we use mixup [@zhang2017mixup] in the embedding space. These two augmentations combined yield a challenging and effective self-supervision task. Assume that only $l$ of $m$ data points are labeled. We denote the embedding layer by $\mathbf{E}$, the SAINT network by $\mathbf{S}$, and 2 projection heads as $g_1(\cdot)$ and $g_2(\cdot)$. The CutMix augmentation probability is denoted $p_{\text{cutmix}}$ and the mixup parameter is $\alpha$. Given point $\mathbf{x_i}$, the original embedding is $\mathbf{p_i} = \mathbf{E}(\mathbf{x_i})$, while the augmented representation is generated as follows: $$\begin{align} +\mathbf{x_i^\prime} &= \mathbf{x_i}\odot \mathbf{m} + \mathbf{x_a}\odot \mathbf{(1-m)} && \textcolor{gray}{\text{CutMix in raw data space}}\\ +\mathbf{p_i^\prime} &= \alpha*\mathbf{E}(\mathbf{x_i^\prime}) +(1-\alpha)*\mathbf{E}(\mathbf{x_b^\prime})&& \textcolor{gray}{\text{mixup in embedding space}} +\end{align}$$ where $\mathbf{x_a}$, $\mathbf{x_b}$ are random samples from the current batch, $\mathbf{x_b^\prime}$ is the CutMix version of $\mathbf{x_b}$, $\mathbf{m}$ is the binary mask vector sampled from a Bernoulli distribution with probability $p_{\text{cutmix}}$, and $\alpha$ is the mixup parameter. Note that we first obtain a CutMix version of every data point in a batch by randomly selecting a partner to mix with. We then embed the samples and choose new partners before performing mixup. + +**SAINT and projection heads** Now that we have both the clean $\mathbf{p_i}$ and mixed $\mathbf{p_i^\prime}$ embeddings, we pass them through SAINT, then through two projection heads, each consisting of an MLP with one hidden layer and a ReLU. The use of a projection head to reduce dimensionality before computing contrastive loss is common in vision [@chen2020simple; @he2020momentum; @grill2020bootstrap] and indeed also improves results on tabular data. Ablation studies and further discussion are available in Appendix [\[appendix_sec:analysis\]](#appendix_sec:analysis){reference-type="ref" reference="appendix_sec:analysis"}. + +**Loss functions** We consider two losses for the pre-training phase. (i) The first is a contrastive loss that pushes the latent representations of two views of the same data point ($z_i$ and $z_i^\prime$) close together and encourages different points ($z_i$ and $z_j$, $i\neq j$) to lie far apart. For this, we borrow the InfoNCE loss from metric-learning works [@sohn2016improved; @oord2018representation; @chen2020simple; @wu2018unsupervised]; (ii) The second loss comes from a denoising task. For denoising, we try to predict the original data sample from a noisy view. Formally, we are given $\mathbf{r_i^\prime}$ and we reconstruct the inputs as $\mathbf{x_i^{\prime \prime}}$ to minimize the difference between the original and the reconstruction. The combined pre-training loss is: $$\begin{align} +% \mathcal{L_\text{contrastive}} &= -\sum_{i=1}^{m}{\log{\frac{\exp(z_i \cdot z_i^\prime/\tau)}{\sum_{k=1}^{m}{\exp(z_i \cdot z_k^\prime/\tau)}}}}\\ +% \mathcal{L_\text{denoising}} &= \sum_{i=1}^{m}\sum_{j=1}^{n} [L_j(MLP_j(\mathbf{r_i^\prime}),\mathbf{x_i})] \\ +% \mathcal{L_\text{pretraining}} &= \mathcal{L_\text{contrastive}} + \lambda_{pt} \mathcal{L_\text{denoising}}\\ +\mathcal{L_\text{pre-training}} &= \underbrace{-\sum_{i=1}^{m}{\log{\frac{\exp(z_i \cdot z_i^\prime/\tau)}{\sum_{k=1}^{m}{\exp(z_i \cdot z_k^\prime/\tau)}}}}}_{\text{Contrastive Loss}} + \lambda_{\text{pt}} \underbrace{\sum_{i=1}^{m}\sum_{j=1}^{n} [\mathcal{L}_j(\text{MLP}_j(\mathbf{r_i^\prime}),\mathbf{x_i})]}_{\text{Denoising Loss}} +\end{align}$$ where $\mathbf{r_i} = \mathbf{S}(\mathbf{p_i}), \mathbf{r_i^\prime} = \mathbf{S}(\mathbf{p_i^\prime}), z_i = g_1(\mathbf{r_i}),z_i^\prime = g_2(\mathbf{r_i^\prime})$. $\mathcal{L}_j$ is cross-entropy loss or mean squared error depending on the $j^{th}$ feature being categorical or continuous. Each $\text{MLP}_j$ is a single hidden layer perceptron with a ReLU non-linearity. There are $n$ in number, one for each input feature. $\lambda_{\text{pt}}$ is a hyper-parameter and $\tau$ is temperature parameter and both of these are tuned using validation data. + +**Finetuning** Once SAINT is pre-trained on all unlabeled data, we finetune the model on the target prediction task using the $l$ labeled samples. The pipeline of this step is shown in Figure [1](#fig:saint_arch_training){reference-type="ref" reference="fig:saint_arch_training"}(b). For a given point $\mathbf{x_i}$, we learn the contextual embedding $\mathbf{r_i}$. For the final prediction step, we pass the embedding corresponding only to the `[CLS]` token through a simple MLP with a single hidden layer with ReLU activation to get the final output. We evaluate cross-entropy loss on the outputs for classification tasks and mean squared error for regression tasks. diff --git a/2109.02832/main_diagram/main_diagram.drawio b/2109.02832/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..32e1dd09034b55d6b132a614720e05ec61485650 --- /dev/null +++ b/2109.02832/main_diagram/main_diagram.drawio @@ -0,0 +1 @@ +7V1bc5s4FP41ntl9iAeQuD3G2Gm723TTprtN9yUjg2KzxUABx3Z//QojwAjFYAO2EzvTaayDEJfvOxcdHTk9YMyW7wLkT289Czs9SbCWPTDsSZIo6Cr5FUtWiUTR9EQwCWyLdsoF9/YvnJ5JpXPbwmGhY+R5TmT7RaHpuS42o4IMBYG3KHZ78pziVX00wSXBvYmcsvSbbUXTRKpJai5/j+3JNL2yqNDnm6G0Mx0inCLLWySi9cOBUQ8YgedFyafZ0sBO/PLS95K8gZsXjmY3FmA3qnOC9FkRbj8Hzj/X4vh+cP/nx38fb68gvdtn5MzpE9O7jVbpKwi8uWvheBSxBwaLqR3hex+Z8dEFAZ3IptHMoYctFE6zvnHjDkURDty1RBLIUw/CKPB+ZG8z7vdkO47hOV6wviC4uRkphpH1LBwxDF2Pz/DcaEOe/FA5JZEk0/YNmtlOTL+v9oxQSRI+4QX5/4s3Q+S2BuX3SF/tMw4ivNwQ0ff6DnszHAUr0oUehZCylZJcVGl7kVMGCFQ23aCLpFMhojSdZGPnSJIPFMxdgFUvwDYHNgMyB7avSCVsM2t1IGyFlrGtBG4oj7Qh5AGnXxvq4KYeP46EoVbEUJLLyinzlFMUugIQtG11zwpAoIj1AExDjfYVsBo+bJEwgja9IJp6E89FziiXDnKABdLK+3z0PJ+i8B+OohV94WgeeUXQyRsNVg/0/HXje9zoy2lzuNw8OFzRFkONJ83EpsmjxkCToSyU6BUzAS/t6CG9D/J548qklV84bmTXbZE9oTcPTPqqPw5/ea4/nd89TL5e3Vr6H676/iqNFVEwwdEWKAHtGOO1lYwBdlBkPxfDQh6z6Kl3nk3uOSNx5g1SKwQZ45LcKT0r5+d1EKDVRjc/7hCWGJzd6P6kFjmkVpyIAlVgt/Jz7qUHrsI1pNexiZD95Rq09Dj5NIl/f3D9ORkBxv9kI46Mx089dbDsqUMqpVciN55cLDmvmVFkeI5FS8Yq1wQqKkDK4e1bFlRIfa1s0lSeSwKwD7SOrJrEIQCFLLIdCxPEUvTG5LMYo6cOHxPxEDsRiiWyYdqBScbZOM2K+4k9eZSO2CauFsLaE9d+KaaGCdGO5rd0mQk89BLIoiiXUVa6ijuUaoSfKKTk0QyRxdOf2o/2OUMIFLkehLLUEYS8aV1ZSaP1C8mt6zlgBSCDlaiWsIIcmwphR1CJWo0w0bWu4ywVabmei3eacufhfBqZgXWsZ93Y8Y22HnNVhkiFeJyjE1TWMJKCgrg9kkpiw1IkVR4I7heStRVwSaBbh0t+G4/5GCZySOM2GWR94nm5Y6jq1e5Y4/C2M3csyfX98ctISsYnnp+uQv+Mked6cR7ynXlxqalr2Hfuz3iHupmA8cuZABbSYlojJDY0Sh/DdFAY2mYqpjeSphCyp9khndBG+mCrbT4Vn6fAIoMBk9qq7fPkttMQNa9Dc3G50iQjtupQQY28+AHycl1o1d65vgZKspk7s8bA+Ii+PfztXy0+/RkJys8ZvtJq6khT8gN5O2frkl8R9e0DHT0HB3hJOIbB4RT5CXtHjmP7YcxZHwcEWzI3IHKcSO9yUZUfHyPzx2RN+7/mkWPHvqbeqsLJenzAmh8J9vVyuAc4Fht0tuhTDvdvjGYB2OsFSGUBEpS+Xo7KFF44rvSFrlKggLey01ISnOjWOgueDDcOGg2WzRPGKKCzhN/Y1Prv55xbp0c1UORZsoAv5D+gxDkdlDknAq2vyl2RrkapxsZkII+iX7QE+67Qdb6imEU8Xa0FbrW81YuBsGZEc5ioX4ESE0RLfZExfbVjH50ZC5TH6jjfBVsyrhvGUDAzbuVCIAiKQvhXMpoGCiybRPKkjWaxQXPHYfxrcBX667CHvArfD7ylTeyo7bl1LSexRRFPHVPi0+n7ppZQEXLsSZzONQmn15FabNlsEznX9MDMtqz1rIRnj4szleMYWIkxsFAEvAopyKvPgKC7xUzIy6215MnTgUKfvMUmA5EnyCmWjFa8wutlXjrHbjoJ3YGJbEiZFuVVGOvOAn7IW269UPANU1CG8MQoyMszFyhY04E2YNgXHNrWPPa6wtjxzB+vi1ivpiqSV3Mu6jyv2wLZtpXccbi2k5UqEegOvWyjzjN7wtYSCmXoedVjWlfISxzkjzKB3X0i2iSxfrAJLCchv00DT2T6qsF+nmeRhTSvkiff+4qsZz/anhNbjV0cYO6zvUnt3Tx8no5Wi+WXD9f6TI/gz8GSa/R2pH62+JmR+/smt7lELzqpnMQ5b7/3NlaKW8vbVK9UtaUI3BUnqawIjVamDqMIKnhhZrIr11V2nxl7L+2RfdvrvxQfljxytq31CMWHXKh4tWUv26WuCkw45qJxqvhtFJgoJ2WhmLkEZCekdQ0Uu1+yq/IS9jpAFxmNaVZewlWpOru+XqdKdahE5TRNdzVbJ6VSJYpCuc+sIJ66VrVdtMXVKl6e/pXMHdsq0jq8jlSuhKonpUuAzXXBPSeLssQO1I0mAWmrJlX2B7DwXR7daB5veeKieR0HfK9N81gip1V0u2oekJndWZAZqCPNYyLD6v7gAD6v04KfXdJIbWrejqmnLnVNralrtTf/n4mba4vfNXa3XKqYaySVdGY5+bBVzFxoa3xzze6m6y0EAJXGJt3tdyLGpnInUG1jI+gb6zyCIO1leppuMDpIBijN219C5sNMVjl1u9vWJU5Es4CyXSH2depH0TNFKS5sVM1k2f4HiafFFlZhL3pZXy/li16eml5WzXOPo5fS29HLbneonNxEtAtNyTYB7LiO0RodeYvil2rYt1ANC5mSyPV36vXYmXJn9bB8uu22YHzS1u8Y3z9QWQ2mlcOQrfv/Tse6thR5HC7Nx99dwHPvB93QN0VBfBkrzhDObPeya6/u9yMw5pJXribJnO81THfI7mAsSTP/IwcJ9fI/FQFG/wM= \ No newline at end of file diff --git a/2109.02832/main_diagram/main_diagram.pdf b/2109.02832/main_diagram/main_diagram.pdf new file mode 100644 index 0000000000000000000000000000000000000000..4ed87f2966f9b8da1f10b3dab624a87a1e072a20 Binary files /dev/null and b/2109.02832/main_diagram/main_diagram.pdf differ diff --git a/2109.02832/paper_text/intro_method.md b/2109.02832/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..a6cf55b171c1678b9d2e9d6735e11733978575bd --- /dev/null +++ b/2109.02832/paper_text/intro_method.md @@ -0,0 +1,433 @@ +# Introduction + +Deep learning has achieved significant success in various practical applications with high-dimensional data set, such as computer vision (Krizhevsky et al., 2012), natural language processing (Graves et al., 2013; Young et al., 2018; Wu et al., 2016), health care (Miotto et al., 2018; Jiang et al., 2017) and bioinformatics (Alipanahi et al., 2015; Zhou and Troyanskaya, 2015). + +The success of deep learning clearly demonstrates the great power of neural networks in representing complex data. In the past decades, the representation power of neural networks has been extensively studied. The most commonly studied architecture is the feedforward neural network (FNN), as it has a simple composition form. The representation theory of FNNs has been developed with smooth activation functions (e.g., sigmoid) in Cybenko (1989); Barron (1993); McCaffrey and Gallant (1994); Hamers and Kohler (2006); Kohler and Krzyżak (2005); Kohler and Mehnert (2011) or nonsmooth activations (e.g., ReLU) in Lu et al. (2017); Yarotsky (2017); Lee et al. (2017); Suzuki (2019). These works show that if the network architecture is properly chosen, FNNs can approximate uniformly smooth functions (e.g., Hölder or Sobolev) with arbitrary accuracy. + +In real-world applications, convolutional neural networks (CNNs) are more popular than FNNs (LeCun et al., 1989; Krizhevsky et al., 2012; Sermanet et al., 2013; He et al., 2016; Chen et al., 2017; Long et al., 2015; Simonyan and Zisserman, 2014; Girshick, 2015). In a CNN, each layer consists of several filters (channels) which are convolved with the input, as demonstrated in Figure 1(a). Due to such complexity in the CNN architecture, there are limited works on the + +\*Hao Liu is affiliated with the Department of Mathematics at Hong Kong Baptist University; Wenjing Liao is affiliated with the School of Mathematics at Georgia Tech; Minshuo Chen and Tuo Zhao are affiliated with the ISYE department at Georgia Tech; Email: haoliu@hkbu.edu.hk, {mchen393, wliao60, tzhao80}@gatech.edu. + +![](_page_1_Figure_0.jpeg) + +Figure 1: Illustration of (a) convolution and (b) skip-layer connection. + +representation theory of CNNs (Zhou, 2020b,a; Fang et al., 2020; Petersen and Voigtlaender, 2020). The constructed CNNs in these works become extremely wide (in terms of the size of each layer's output) as the approximation error goes to 0. In most real-life applications, the network width does not exceed 2048 (Zagoruyko and Komodakis, 2016; Zhang et al., 2020). + +Convolutional residual networks (ConvResNet) is a special CNN architecture with skip-layer connections, as shown in Figure 1(b). Specifically, in addition to CNNs, ConvResNets have identity connections between inconsecutive layers. In many applications, ConvResNets outperform CNNs in terms of generalization performance and computational efficiency, and alleviate the vanishing gradient issue. Using this architecture, He et al. (2016) won the 1st place on the ImageNet classification task with a 3.57% top 5 error in 2015. + +Table 1: Comparison of our approximation theory and existing theoretical results. + +| | Network type | Function class | Low dim. structure | Fixed
width | Training | +|---------------------------|--------------|----------------|--------------------|----------------|-----------------------------------| +| Yarotsky (2017) | FNN | Sobolev | Х | Х | difficult to train | +| Suzuki (2019) | FNN | Besov | Х | Х | due to the cardinality constraint | +| Chen et al. (2019b) | FNN | Hölder | ✓ | Х | | +| Petersen and Voigtlaender | CNN | FNN | Х | Х | | +| (2020) | | | | | | +| Zhou (2020b) | CNN | Sobolev | Х | Х | can be trained | +| Oono and Suzuki (2019) | ConvResNet | Hölder | Х | ✓ | without the | +| Ours | ConvResNet | Besov | ✓ | ✓ | cardinality constraint | + +Recently, Oono and Suzuki (2019) develops the only representation and statistical estimation theory of ConvResNets. Oono and Suzuki (2019) proves that if the network architecture is properly set, ConvResNets with a fixed filter size and a fixed number of channels can universally approximate Hölder functions with arbitrary accuracy. However, the sample complexity in Oono and Suzuki (2019) grows exponentially with respect to the data dimension and therefore cannot well explain the empirical success of ConvResNets for high dimensional data. In order to estimate a $C^s$ function in $\mathbb{R}^D$ with accuracy $\varepsilon$ , the sample size required by Oono and Suzuki (2019) scales as $\varepsilon^{-\frac{2s+D}{s}}$ , which is far beyond the sample size used in practical applications. For example, the ImageNet data set consists of 1.2 million labeled images of size $224 \times 224 \times 3$ . According to this theory, to achieve a 0.1 error, the sample size is required to be in the order of $10^{224 \times 224 \times 3}$ which greatly exceeds 1.2 million. Due to the curse of dimensionality, there is a huge gap between theory and practice. + +We bridge such a gap by taking low-dimensional geometric structures of data sets into consideration. It is commonly believed that real world data sets exhibit low-dimensional structures due to rich local regularities, global symmetries, or repetitive patterns (Hinton and Salakhutdinov, 2006; Osher et al., 2017; Tenenbaum et al., 2000). For example, the ImageNet + +data set contains many images of the same object with certain transformations, such as rotation, translation, projection and skeletonization. As a result, the degree of freedom of the ImageNet data set is significantly smaller than the data dimension (Gong et al., 2019). + +The function space considered in Oono and Suzuki (2019) is the Hölder space in which functions are required to be differentiable everywhere up to certain order. In practice, the target function may not have high order derivatives. Function spaces with less restrictive conditions are more desirable. In this paper, we consider the Besov space $B_{p,q}^s$ , which is more general than the Hölder space. In particular, the Hölder and Sobolev spaces are special cases of the Besov space: + +$$W^{s+\alpha,\infty} = \mathcal{H}^{s,\alpha} \subseteq B^{s+\alpha}_{\infty,\infty} \subseteq B^{s+\alpha}_{p,a}$$ + +for any $0 < p, q \le \infty, s \in \mathbb{N}$ and $\alpha \in (0,1]$ . For practical applications, it has been demonstrated in image processing that Besov norms can capture important features, such as edges (Jaffard et al., 2001). Due to the generality of the Besov space, it is shown in Suzuki and Nitanda (2019) that kernel ridge estimators have a sub-optimal rate when estimating Besov functions. + +In this paper, we establish theoretical guarantees of ConvResNets for the approximation of Besov functions on a low-dimensional manifold, and a statistical theory on binary classification. Let $\mathcal{M}$ be a d-dimensional compact Riemannian manifold isometrically embedded in $\mathbb{R}^D$ . Denote the Besov space on $\mathcal{M}$ as $B_{p,q}^s(\mathcal{M})$ for $0 < p,q \le \infty$ and $0 < s < \infty$ . Our function approximation theory is established for $B_{p,q}^s(\mathcal{M})$ . For binary classification, we are given n i.i.d. samples $\{(\mathbf{x}_i,y_i)\}_{i=1}^n$ where $\mathbf{x}_i \in \mathcal{M}$ and $y_i \in \{-1,1\}$ is the label. The label y follows the Bernoulli-type distribution + +$$\mathbb{P}(y = 1|\mathbf{x}) = \eta(\mathbf{x}), \ \mathbb{P}(y = -1|\mathbf{x}) = 1 - \eta(\mathbf{x})$$ + +for some $\eta: \mathcal{M} \to [0,1]$ . Our results (Theorem 1 and 2) are summarized as follows: + +**Theorem** (informal). Assume $s \ge d/p + 1$ . + +1. Given $\varepsilon \in (0,1)$ , we construct a ConvResNet architecture such that, for any $f^* \in B^s_{p,q}(\mathcal{M})$ , if the weight parameters of this ConvResNet are properly chosen, it gives rises to $\bar{f}$ satisfying + +$$\|\bar{f} - f^*\|_{L^{\infty}} \le \varepsilon.$$ + +2. Assume $\eta \in B^s_{p,q}(\mathcal{M})$ . Let $f^*_{\phi}$ be the minimizer of the population logistic risk. If the ConvResNet architecture is properly chosen, minimizing the empirical logistic risk gives rise to $\widehat{f}_{\phi,n}$ with the following excess risk bound + +$$\mathbb{E}(\mathcal{E}_{\phi}(\widehat{f}_{\phi,n}, f_{\phi}^*)) \le C n^{-\frac{s}{2s+2(s\vee d)}} \log^4 n,$$ + +where $\mathcal{E}_{\phi}(\widehat{f_{\phi}}, n, f_{\phi}^*)$ denotes the excess logistic risk of $\widehat{f_{\phi}}, n$ against $f_{\phi}^*$ and C is a constant independent of n. + +We remark that the first part of the theorem above requires the network size to depend on the intrinsic dimension d and only weakly depend on D. The second part is built upon the first part and shows a fast convergence rate of the excess risk in terms of n where the exponent depends on d instead of D. Our results demonstrate that ConvResNets are adaptive to low-dimensional structures of data sets. + +**Related work.** Approximation theories of FNNs with the ReLU activation have been established for Sobolev (Yarotsky, 2017), Hölder (Schmidt-Hieber, 2017) and Besov (Suzuki, 2019) spaces. The networks in these works have certain cardinality constraint, i.e., the number of nonzero parameters is bounded by certain constant, which requires a lot of efforts for training. + +Approximation theories of CNNs are developed in Zhou (2020b); Petersen and Voigtlaender (2020); Oono and Suzuki (2019). Among these works, Zhou (2020b) shows that CNNs can + +approximate Sobolev functions in $W^{s,2}$ for $s \ge D/2 + 2$ with an arbitrary accuracy $\varepsilon \in (0,1)$ . The network in Zhou (2020b) has width increasing linearly with respect to depth and has depth growing in the order of $\varepsilon^{-2}$ as $\varepsilon$ decreases to 0. It is shown in Petersen and Voigtlaender (2020); Zhou (2020a) that any approximation error achieved by FNNs can be achieved by CNNs. Combining Zhou (2020a) and Yarotsky (2017), we can show that CNNs can approximate $W^{s,\infty}$ functions in $\mathbb{R}^D$ with arbitrary accuracy $\varepsilon$ . Such CNNs have the number of channels in the order of $\varepsilon^{-D/s}$ and a cardinality constraint. The only theory on ConvResNet can be found in Oono and Suzuki (2019), where an approximation theory for Hölder functions is proved for ConvResNets with fixed width. + +Statistical theories for binary classification by FNNs are established with the hinge loss (Ohn and Kim, 2019; Hu et al., 2020) and the logistic loss (Kim et al., 2018). Among these works, Hu et al. (2020) uses a parametric model given by a teacher-student network. The non-parametric results in Ohn and Kim (2019); Kim et al. (2018) are cursed by the data dimension, and therefore require a large number of samples for high-dimensional data. + +Binary classification by CNNs has been studied in Kohler et al. (2020); Kohler and Langer (2020); Nitanda and Suzuki (2018); Huang et al. (2018). Image binary classification is studied in Kohler et al. (2020); Kohler and Langer (2020) in which the probability function is assumed to be in a hierarchical max-pooling model class. ResNet type classifiers are considered in Nitanda and Suzuki (2018); Huang et al. (2018) while the generalization error is not given explicitly. + +Low-dimensional structures of data sets are explored for neural networks in Chui and Mhaskar (2018); Shaham et al. (2018); Chen et al. (2019b,a); Schmidt-Hieber (2019); Nakada and Imaizumi (2019); Cloninger and Klock (2020); Chen et al. (2020); Montanelli and Yang (2020). These works show that, if data are near a low-dimensional manifold, the performance of FNNs depends on the intrinsic dimension of the manifold and only weakly depends on the data dimension. Our work focuses on ConvResNets for practical applications. + +The networks in many aforementioned works have a cardinality constraint. From the computational perspective, training such networks requires substantial efforts (Han et al., 2016, 2015; Blalock et al., 2020). In comparison, the ConvResNet in Oono and Suzuki (2019) and this paper does not require any cardinality constraint. Additionally, our constructed network has a fixed filter size and a fixed number of channels, which is desirable for practical applications. + +As a summary, we compare our approximation theory and existing results in Table 1. + +The rest of this paper is organized as follows: In Section 2, we briefly introduce manifolds, Besov functions on manifolds and convolution. Our main results are presented in Section 3. We give a proof sketch in Section 4 and conclude this paper in Section 5. + +# Method + +**Notations**: We use bold lower-case letters to denote vectors, upper-case letters to denote matrices, calligraphic letters to denote tensors, sets and manifolds. For any x>0, we use $\lceil x \rceil$ to denote the smallest integer that is no less than x and use $\lfloor x \rfloor$ to denote the largest integer that is no larger than x. For any $a,b \in \mathbb{R}$ , we denote $a \lor b = \max(a,b)$ . For a function $f: \mathbb{R}^d \to \mathbb{R}$ and a set $\Omega \subset \mathbb{R}^d$ , we denote the restriction of f to $\Omega$ by $f|_{\Omega}$ . We use $\|f\|_{L^p}$ to denote the $L^p$ norm of f. We denote the Euclidean ball centered at $\mathbf{c}$ with radius $\omega$ by $B_{\omega}(\mathbf{c})$ . + +We first introduce some concepts on manifolds. We refer the readers to Tu (2010); Lee (2006) for details. Throughout this paper, we let $\mathcal{M}$ be a d-dimensional Riemannian manifold $\mathcal{M}$ isometrically embedded in $\mathbb{R}^D$ with $d \leq D$ . We first introduce charts, an atlas and the partition of unity. + +**Definition 1** (Chart). A chart on $\mathcal{M}$ is a pair $(U, \phi)$ where $U \subset \mathcal{M}$ is open and $\phi : U \to \mathbb{R}^d$ , is a homeomorphism (i.e., bijective, $\phi$ and $\phi^{-1}$ are both continuous). + +In a chart $(U, \phi)$ , U is called a coordinate neighborhood and $\phi$ is a coordinate system on U. A collection of charts which covers $\mathcal{M}$ is called an atlas of $\mathcal{M}$ . + +**Definition 2** ( $C^k$ Atlas). A $C^k$ atlas for $\mathcal{M}$ is a collection of charts $\{(U_\alpha, \phi_\alpha)\}_{\alpha \in \mathcal{A}}$ which satisfies $\bigcup_{\alpha \in \mathcal{A}} U_\alpha = \mathcal{M}$ , and are pairwise $C^k$ compatible: + +$$\phi_{\alpha} \circ \phi_{\beta}^{-1} : \phi_{\beta}(U_{\alpha} \cap U_{\beta}) \to \phi_{\alpha}(U_{\alpha} \cap U_{\beta}) \quad and \quad \phi_{\beta} \circ \phi_{\alpha}^{-1} : \phi_{\alpha}(U_{\alpha} \cap U_{\beta}) \to \phi_{\beta}(U_{\alpha} \cap U_{\beta})$$ + +are both $C^k$ for any $\alpha, \beta \in A$ . An atlas is called finite if it contains finitely many charts. + +**Definition 3** (Smooth Manifold). A smooth manifold is a manifold $\mathcal{M}$ together with a $C^{\infty}$ atlas. + +The Euclidean space, the torus and the unit sphere are examples of smooth manifolds. $C^s$ functions on a smooth manifold $\mathcal{M}$ are defined as follows: + +**Definition 4** ( $C^s$ functions on $\mathcal{M}$ ). Let $\mathcal{M}$ be a smooth manifold and $f: \mathcal{M} \to \mathbb{R}$ be a function on $\mathcal{M}$ . We say f is a $C^s$ function on $\mathcal{M}$ , if for every chart $(U, \phi)$ on $\mathcal{M}$ , the function $f \circ \phi^{-1} : \phi(U) \to \mathbb{R}$ is a $C^s$ function. + +We next define the $C^{\infty}$ partition of unity which is an important tool for the study of functions on manifolds. + +**Definition 5** (Partition of Unity). A $C^{\infty}$ partition of unity on a manifold $\mathcal{M}$ is a collection of $C^{\infty}$ functions $\{\rho_{\alpha}\}_{{\alpha}\in\mathcal{A}}$ with $\rho_{\alpha}:\mathcal{M}\to[0,1]$ such that for any $\mathbf{x}\in\mathcal{M}$ , + +1. there is a neighbourhood of $\mathbf{x}$ where only a finite number of the functions in $\{\rho_{\alpha}\}_{{\alpha}\in\mathcal{A}}$ are nonzero, and + +$$2. \sum_{\alpha \in \mathcal{A}} \rho_{\alpha}(\mathbf{x}) = 1.$$ + +An open cover of a manifold $\mathcal{M}$ is called locally finite if every $\mathbf{x} \in \mathcal{M}$ has a neighbourhood which intersects with a finite number of sets in the cover. The following proposition shows that a $C^{\infty}$ partition of unity for a smooth manifold always exists (Spivak, 1970, Chapter 2, Theorem 15). + +**Proposition 1** (Existence of a $C^{\infty}$ partition of unity). Let $\{U_{\alpha}\}_{{\alpha}\in\mathcal{A}}$ be a locally finite cover of a smooth manifold $\mathcal{M}$ . There is a $C^{\infty}$ partition of unity $\{\rho_{\alpha}\}_{{\alpha}=1}^{\infty}$ such that $\operatorname{supp}(\rho_{\alpha})\subset U_{\alpha}$ . + +Let $\{(U_{\alpha}, \phi_{\alpha})\}_{\alpha \in \mathcal{A}}$ be a $C^{\infty}$ atlas of $\mathcal{M}$ . Proposition 1 guarantees the existence of a partition of unity $\{\rho_{\alpha}\}_{\alpha \in \mathcal{A}}$ such that $\rho_{\alpha}$ is supported on $U_{\alpha}$ . + +The reach of $\mathcal{M}$ introduced by Federer (Federer, 1959) is an important quantity defined below. Let $d(\mathbf{x}, \mathcal{M}) = \inf_{\mathbf{y} \in \mathcal{M}} ||\mathbf{x} - \mathbf{y}||_2$ be the distance from $\mathbf{x}$ to $\mathcal{M}$ . + +Definition 6 (Reach (Federer, 1959; Nivogi et al., 2008)). Define the set + +$$G = \{ \mathbf{x} \in \mathbb{R}^D : \exists \text{ distinct } \mathbf{p}, \mathbf{q} \in \mathcal{M} \text{ such that } d(\mathbf{x}, \mathcal{M}) = ||\mathbf{x} - \mathbf{p}||_2 = ||\mathbf{x} - \mathbf{q}||_2 \}.$$ + +The closure of G is called the medial axis of M. The reach of M is defined as + +$$\tau = \inf_{\mathbf{x} \in \mathcal{M}} \inf_{\mathbf{y} \in G} ||\mathbf{x} - \mathbf{y}||_2.$$ + +We illustrate large and small reach in Figure 2. + +![](_page_5_Picture_0.jpeg) + +Slow Change: Large au Rapid Change: Small au + +Figure 2: Illustration of manifolds with large and small reach. + +We next define Besov function spaces on $\mathcal{M}$ , which generalizes more elementary function spaces such as the Sobolev and Hölder spaces. To define Besov functions, we first introduce the modulus of smoothness. + +**Definition 7** (Modulus of Smoothness (DeVore and Lorentz, 1993; Suzuki, 2019)). Let $\Omega \subset \mathbb{R}^D$ . For a function $f: \mathbb{R}^D \to \mathbb{R}$ be in $L^p(\Omega)$ for p > 0, the r-th modulus of smoothness of f is defined by + +$$\begin{split} w_{r,p}(f,t) &= \sup_{\|\mathbf{h}\|_2 \leq t} \|\Delta_{\mathbf{h}}^r(f)\|_{L^p}, \ where \\ \Delta_{\mathbf{h}}^r(f)(\mathbf{x}) &= \begin{cases} \sum_{j=0}^r \binom{r}{j} (-1)^{r-j} f(\mathbf{x}+j\mathbf{h}) & if \ \mathbf{x} \in \Omega, \mathbf{x}+r\mathbf{h} \in \Omega, \\ 0 & otherwise. \end{cases} \end{split}$$ + +**Definition 8** (Besov Space $B_{p,q}^s(\Omega)$ ). For $0 < p,q \le \infty, s > 0, r = \lfloor s \rfloor + 1$ , define the seminorm $|\cdot|_{B_{p,q}^s}$ as + +$$|f|_{B^s_{p,q}(\Omega)} := \begin{cases} \left(\int_0^\infty (t^{-s}w_{r,p}(f,t))^q \frac{dt}{t}\right)^{\frac{1}{q}} & \text{if } q < \infty, \\ \sup_{t>0} t^{-s}w_{r,p}(f,t) & \text{if } q = \infty. \end{cases}$$ + +The norm of the Besov space $B^s_{p,q}(\Omega)$ is defined as $||f||_{B^s_{p,q}(\Omega)} := ||f||_{L^p(\Omega)} + |f|_{B^s_{p,q}(\Omega)}$ . The Besov space is $B^s_{p,q}(\Omega) = \{f \in L^p(\Omega) |||f||_{B^s_{p,q}} < \infty\}$ . + +We next define $B_{p,q}^s$ functions on $\mathcal{M}$ (Geller and Pesenson, 2011; Triebel, 1983, 1992). + +**Definition 9** ( $B_{p,q}^s$ Functions on $\mathcal{M}$ ). Let $\mathcal{M}$ be a compact smooth manifold of dimension d. Let $\{(U_i, \phi_i)\}_{i=1}^{C_{\mathcal{M}}}$ be a finite atlas on $\mathcal{M}$ and $\{\rho_i\}_{i=1}^{C_{\mathcal{M}}}$ be a partition of unity on $\mathcal{M}$ such that $\operatorname{supp}(\rho_i) \subset U_i$ . A function $f: \mathcal{M} \to \mathbb{R}$ is in $B_{p,q}^s(\mathcal{M})$ if + +$$||f||_{B_{p,q}^s(\mathcal{M})} := \sum_{i=1}^{C_{\mathcal{M}}} ||(f\rho_i) \circ \phi_i^{-1}||_{B_{p,q}^s(\mathbb{R}^d)} < \infty.$$ + (1) + +Since $\rho_i$ is supported on $U_i$ , the function $(f\rho_i) \circ \phi_i^{-1}$ is supported on $\phi(U_i)$ . We can extend $(f\rho_i) \circ \phi_i^{-1}$ from $\phi(U_i)$ to $\mathbb{R}^d$ by setting the function to be 0 on $\mathbb{R}^d \setminus \phi(U_i)$ . The extended function lies in the Besov space $B_{p,q}^s(\mathbb{R}^d)$ (Triebel, 1992, Chapter 7). + +In this paper, we consider one-sided stride-one convolution in our network. Let $\mathcal{W} = \{\mathcal{W}_{j,k,l}\} \in \mathbb{R}^{C' \times K \times C}$ be a filter where C' is the output channel size, K is the filter size and C is the input channel size. For $z \in \mathbb{R}^{D \times C}$ , the convolution of $\mathcal{W}$ with z gives $y \in \mathbb{R}^{D \times C'}$ such that + +$$y = \mathcal{W} * z$$ +, $y_{i,j} = \sum_{k=1}^{K} \sum_{l=1}^{C} \mathcal{W}_{j,k,l} z_{i+k-1,l}$ , (2) + +where $1 \le i \le D$ , $1 \le j \le C'$ and we set $z_{i+k-1,l} = 0$ for i+k-1 > D, as demonstrated in Figure 3(a). + +The building blocks of ConvResNets are residual blocks. For an input x, each residual block computes + +$$\mathbf{x} + F(\mathbf{x})$$ + +where F is a subnetwork consisting of convolutional layers (see more details in Section 3.1). A residual block is demonstrated in Figure 3(b). + +![](_page_6_Figure_4.jpeg) + +Figure 3: (a) Demonstration of W\*z, where the input is $z \in \mathbb{R}^{D \times C}$ , and the output is $W*z \in$ $\mathbb{R}^{D \times C'}$ . Here $\mathcal{W} = \{\mathcal{W}_{i,k,l}\} \in \mathbb{R}^{C' \times K \times C}$ is a filter where C' is the output channel size, K is the filter size and C is the input channel size. $W_{i...}$ is a $D \times C$ matrix for the j-th output channel. (b) Demonstration of a residual block. + +In this section, we first introduce the ConvResNet architecture, and then present our main results. + +We study the ConvResNet with the rectified linear unit (ReLU) activation function: $ReLU(z) = \frac{1}{2} \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2} \right) \left( \frac{1}{2}$ max(z, 0). The ConvResNet we consider consists of a padding layer and several residual blocks followed by a fully connected feedforward layer. + +We first define the padding layer. Given an input $A \in \mathbb{R}^{D \times C_1}$ , the network first applies a padding operator $P: \mathbb{R}^{D \times C_1} \to \mathbb{R}^{D \times C_2}$ for some integer $C_2 \ge C_1$ such that + +$$Z = P(A) = \begin{bmatrix} A & \mathbf{0} & \cdots & \mathbf{0} \end{bmatrix} \in \mathbb{R}^{D \times C_2}.$$ + +Then the matrix Z is passed through M residual blocks. In the m-th block, let $\mathcal{W}_m = \{\mathcal{W}_m^{(1)},...,\mathcal{W}_m^{(L_m)}\}$ and $\mathcal{B}_m = \{B_m^{(1)},...,B_m^{(L_m)}\}$ be a collection of filters and biases. The m-th residual block maps a matrix from $\mathbb{R}^{D \times C}$ to $\mathbb{R}^{D \times C}$ by + +$$Conv_{\mathcal{W}_{m},\mathcal{B}_{m}} + id$$ + +where id is the identity operator and + +$$\operatorname{Conv}_{\mathcal{W}_m,\mathcal{B}_m}(Z) = \operatorname{ReLU}\left(\mathcal{W}_m^{(L_m)} * \cdots * \operatorname{ReLU}\left(\mathcal{W}_m^{(1)} * Z + B_m^{(1)}\right) \cdots + B_m^{(L_m)}\right), \tag{3}$$ + +with ReLU applied entrywise. Denote + +$$Q(\mathbf{x}) = \left(\operatorname{Conv}_{\mathcal{W}_{M}, \mathcal{B}_{M}} + \operatorname{id}\right) \circ \cdots \circ \left(\operatorname{Conv}_{\mathcal{W}_{1}, \mathcal{B}_{1}} + \operatorname{id}\right) \circ P(\mathbf{x}). \tag{4}$$ + +For networks only consisting of residual blocks, we define the network class as + + $C^{\text{Conv}}(M, L, J, K, \kappa) = \{Q | Q(\mathbf{x}) \text{ is in the form of (4) with } M \text{ residual blocks. Each block has} \}$ filter size bounded by K, number of channels bounded by J, + +$$\max_{m} L_{m} \le L, \ \max_{m,l} \|\mathcal{W}_{m}^{(l)}\|_{\infty} \lor \|B_{m}^{(l)}\|_{\infty} \le \kappa \}, \tag{5}$$ + +where $\|\cdot\|_{\infty}$ denotes $\ell^{\infty}$ norm of a vector, and for a tensor $\mathcal{W}$ , $\|\mathcal{W}\|_{\infty} = \max_{i,k,l} |\mathcal{W}_{i,k,l}|$ . + +Based on the network Q in (4), a ConvResNet has an additional fully connected layer and can be expressed as + +$$f(\mathbf{x}) = WQ(\mathbf{x}) + b \tag{6}$$ + +where W and b are the weight matrix and the bias in the fully connected layer. The class of ConvResNets is defined as + +$$C(M, L, J, K, \kappa_1, \kappa_2, R) = \{ f | f(\mathbf{x}) = WQ(\mathbf{x}) + b \text{ with } Q \in C^{\text{Conv}}(M, L, J, K, \kappa_1),$$ + +$$||W||_{\infty} \vee |b| \leq \kappa_2, ||f||_{L^{\infty}} \leq R \}.$$ + +$$(7)$$ + +Sometimes we do not have restriction on the output, we omit the parameter R and denote the network class by $C(M, L, J, K, \kappa_1, \kappa_2)$ . + +Our approximation theory is based on the following assumptions of $\mathcal{M}$ and the object function $f^* : \mathcal{M} \to \mathbb{R}$ . + +**Assumption 1.** $\mathcal{M}$ is a d-dimensional compact smooth Riemannian manifold isometrically embedded in $\mathbb{R}^D$ . There is a constant B such that for any $\mathbf{x} \in \mathcal{M}$ , $\|\mathbf{x}\|_{\infty} \leq B$ . + +**Assumption 2.** *The reach of* $\mathcal{M}$ *is* $\tau > 0$ . + +**Assumption 3.** Let $0 < p, q \le \infty$ , $d/p + 1 \le s < \infty$ . Assume $f^* \in B^s_{p,q}(\mathcal{M})$ and $||f^*||_{B^s_{p,q}(\mathcal{M})} \le c_0$ for a constant $c_0 > 0$ . Additionally, we assume $||f^*||_{L^\infty} \le R$ for a constant R > 0. + +Assumption 3 implies that $f^*$ is Lipschitz continuous (Triebel, 1983, Section 2.7.1 Remark 2 and Section 3.3.1). + +Our first result is the following universal approximation error of ConvResNets for Besov functions on $\mathcal{M}$ . + +![](_page_7_Figure_16.jpeg) + +Figure 4: The ConvResNet in Theorem 1 contains a padding layer, *M* residual blocks, and a fully connected (FC) layer. + +**Theorem 1.** Assume Assumption 1-3. For any $\varepsilon \in (0,1)$ and positive integer $K \in [2,D]$ , there is a ConvResNet architecture $C(M,L,J,K,\kappa_1,\kappa_2)$ such that, for any $f^* \in B^s_{p,q}(\mathcal{M})$ , if the weight parameters of this ConvResNet are properly chosen, the network yields a function $\bar{f} \in C(M,L,J,K,\kappa_1,\kappa_2)$ satisfying + +$$\|\bar{f} - f^*\|_{L^{\infty}} \le \varepsilon. \tag{8}$$ + +Such a network architecture has + +$$M = O\left(\varepsilon^{-d/s}\right), \ L = O(\log(1/\varepsilon) + D + \log D), \ J = O(D), \ \kappa_1 = O(1), \ \log \kappa_2 = O(\log^2(1/\varepsilon)). \tag{9}$$ + +The constant hidden in $O(\cdot)$ depend on d, s, $\frac{2d}{sp-d}$ , p, q, $c_0$ , $\tau$ and the surface area of $\mathcal{M}$ . + +The architecture of the ConvResNet in Theorem 1 is illustrated in Figure 4. It has the following properties: + +- The network has a fixed filter size and a fixed number of channels. +- There is no cardinality constraint. +- The network size depends on the intrinsic dimension d, and only weakly depends on D. + +Theorem 1 can be compared with Suzuki (2019) on the approximation theory for Besov functions in $\mathbb{R}^D$ by FNNs as follows: (1) To universally approximate Besov functions in $\mathbb{R}^D$ with $\varepsilon$ error, the FNN constructed in Suzuki (2019) requires $O(\log(1/\varepsilon))$ depth, $O(\varepsilon^{-D/s})$ width and $O(\varepsilon^{-D/s}\log(1/\varepsilon))$ nonzero parameters. By exploiting the manifold model, our network size depends on the intrinsic dimension d and weakly depends on D. (2) The ConvResNet in Theorem 1 does not require any cardinality constraint, while such a constraint is needed in Suzuki (2019). + +We next consider binary classification on $\mathcal{M}$ . For any $\mathbf{x} \in \mathcal{M}$ , denote its label by $y \in \{-1,1\}$ . The label y follows the following Bernoulli-type distribution + +$$\mathbb{P}(v=1|\mathbf{x}) = n(\mathbf{x}), \ \mathbb{P}(v=-1|\mathbf{x}) = 1 - n(\mathbf{x}) \tag{10}$$ + +for some $\eta: \mathcal{M} \to [0,1]$ . + +We assume the following data model: + +**Assumption 4.** We are given i.i.d. sample $\{(\mathbf{x}_i, y_i)\}_{i=1}^n$ , where $\mathbf{x}_i \in \mathcal{M}$ , and the $y_i$ 's are sampled according to (10). + +In binary classification, a classifier f predicts the label of $\mathbf{x}$ as $\mathrm{sign}(f(\mathbf{x}))$ . To learn the optimal classifier, we consider the logistic loss $\phi(z) = \log(1 + \exp(-z))$ . The logistic risk $\mathcal{E}_{\phi}(f)$ of a classifier f is defined as + +$$\mathcal{E}_{\phi}(f) = \mathbb{E}(\phi(\gamma f(\mathbf{x}))). \tag{11}$$ + +The minimizer of $\mathcal{E}_{\phi}(f)$ is denoted by $f_{\phi}^*$ , which satisfies + +$$f_{\phi}^{*}(\mathbf{x}) = \log \frac{\eta(\mathbf{x})}{1 - \eta(\mathbf{x})}.$$ +(12) + +For any classifier f, we define its logistic excess risk as + +$$\mathcal{E}_{\phi}(f, f_{\phi}^*) = \mathcal{E}_{\phi}(f) - \mathcal{E}_{\phi}(f_{\phi}^*). \tag{13}$$ + +In this paper, we consider ConvResNets with the following architecture: + +$$C^{(n)} = \left\{ f | f = \bar{g}_2 \circ \bar{h} \circ \bar{g}_1 \circ \bar{\eta} \text{ where } \bar{\eta} \in C^{\text{Conv}}(M_1, L_1, J_1, K, \kappa_1), \ \bar{g}_1 \in C^{\text{Conv}}(1, 4, 8, 1, \kappa_2), \\ \bar{h} \in C^{\text{Conv}}(M_2, L_2, J_2, 1, \kappa_1), \ \bar{g}_2 \in C(1, 3, 8, 1, \kappa_3, 1, R) \right\},$$ +(14) + +where $M_1, M_2, L, J, K, \kappa_1, \kappa_2, \kappa_3$ are some parameters to be determined. + +The empirical classifier is learned by minimizing the empirical logistic risk: + +$$\widehat{f}_{\phi,n} = \underset{f \in \mathcal{C}^{(n)}}{\operatorname{argmin}} \frac{1}{n} \sum_{i=1}^{n} \phi(y_i f(\mathbf{x}_i)). \tag{15}$$ + +We establish an upper bound on the excess risk of $\widehat{f_{\phi,n}}$ : + +**Theorem 2.** Assume Assumption 1, 2 and 4. Assume $0 < p, q \le \infty$ , $0 < s < \infty$ , $s \ge d/p + 1$ and $\eta \in B^s_{p,q}(\mathcal{M})$ with $\|\eta\|_{B^s_{p,q}} \le c_0$ for some constant $c_0$ . For any $2 \le K \le D$ , we set + +$$M_{1} = O\left(n^{\frac{2d}{s+2(s \vee d)}}\right), \ M_{2} = O\left(n^{\frac{2s}{s+2(s \vee d)}}\right), \ L_{1} = O(\log(1/\varepsilon) + D + \log D), \ L_{2} = O(\log(1/\varepsilon)),$$ + +$$J_{1} = O(D), \ J_{2} = O(1), \ \kappa_{1} = O(1), \ \log \kappa_{2} = O(\log^{2} n), \ \kappa_{3} = O(\log n), \ R = O(\log n)$$ + +for $C^{(n)}$ . Then + +$$\mathbb{E}(\mathcal{E}_{\phi}(\widehat{f_{\phi}}, n, f_{\phi}^{*})) \le C n^{-\frac{s}{2s + 2(s \lor d)}} \log^{4} n \tag{16}$$ + +for some constant C. Here C is linear in D log D and additionally depends on $d, s, \frac{2d}{sp-d}, p, q, c_0, \tau$ and the surface area of M. The constant hidden in $O(\cdot)$ depends on $d, s, \frac{2d}{sp-d}, p, q, c_0, \tau$ and the surface area of M. + +Theorem 2 shows that a properly designed ConvResNet gives rise to an empirical classifier, of which the excess risk converges at a fast rate with an exponent depending on the intrinsic dimension d, instead of D. + +Theorem 2 is proved in Appendix A. Each building block of $C^{(n)}$ is constructed for the following purpose: + +- $\bar{g}_1 \circ \bar{\eta}$ is designed to approximate a truncated $\eta$ on $\mathcal{M}$ , which is realized by Theorem 1. +- $\bar{g}_2 \circ \bar{h}$ is designed to approximate a truncated univariate function $\log \frac{z}{1-z}$ . + +We provide a proof sketch of Theorem 1 in this section. More technical details are deferred to Appendix C. + +We prove Theorem 1 in the following four steps: + +- 1. Decompose $f^* = \sum_i f_i$ as a sum of locally supported functions according to the manifold structure. +- 2. Locally approximate each $f_i$ using cardinal B-splines. +- 3. Implement the cardinal B-splines using CNNs. +- 4. Implement the sum of all CNNs by a ConvResNet for approximating $f^*$ . + +![](_page_10_Picture_0.jpeg) + +Figure 5: An atlas given by covering $\mathcal{M}$ using Euclidean balls. + +• Construct an atlas on $\mathcal{M}$ . Since the manifold $\mathcal{M}$ is compact, we can cover $\mathcal{M}$ by a finite collection of open balls $B_{\omega}(\mathbf{c}_i)$ for $i=1,\ldots,C_{\mathcal{M}}$ , where $\mathbf{c}_i$ is the center of the ball and $\omega$ is the radius to be chosen later. Accordingly, the manifold is partitioned as $\mathcal{M}=\bigcup_i U_i$ with $U_i=B_{\omega}(\mathbf{c}_i)\cap\mathcal{M}$ . We choose $\omega<\tau/2$ such that $U_i$ is diffeomorphic to an open subset of $\mathbb{R}^d$ (Niyogi et al., 2008, Lemma 5.4). The total number of partitions is then bounded by $C_{\mathcal{M}} \leq \left\lceil \frac{\mathrm{SA}(\mathcal{M})}{\omega^d} T_d \right\rceil$ , where $\mathrm{SA}(\mathcal{M})$ is the surface area of $\mathcal{M}$ and $T_d$ is the average number of $U_i$ 's that contain a given point on $\mathcal{M}$ (Conway et al., 1987, Chapter 2 Equation (1)). + +On each partition, we define a projection-based transformation $\phi_i$ as + +$$\phi_i(\mathbf{x}) = a_i V_i^{\top}(\mathbf{x} - \mathbf{c}_i) + \mathbf{b}_i,$$ + +where the scaling factor $a_i \in \mathbb{R}$ and the shifting vector $\mathbf{b}_i \in \mathbb{R}^d$ ensure $\phi_i(U_i) \subset [0,1]^d$ , and the column vectors of $V_i \in \mathbb{R}^{D \times d}$ form an orthonormal basis of the tangent space $T_{\mathbf{c}_i}(\mathcal{M})$ . The atlas on $\mathcal{M}$ is the collection $(U_i, \phi_i)$ for $i = 1, ..., \mathcal{M}$ . See Figure 5 for a graphical illustration of the atlas. + +• **Decompose** $f^*$ according to the atlas. We decompose $f^*$ as + +$$f^* = \sum_{i=1}^{C_M} f_i \quad \text{with} \quad f_i = f \rho_i, \tag{17}$$ + +where $\{\rho_i\}_{i=1}^{C_M}$ is a $C^{\infty}$ partition of unity with $\operatorname{supp}(\phi_i) \subset U_i$ . The existence of such a $\{\rho_i\}_{i=1}^{C_M}$ is guaranteed by Proposition 1. As a result, each $f_i$ is supported on a subset of $U_i$ , and therefore, we can rewrite (17) as + +$$f^* = \sum_{i=1}^{C_M} (f_i \circ \phi_i^{-1}) \circ \phi_i \times \mathbb{1}_{U_i} \quad \text{with} \quad f_i = f \rho_i,$$ + (18) + +where $\mathbb{1}_{U_i}$ is the indicator function of $U_i$ . Since $\phi_i$ is a bijection between $U_i$ and $\phi_i(U_i)$ , $f_i \circ \phi_i^{-1}$ is supported on $\phi_i(U_i) \subset [0,1]^d$ . We extend $f_i \circ \phi_i^{-1}$ on $[0,1]^d \setminus \phi_i(U_i)$ by 0. The extended function is in $B_{p,q}^s([0,1]^d)$ (see Lemma 4 in Appendix C.1). This allows us to use cardinal B-splines to locally approximate each $f_i \circ \phi_i^{-1}$ as detailed in **Step 2**. + +Step 2: Local cardinal B-spline approximation. We approximate $f_i \circ \phi_i^{-1}$ using cardinal B-splines $\widetilde{f_i}$ as + +$$f_i \circ \phi_i^{-1} \approx \widetilde{f_i} \equiv \sum_{j=1}^N \widetilde{f_{i,j}} \text{ with } \widetilde{f_{i,j}} = \alpha_{k,j}^{(i)} M_{k,j,m}^d,$$ + (19) + +where $\alpha_{k,\mathbf{j}}^{(i)} \in \mathbb{R}$ is a coefficient and $M_{k,\mathbf{j},m}^d : [0,1]^d \to \mathbb{R}$ denotes a cardinal B-spline with indecies $k,m \in \mathbb{N}^+$ , $\mathbf{j} \in \mathbb{R}^d$ . Here k is a scaling factor, $\mathbf{j}$ is a shifting vector, m is the degree of the B-spline and d is the dimension (see a formal definition in Appendix C.2). + +Since $s \ge d/p + 1$ (by Assumption 3), setting $r = +\infty$ , $m = \lceil s \rceil + 1$ in Lemma 5 (see Appendix C.3) and applying Lemma 4 gives + +$$\|\widetilde{f_i} - f_i \circ \phi_i^{-1}\|_{L^{\infty}} \le Cc_0 N^{-s/d} \tag{20}$$ + +for some constant C depending on s, p, q and d. + +Combining (18) and (19), we approximate $f^*$ by + +$$\widetilde{f}^* \equiv \sum_{i=1}^{C_{\mathcal{M}}} \widetilde{f_i} \circ \phi_i \times \mathbb{1}_{U_i} = \sum_{i=1}^{C_{\mathcal{M}}} \sum_{i=1}^{N} \widetilde{f_{i,j}} \circ \phi_i \times \mathbb{1}_{U_i}.$$ + (21) + +Such an approximation has error + +$$\|\widetilde{f}^* - f^*\|_{I^{\infty}} \le CC_M c_0 N^{-s/d}$$ + +- Step 3: Implement local approximations in Step 2 by CNNs. In Step 2, (21) gives a natural approximation of $f^*$ . In the sequel, we aim to implement all ingredients of $\widetilde{f_{i,j}} \circ \phi_i \times \mathbb{1}_{U_i}$ using CNNs. In particular, we show that CNNs can implement the cardinal B-spline $\widetilde{f_{i,j}}$ , the linear projection $\phi_i$ , the indicator function $\mathbb{1}_{U_i}$ , and the multiplication operation. +- Implement $\mathbb{1}_{U_i}$ by CNNs. Recall our construction of $U_i$ in Step 1. For any $\mathbf{x} \in \mathcal{M}$ , we have $\mathbb{1}_{U_i}(\mathbf{x}) = 1$ if $d_i^2(\mathbf{x}) = ||\mathbf{x} \mathbf{c}_i||_2^2 \le \omega^2$ ; otherwise $\mathbb{1}_{U_i}(\mathbf{x}) = 0$ . + +To implement $\mathbb{1}_{U_i}$ , we rewrite it as the composition of a univariate indicator function $\mathbb{1}_{[0,\omega^2]}$ and the distance function $d_i^2$ : + +$$\mathbb{1}_{U_i}(\mathbf{x}) = \mathbb{1}_{[0,\omega^2]} \circ d_i^2(\mathbf{x}) \quad \text{for} \quad \mathbf{x} \in \mathcal{M}.$$ + (22) + +We show that CNNs can efficiently implement both $\mathbb{1}_{[0,\omega^2]}$ and $d_i^2$ . Specifically, given $\theta \in (0,1)$ and $\Delta \geq 8DB^2\theta$ , there exist CNNs that yield functions $\widetilde{\mathbb{1}}_{\Delta}$ and $\widetilde{d}_i^2$ satisfying + +$$\|\widetilde{d}_{i}^{2} - d_{i}^{2}\|_{L^{\infty}} \le 4B^{2}D\theta \tag{23}$$ + +and + +$$\widetilde{\mathbb{I}}_{\Delta} \circ \widetilde{d}_{i}^{2}(\mathbf{x}) = \begin{cases} +1, & \text{if } \mathbf{x} \in U_{i}, d_{i}^{2}(\mathbf{x}) \leq \omega^{2} - \Delta, \\ +0, & \text{if } \mathbf{x} \notin U_{i}, \\ +\text{between 0 and 1, otherwise.} +\end{cases} \tag{24}$$ + +We also characterize the network sizes for realizing $\widetilde{\mathbb{1}}_{\Delta}$ and $\widetilde{d}_i^2$ : The network for $\widetilde{\mathbb{1}}_{\Delta}$ has $O(\log(\omega^2/\Delta))$ layers, 2 channels and all weight parameters bounded by $\max(2, |\omega^2 - 4B^2D\theta|)$ ; the network for $\widetilde{d}_i^2$ has $O(\log(1/\theta) + D)$ layers, 6D channels and all weight parameters bounded by $4B^2$ . More technical details are provided in Lemma 9 in Appendix C.6. + +• **Implement** $f_{i,j} \circ \phi_i$ **by CNNs.** Since $\phi_i$ is a linear projection, it can be realized by a single-layer perceptron. By Lemma 8 (see Appendix C.5), this single-layer perceptron can be realized by a CNN, denoted by $\phi_i^{\text{CNN}}$ . + +For $f_{i,j}$ , Proposition 3 (see Appendix C.8) shows that for any $\delta \in (0,1)$ and $2 \le K \le d$ , there exists a CNN $\widetilde{f}_{i,j}^{\text{CNN}} \in \mathcal{F}^{\text{CNN}}(L,J,K,\kappa,\kappa)$ with + +$$L = O\left(\log \frac{1}{\delta}\right), J = O(1), \kappa = O\left(\delta^{-(\log 2)(\frac{2d}{sp-d} + \frac{c_1}{d})}\right)$$ + +such that when setting $N = C_1 \delta^{-d/s}$ , we have + +$$\left\| \sum_{j=1}^{N} \widetilde{f}_{i,j}^{\text{CNN}} - f_i \circ \phi_i^{-1} \right\|_{L^{\infty}(\phi_i(U_i))} \le \delta, \tag{25}$$ + +where $C_1$ is a constant depending on s, p, q and d. The constant hidden in $O(\cdot)$ depends on $d, s, \frac{2d}{sp-d}, p, q, c_0$ . The CNN class $\mathcal{F}^{\text{CNN}}$ is defined in Appendix B. + +• Implement the multiplication × by a CNN. According to Lemma 7 (see Appendix C.4) and Lemma 8, for any $\eta \in (0,1)$ , the multiplication operation × can be approximated by a CNN $\widetilde{\times}$ with $L^{\infty}$ error $\eta$ : + +$$||a \times b - \widetilde{\times}(a, b)||_{L^{\infty}} \le \eta. \tag{26}$$ + +Such a CNN has $O(\log 1/\eta)$ layers, 6 channels. All parameters are bounded by $\max(2c_0^2, 1)$ . + +**Step 4: Implement** $\widetilde{f}^*$ **by a ConvResNet.** We assemble all CNN approximations in **Step 3** together and show that the whole approximation can be realized by a ConvResNet. + +• Assemble all ingredients together. Assembling all CNN approximations together gives an approximation of $\widetilde{f_{i,j}} \circ \phi_i \times \mathbb{1}_{U_i}$ as + +$$\hat{f}_{i,j} \equiv \widetilde{\times} \left( \widetilde{f}_{i,j}^{\text{CNN}} \circ \phi_i^{\text{CNN}}, \widetilde{\mathbb{1}}_{\Delta} \circ \widetilde{d}_i^2 \right). \tag{27}$$ + +After substituting (27) into (21), we approximate the target function $f^*$ by + +$$\mathring{f} = \sum_{i=1}^{C_{\mathcal{M}}} \sum_{j=1}^{N} \mathring{f}_{i,j}.$$ + (28) + +The approximation error of $\mathring{f}$ is analyzed in Lemma 12 (see Appendix C.9). According to Lemma 12, the approximation error can be bounded as follows: + +$$\|\mathring{f} - f^*\|_{L^{\infty}} \leq \sum_{i=1}^{C_{\mathcal{M}}} (A_{i,1} + A_{i,2} + A_{i,3}) \quad \text{with}$$ + +$$A_{i,1} = \sum_{j=1}^{N} \left\| \widetilde{\times} (\widetilde{f}_{i,j}^{\text{CNN}} \circ \phi_i^{\text{CNN}}, \widetilde{\mathbb{I}}_{\Delta} \circ \widetilde{d}_i^2) - (\widetilde{f}_{i,j}^{\text{CNN}} \circ \phi_i^{\text{CNN}}) \times (\widetilde{\mathbb{I}}_{\Delta} \circ \widetilde{d}_i^2) \right\|_{L^{\infty}} \leq N\eta,$$ + +$$A_{i,2} = \left\| \left( \sum_{j=1}^{N} \left( \widetilde{f}_{i,j}^{\text{CNN}} \circ \phi_i^{\text{CNN}} \right) \right) \times (\widetilde{\mathbb{I}}_{\Delta} \circ \widetilde{d}_i^2) - f_i \times (\widetilde{\mathbb{I}}_{\Delta} \circ \widetilde{d}_i^2) \right\|_{L^{\infty}} \leq \delta,$$ + +$$A_{i,3} = \|f_i \times (\widetilde{\mathbb{I}}_{\Delta} \circ \widetilde{d}_i^2) - f_i \times \mathbb{1}_{U_i} \|_{L^{\infty}} \leq \frac{c(\pi + 1)}{\omega(1 - \omega/\tau)} \Delta,$$ + +where $\delta$ , $\eta$ , $\Delta$ and $\theta$ are defined in (25), (26), (24) and (23), respectively. For any $\varepsilon \in (0, 1)$ , with properly chosen $\delta$ , $\eta$ , $\Delta$ and $\theta$ as in (53) in Lemma 12, one has + +$$\|\mathring{f} - f^*\|_{L^{\infty}} \le \varepsilon. \tag{29}$$ + +With these choices, the network size of each CNN is quantified in Appendix C.10. + +• **Realize** $\mathring{f}$ **by a ConvResNet.** Lemma 17 (see Appendix C.15) shows that for every $\mathring{f}_{i,j}$ , there exists $\bar{f}_{i,j}^{\text{CNN}} \in \mathcal{F}^{\text{CNN}}(L,J,K,\kappa_1,\kappa_2)$ with $L = O(\log 1/\varepsilon + D + \log D), J = O(D), \kappa_1 = O(1), \log \kappa_2 = O(\log^2 1/\varepsilon)$ such that $\bar{f}_{i,j}^{\text{CNN}}(\mathbf{x}) = \mathring{f}_{i,j}(\mathbf{x})$ for any $\mathbf{x} \in \mathcal{M}$ . As a result, the function $\mathring{f}$ in (28) can be expressed as a sum of CNNs: + +$$\mathring{f} = \bar{f}^{\text{CNN}} \equiv \sum_{i=1}^{C_{\mathcal{M}}} \sum_{j=1}^{N} \bar{f}_{i,j}^{\text{CNN}},$$ +(30) + +where N is chosen of $O\left(\varepsilon^{-d/s}\right)$ (see Proposition 3 and Lemma 12). Lemma 18 (see Appendix C.16) shows that $\bar{f}^{\text{CNN}}$ can be realized by $\bar{f} \in \mathcal{C}(M, L, J, \kappa_1, \kappa_2)$ with + +$$M = O\left(\varepsilon^{-d/s}\right), L = O(\log(1/\varepsilon) + D + \log D), J = O(D), \kappa_1 = O(1), \log \kappa_2 = O\left(\log^2(1/\varepsilon)\right)$$ diff --git a/2201.10222/main_diagram/main_diagram.drawio b/2201.10222/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..14ec57065080c5421311677f16ca63491e04b6e8 --- /dev/null +++ b/2201.10222/main_diagram/main_diagram.drawio @@ -0,0 +1 @@ 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 \ No newline at end of file diff --git a/2201.10222/paper_text/intro_method.md b/2201.10222/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..f9572531393daf83c0926a26ba30e2decca77d19 --- /dev/null +++ b/2201.10222/paper_text/intro_method.md @@ -0,0 +1,123 @@ +# Introduction + +Making accurate predictions about the future is a key ability to survive and thrive in a habitat. Living beings have evolved many systems to this end, such as memory [@mcconnell1962memory], and several can predict the course of complex phenomena [@Taylor16389]. However, no animal comes even close to the prediction ability of humans, which stems from a unique-in-nature system. + +At the core of this system lies an object called *explanation*, formed by the proposition of a language, which has a remarkable property: it can be installed with ease into another human speaking the same language, allowing to make predictions on new phenomena without ever having experienced them. When the installation is successful, we say that the human has *understood* the explanation. + +This process is key to the success of human beings. An individual can provide accurate predictions for a multitude of phenomena without going through a painful discovery process for all of them, but only needs an operating system -- mastering a language -- and someone who communicates the relevant explanations; this way, the individual can focus on unexplained phenomena. When an explanation is found for them, it is added to the existing shared collection, which we call *knowledge*. + +How can we make machines take part in this orchestra? With this work, we try to shed new light on this problem. Specifically, we propose a learning procedure to allow machines (i) to *understand* existing explanations, in the sense described above, and (ii) create new explanations for unexplained phenomena, much like human scientists do. + +Our contribution in this sense is threefold: + +i\) We formulate the challenge of creating a machine that masters a language as the problem of learning an interpreter from a collection of examples in the form $(\text{explanation}, \text{observations})$. The only assumption we make is this dual structure of data; explanations are free strings, and are not required to fit any formal grammar. This results in the *Explanatory Learning* (EL) framework described in Sec. [2](#sec-el){reference-type="ref" reference="sec-el"}. + +ii\) We present Odeen, a basic environment to test EL approaches, which draws inspiration from the board game Zendo [@zendo]. Odeen simulates the work of a scientist in a small universe of simple geometric figures, see Figure [1](#fig-galileo){reference-type="ref" reference="fig-galileo"}. We present it in Sec. [3](#sec-odeen){reference-type="ref" reference="sec-odeen"}, and release it with this work[^1]. + +iii\) We argue that the dominating empiricist ML approaches are not suitable for EL problems. We propose *Critical Rationalist Networks* (CRNs), a family of models designed according to the epistemological philosophy pushed forward by [@popper1935logic]. Although a CRN is implemented using two neural networks, the working hypothesis of such a model does not coincide with the adjustable network parameters, but rather with a language proposition that can only be accepted or refused *in toto*. We will present CRNs in Sec. [4](#sec-crns){reference-type="ref" reference="sec-crns"}, and test their performance on Odeen in Sec. [5](#sec-exp){reference-type="ref" reference="sec-exp"}. + +Humans do not master a language from birth. A baby can not use the message "this soap stings" to predict the burning sensation caused by contact with the substance. Instead, the baby gradually *learns* to interpret such messages and make predictions for an entire universe of phenomena [@schulz2007preschool]. We refer to this state of affairs as *mastering a language*, and we aim to replicate it in a machine as the result of an analogous learning process. + +Using a batch of explanations paired with observations of several phenomena, we want to learn an interpreter to make predictions about novel phenomena for which we are given explanations in the same language. Going a step further, we also want to discover these explanations, when all we have is a handful of observations of the novel phenomena. We first describe the problem setup in the sequel, comparing it to existing ML problems; then we detail our approach in Sec. [4](#sec-crns){reference-type="ref" reference="sec-crns"}. + +Formally, let phenomena $P_1, P_2, P_3, \dots$ be subsets of a universe $U$, which is a large set with no special structure (i.e., all the possible observations $U = \{x_1, \dots, x_z\}$). Over a universe $U$, one can define a language $L$ as a pair $(\Sigma_L, \mathcal{I}_L)$, where $\Sigma_L$ is a finite collection of short strings over some alphabet $A$, with $|\Sigma_L| \gg |A|$, and $\mathcal{I}_L$ is a binary function $\mathcal{I}_L: U \times \Sigma_L \rightarrow \{ 0, 1 \}$, which we call *interpreter*. We say that a phenomenon $P_i$ is *explainable* in a language $L$ if there exists a string $e \in \Sigma_{L}$ such that, for any $x \in U$, it occurs $\mathcal{I}_{L}(x, e) = \mathbf{1}_{P_i}(x)$, where $\mathbf{1}_{P_i}(x)$ is the indicator function of $P_i$. We call the string $e$ an explanation, in the language $L$, for the phenomenon $P_i$. + +Our first contribution is the introduction of a new class of machine learning problems, which we refer to as *Explanatory Learning* (EL). + +Consider the general problem of making a new prediction for a phenomenon $P_0\subset U$. In our setting, this is phrased as a binary classification task: given a sample $x' \in U$, establish whether $x' \in P_0$ or not. We are interested in two instances of this problem, with different underlying assumptions: + +- **The communication problem: we have an explanation**. We are given an explanation $e_0$ for $P_0$, in an unknown language $L$. This means that we do not have access to an interpreter $\mathcal{I}_L$; $e_0$ looks like Japanese to a non-Japanese speaker. Instead, we are also given other explanations $\{e_1, \dots, e_n\}$, in the same language, for other phenomena $P_1, \dots, P_n$, as well as observations of them, i.e., datasets $\{D_1, \dots, D_n\}$ in the form $D_i = \{(x_1, \mathbf{1}_{P_i}(x_1)), \dots, (x_m, \mathbf{1}_{P_i}(x_m))\}$, with $m \ll |U|$. Intuitively, here we expect the learner to use the explanations paired with the observations to build an approximated interpreter $\mathcal{\hat{I}}_L$, and then use it to make the proper prediction for $x'$ by evaluating $\mathcal{\hat{I}}_L(x', e_0)$. + +- **The scientist problem: we do not have an explanation**. We are given explanations $\{e_1, \dots, e_n\}$ in an unknown language $L$ for other phenomena $P_1, \dots, P_n$ and observations of them $\{D_1, \dots, D_n\}$. However, we do not have an explanation for $P_0$; instead, we are given just a small set of observations $D_0 = \{(x_1, \mathbf{1}_{P_0}(x_1)), \dots, (x_k, \mathbf{1}_{P_0}(x_k))\}$ and two guarantees, namely that $P_0$ is explainable in $L$, and that $D_0$ is *representative* for $P_0$ in $L$. That is, for every phenomenon $P \neq P_0$ explainable in $L$ there should exist at least a $x_i\in D_0$ such that $\mathbf{1}_{P_0}(x_i) \neq \mathbf{1}_{P}(x_i)$. Again, we expect the learner to build the interpreter $\mathcal{\hat{I}}_L$, which should first guide the search for the missing explanation $e_0$ based on the clues $D_0$, and then provide the final prediction through $\mathcal{\hat{I}}_L(x', e_0)$. + +Several existing works fall within the formalization above. The seminal work of [@Angluin] on learning regular sets is an instance of the scientist problem, where finite automata take the role of explanations, while regular sets are the phenomena. More recently, CLEVR [@clevr] posed a communication problem in a universe of images of simple solids, where explanations are textual and read like *"There is a sphere with the same size as the metal cube"*. Another recent example is CLIP [@clip], where 400,000,000 captioned internet images are arranged in a communication problem to train an interpreter, thereby elevating captions to the status of explanations rather than treating them as simple labels[^2]. With EL, we aim to offer a unified perspective on these works, making explicit the core problem of learning an interpreter purely from observations. + +We briefly discuss the relationship between EL and other problems in ML, pointing to Sec. [6](#sec:related){reference-type="ref" reference="sec:related"} for additional discussion on the related work. + +EL can be framed in the general meta-learning framework. The learner gains experience over multiple tasks to improve its general learning algorithm, thus requiring fewer data and less computation on new tasks. However, differently from current meta-learning approaches [@hospedales], we are not optimizing for any meta-objective. Instead, we expect the sought generality to be a consequence of implicitly defining an interpreter through a limited set of examples rather than an explicit goal to optimize for. + +To many, the concept of explanation may sound close to the concept of program; similarly, the scientist problem may seem a rephrasing of the fundamental problem of Inductive Logic Programming (ILP) [@shapiro1981inductive] or Program Synthesis (PS) [@balog2019deepcoder]. This is not the case. ILP has the analogous goal of producing a hypothesis from positive/negative examples accompanied by background knowledge. Yet, ILP requires observations to be expressed as logic formulas, a task requiring a human; only then the ILP solver outputs an explanation in the form of a logic proposition, which in turn is interpreted by a human expert. With EL, data can be fed as-is without being translated into logic propositions, and a learned interpreter plays the expert's role. PS also admits raw data as input, it yields a program as output, and replaces the expert with a handcrafted interpreter; still, the sequence of symbols produced by a PS system only makes sense to a human (who designed the interpreter), not to the system itself. Instead, in EL, the interpreter is learned from data rather than hardcoded. An empirical comparison demonstrating the benefits of EL over PS is given in Sec. [5](#sec-exp){reference-type="ref" reference="sec-exp"}. + +Next we introduce Odeen, an environment and benchmark to experiment with the EL paradigm. + +:::: wrapfigure +r0.47 + +::: overpic +./figures/odeen-game.jpg +::: +:::: + +**Single game.** The inset shows a typical situation in a game of Odeen. The players look at a set of structures made of simple geometric figures. Each structure is tagged red or green according to a secret rule, and the players' goal is to guess this rule. In the example, the rule can not possibly be "A structure must contain at least one red square" since the fifth structure on the left does not contain a red square, but respects the rule (green tag). To win the game, a player must prove to know the rule by correctly tagging a large set of new structures[^3]. We made a simplified interactive version of Odeen available at [ https://huggingface.co/spaces/gladia/odeen](https://huggingface.co/spaces/gladia/odeen). + +**Odeen challenge.** We can see each game of Odeen as a different phenomenon of a universe, where each element is a sequence of geometric figures. In this universe, players are scientists like Galileo, trying to explain the new phenomenon; see Figure [1](#fig-galileo){reference-type="ref" reference="fig-galileo"}. We can phrase the challenge for an Odeen scientist in this way: make correct predictions for a new phenomenon given few observations of it in addition to explanations and observations of some other phenomena. This is the essence of the Odeen Explanatory Learning problem, see Figure [2](#fig-teaser){reference-type="ref" reference="fig-teaser"} (A and B). + +*- Why do we need explanations and observations from phenomena different from the one of interest? Indeed, we are able to play Odeen from the very first game.* + +*- We are able to do so only because we are -already- fluent in the Odeen language, which is a subset of English in the above case. We already have and understand all necessary concepts, such as being "at the right of" something, but also being a "square" or "at least". Otherwise, we would need past explanations and observations to first build this understanding. Before explaining the dynamic of the Jupiter moons, Galileo learned what "Jupiter\" is and what does it mean to "have a period around\" something from past explanations and examples provided to him by books and teachers.* + +:::: wrapfigure +r0.44 + +::: overpic +./figures/rules.png +::: +:::: + +In Odeen, consider the point of view of someone who does *not* speak the language in which the rules are written; an example of this is in the inset, where the secret explanations are given in hieroglyphics rather than English. Such a player would not be able to tag any structure according to the secret rule, even if the latter is given. However, assume the player has been watching several games together with their secret rules. Reasonably, the player will grow an idea of what those strange symbols mean. If the player then wins several Odeen games, it would be strong evidence of mastering the Odeen language. + +**Problem formulation.** Each game of Odeen is a different phenomenon $P_i$ of a universe $U$ whose elements $x$ are sequences of geometric figures. The specific task is to make correct predictions for a new phenomenon $P_0$ (a new game) given: (i) a few observations $D_0$ of $P_0$ (tagged structures), in conjunction with (ii) explanations $\{e_1, \dots, e_n\}$ and observations $\{D_1, \dots, D_n\}$ of other phenomena (other games and their secret rules). More formally: + +::: mdframed +Let us be given $s$ unexplained phenomena with $k$ observations each, and $n$ explained phenomena with $m$ observations each; let the $n$ phenomena be explained in an unknown language, i.e., $e_1, \dots e_n$ are plain strings without any interpreter. The task is to make $\ell$ correct predictions for each of the $s$ unexplained phenomena. +::: + +We consider $\ell=1176$ (1$\%$ of structures); $s\hspace{-0.2mm}=\hspace{-0.1mm}1132$; $k\hspace{-0.1mm}=\hspace{-0.1mm}32$; $m\hspace{-0.1mm}=\hspace{-0.1mm}10K$, $1K$, $100$; $n\hspace{-0.1mm}=\hspace{-0.1mm}1438$ or $500$. + +
+
+

./figures/zendo_teaser.png

+
+
Odeen Explanatory Learning problem. Given observations and explanations in an unknown language for some phenomena (A), plus a few observations of a new phenomenon, explain the latter and prove this knowledge by correctly tagging a large set of new samples (B). An empiricist approach attempts to extract this knowledge from data (C, left); a rationalist one conceives data as theory-laden observations, used to find the true explanation among a set of conjectures (C, right).
+
+ +**Why not explicitly ask for the rule?** Instead of requiring the player to reveal the secret explanation explicitly, we follow the principle of zero-knowledge proofs [@zero]. In our setting, this is done by asking the player to correctly tag many unseen structures according to the discovered rule. This makes it possible for any binary classification method to fit our EL environment without generating text. A winning condition is then defined by counting the correct predictions, instead of a textual similarity between predicted and correct explanation, which would require the player to guess word-by-word the secret rule. In fact, different phrasings with the same meaning should grant a victory, e.g., "at least one pyramid pointing up and at most one pyramid pointing up" is a winning guess for the secret rule "exactly one pyramid pointing up"[^4]. A brute-force enumeration of all equivalent phrasings, in turn, would not allow solutions like "exactly one *one* pyramid pointing up", where "one" is mistakenly repeated twice; intuitively, we want to accept this as correct and dismiss the grammatical error. Similarly, a solution like "exactly one pointing up", where "pyramid" is omitted, should be accepted in a universe where only pyramids point up. We will reencounter these examples in Sec. [5](#sec-exp){reference-type="ref" reference="sec-exp"} when we discuss the key properties of our approach. + +**Dataset generation.** Odeen structures are sequences of six elements including spaces, blues or reds, squares or pyramids, the latter pointing up or down. The size of the universe is $|U| = 7^6=117,649$ possible structures. We further created a small language with objects, attributes, quantifiers, logical conjunctions, and interactions (e.g., "touching", see Appendix A). The grammar generates $\approx$``{=html}25k valid rules in total. Each of the $|U|$ structures is tagged according to all the rules. The tagging is done by an interpreter implemented via regular expressions. + +**Metrics.** As described above, the task is to tag $\ell$ new structures for each of $s$ unexplained games. An EL algorithm addressing this task encodes the predicted rule as an $\ell$-dimensional binary vector $\textbf{v}$ per game (predicted vector), where $v_i=1$ means that + +:::: wrapfigure +r0.39 + +::: overpic +./figures/meetriche5.png (12,27)[Predicted vector v]{style="color: red"} +::: +:::: + +the $i$-th structure satisfies the predicted rule, and $v_i=0$ otherwise (see inset). Let $\textbf{w}^\ast$ be the ground-truth vector, obtained by tagging the $\ell$ structures according to the correct secret rule. Then, the Hamming distance $d_H(\textbf{v},\textbf{w}^\ast)$ measures the number of wrong tags assigned by the EL algorithm; if $d_H(\textbf{v},\textbf{w}^\ast)`{=html}25k rules, then the predicted rule $\textbf{v}$ made by the algorithm is deemed correct. + +According to this, the *Nearest Rule Score* (NRS) is the number of correctly predicted rules over a total of $s$ games. A second score, the *Tagging Accuracy* (T-Acc), directly counts the number of correct tags averaged over $s$ games; this is more permissive in the following sense. Consider two different rules $A$ and $B$ sharing $99\%$ of the taggings, and let $A$ be the correct one; if an EL model tags all the structures according to the *wrong* rule $B$, it still reaches a T-Acc of $99\%$, but the NRS would be $0$. An EL algorithm with these scores would be good at making predictions, but would be based on a wrong explanation. + +In principle, an EL problem like Odeen can be approached by training an end-to-end neural network to predict $\hat{y} = \mathbf{1}_{P_i}(x')$, given as input a set of observations $D_i$ and a single sample $x'$ (see Figure [2](#fig-teaser){reference-type="ref" reference="fig-teaser"} C, left). Such a model would assume that all the information needed to solve the task is embedded in the data, ignoring the explanations; we may call it a "radical empiricist" approach [@pearl2021radical]. A variant that includes the explanations in the pipeline can be done by adding a textual head to the network. This way, we expect performance to improve because predicting the explanation string can aid the classification task. As we show in the experiments, the latter approach (called "conscious empiricist") indeed improves upon the former; yet, it treats the explanations as mere data, nothing more than mute strings to match, in a Chinese room fashion [@chinese-room; @bender-koller-2020-climbing]. + +In the following, we introduce a "rationalist" approach to solve EL problems. This approach recognizes the given explanations as existing knowledge, and focuses on interpreting them. Here theory comes first, while the data become theory-laden observations. + +Our *Critical Rationalist Networks* (CRNs) tackle the EL scientist problem introduced in Sec. [2](#sec-el){reference-type="ref" reference="sec-el"}: to find $y=\mathbf{1}_{P_0}(x')$ given $x'$, $D_0$, $\{D_1, \dots, D_n\}$, $\{e_1, \dots, e_n\}$. They are formed by two independently trained models: + +\(i\) A stochastic *Conjecture Generator* $$\mathcal{CG}: \{ (x,\mathbf{1}_{P}(x))_j \}_{j=1}^k \mapsto e \,,$$ taking $k \leq |D_0|$ pairs $(x,\mathbf{1}_{P}(x)) \in D_i$ as input, and returning an explanation string $e \in \Sigma$ as output. $\mathcal{CG}$ is trained to maximize the probability that $\mathcal{CG}$$(\Tilde{D_i}) = e_i$ for all $i=1,\dots,n$, where $\Tilde{D_i} \subset D_i$ is a random sampling of $D_i$, and $|\Tilde{D_i}| = k$. + +\(ii\) A learned *Interpreter* $$\mathcal{I} : (e, x) \mapsto \hat{y}\,,$$ which takes as input a string $e \in \Sigma$ and a sample $x \in U$, to output a prediction $\hat{y} \in \{0,1\}$. $\mathcal{I}$ is trained to maximize the probability that $\mathcal{I}$$(e_i, x) = \mathbf{1}_{P_i}(x)$, with $i = 1, \dots, n$ and $(x,\mathbf{1}_{P_i}(x)) \in D_i$. + +At test time, we are given a trained $\mathcal{CG}$ and a trained $\mathcal{I}$, and we must predict whether some $x' \notin D_0$ belongs to $P_0$ or not. The idea is to first generate $t$ conjectures by applying $\mathcal{CG}$ $t$ times to the dataset $D_0$; then, each conjecture is verified by counting how many times the interpreter $\mathcal{I}$ outputs a correct prediction over $D_0$. The conjecture with the highest hit rate is our candidate explanation $\hat{e}_0$ for $P_0$. Finally, we obtain the prediction $\hat{y}'$ as $\mathcal{I}$$(\hat{e}_0, x')$. See Figure [3](#fig-implementation){reference-type="ref" reference="fig-implementation"} (left) for a step-by-step pseudo code. + +
+
+

./figures/impl.png (52,38)ℐ (47.5,3)𝒞𝒢 (91,1.2)EMP-R

+
+
Left: Test-time algorithm of CRNs. Right: CRNs are implemented using encoder-decoder transformers blocks, details of the parameters in Appendix 9. Right-top: ℐ denotes the interpreter model (rule encoder and label decoder). Right-bottom: The conjecture generator 𝒞𝒢 is composed by blue blocks. The “radical empiricist” (EMP-R) is composed by orange blocks. The “conscious empiricist” (EMP-C) baseline model consists of all the transformer blocks in the right-bottom figure, board encoder with rule and label decoders (all the blue and orange blocks).
+
+ +The interpreter $\mathcal{I}$ is a crucial component of our approach. A poor $\mathcal{I}$ may fail to identify ${e}_0$ among the generated conjectures, or yield a wrong prediction $y'$ when given the correct ${e}_0$. On the other hand, we can work with a $\mathcal{CG}$ of any quality and safely return as output an *unknown* token, rather than a wrong prediction, whenever ${e}_0$ does not appear among the generated conjectures. The role of $\mathcal{CG}$ is to trade-off performance for computational cost, and is controlled by the parameter $t$. Larger values for $t$ imply more generated conjectures, corresponding to exhaustive search if taken to the limit (as done, e.g., in @clip). This potential asymmetry in quality between $\mathcal{CG}$ and $\mathcal{I}$ is tolerated, since the learning problem solved by $\mathcal{CG}$ is generally harder. + +Secondly, although a CRN is implemented using neural networks, as we shall see shortly, its working hypothesis does not coincide with a snapshot of the countless network's parameters; rather, the working hypothesis is but the small conjecture analyzed at a given moment. This way, the CRN hypothesis is detached from the model and can only be accepted or refused in its entirety, rather than being slightly adjusted at each new data sample (Figure [2](#fig-teaser){reference-type="ref" reference="fig-teaser"} C, the hypotheses are in orange). diff --git a/2204.08453/main_diagram/main_diagram.drawio b/2204.08453/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..ac66d86f1cbe31e09594d61db5782598ed8ad5c0 --- /dev/null +++ b/2204.08453/main_diagram/main_diagram.drawio @@ -0,0 +1 @@ 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 \ No newline at end of file diff --git a/2204.08453/main_diagram/main_diagram.pdf b/2204.08453/main_diagram/main_diagram.pdf new file mode 100644 index 0000000000000000000000000000000000000000..2aba15be38236cdd030d5415676138322dc4a0f2 Binary files /dev/null and b/2204.08453/main_diagram/main_diagram.pdf differ diff --git a/2204.08453/paper_text/intro_method.md b/2204.08453/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..786979b540a44fb2ec83dce21f1191ab13b6bde0 --- /dev/null +++ b/2204.08453/paper_text/intro_method.md @@ -0,0 +1,22 @@ +# Introduction + +In any form of digital communication, information is transmitted via a sequence of discrete symbols. This includes images and videos, even though they are inherently signals with two spatial dimensions (2D). The modus operandi for transmitting such signals is to (1) efficiently encode and quantize their values in the spatial or spectral domain, (2) linearize the signal to a one-dimensional (1D) sequence by using a standard scanning order such as raster, zig-zag, or Hilbert Curve order [@hilbert1935stetige], and finally (3) apply a Shannon [@shannon1948mathematical] style entropy coding technique such as Arithmetic coding [@witten1987arithmetic] or Huffman coding [@huffman1952method] to further compress the 1D sequence. Given the ubiquity of images and videos in our lives, a large amount of effort has gone into optimizing each of these steps of digital communication. The focus of this work is the second step of linearizing the 2D spatial signal to a 1D sequence. A continuous scan order that traverses all spatial locations in two or higher dimensional signals exactly once is also known as the space-filling curve (SFC) [@peano1890courbe]. + +
+ +
Given a set of images, a gif, or a video, Neural Space-filling Curves (SFC) can provide a more spatially coherent scan order for images as compared to universal scan orders such as S-curve, or Peano-Hilbert curves. As shown in the example of a trouser and a face, the scan line tends to cover the background before moving to the foreground. (SFCs generated here using half-resolution images and resized for clarity. Best viewed in color.)
+
+ +Prior works have proposed various space-filling curves (SFCs) in the last hundred years, most of them context-agnostic, *i.e.*, they are completely defined by the size and dimension of the space without taking into account spatial information of the space, *e.g.*, pixels in the case of two-dimensional images. These universal context-agnostic SFCs are typically defined recursively to ensure simplicity and scale. Some of the SFCs also have spatial coherence properties and have been used in various image-based applications [@moon2001analysis; @kamata1993implementation; @ansari1992image; @thyagarajan1991fractal]. + +However, in many applications such as video conferencing, health-care, or social media, the images being transmitted, are often repetitive with a similar layout and content with minor variations transmitted over and over again. GIFs are another great example that consists of highly repetitive content and need to be stored efficiently and often, losslessly. Since universal SFCs do not utilize the intrinsic information of image content, they are far from optimal for a single image or a set of images with repetitive structure (refer to [1](#fig:teaser){reference-type="ref+label" reference="fig:teaser"} for an example). Dafner *et al.* [@dafner2000context] proposed SFCs that exploit the inherent correlation between pixel values in an image. Our work improves upon Dafner *et al.* , in two aspects. + +- Instead of discovering a single SFC for every image independently, we propose a data-driven technique to find optimal SFCs for a set of images. We postulate that context-based SFCs are more suitable for linearizing a group of images (or a short video/gif), since the cost of storing the SFC itself can be amortized by the number of images. + +- We devise a novel alternating minimization technique to train an SFC weights generator, which allows us to optimize for any given objective function, even when not differentiable. + +To the best of our knowledge, ours is the first work to propose a machine learning method for computing context-based SFCs and opens new directions for future research on optimal scanning of 2D and 3D grid-based data structures such as images, videos, and voxels. We demonstrate both quantitatively and qualitatively the benefit of our approach in various applications. + +# Method + +We first describe the algorithm for computing SFC for one image as proposed by Dafner *et al.* [@dafner2000context] in [3.1](#sec:cbsfc){reference-type="ref+Label" reference="sec:cbsfc"}. We then extend this treatment to a more general setting where we can optimize the SFC for any non-differential objective function for multiple images in [3.2](#sec:nsfc){reference-type="ref+Label" reference="sec:nsfc"}. The rest of the [3](#sec:approach){reference-type="ref+Label" reference="sec:approach"} describes major components of our framework and training procedure in detail. diff --git a/2205.11029/main_diagram/main_diagram.drawio b/2205.11029/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..90c1bf28ef2a9f40a2144db9617c7f9805f4aa7f --- /dev/null +++ b/2205.11029/main_diagram/main_diagram.drawio @@ -0,0 +1 @@ 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 \ No newline at end of file diff --git a/2205.11029/paper_text/intro_method.md b/2205.11029/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..7936d621dfe530fe5b9cff0c1a70510f21133ddb --- /dev/null +++ b/2205.11029/paper_text/intro_method.md @@ -0,0 +1,145 @@ +# Introduction + +
+ +
An example of the GUI-based task-oriented dialogue system(GUI-TOD). The Action Executor will execute tasks on GUI and the system will generate a response based on the execution result.
+
+ +Recent years have witnessed the rapid development of task-oriented dialogue systems [@zhang2020recent; @ni2021recent; @chen2022dialogzoo; @chen2017survey]. They have been widely applied to customer support, booking system and especially intelligent personal assistant. These task-oriented dialogue systems work in a similar pipeline: firstly identify the user intent, then extract necessary information by the process of slot-filling. After getting enough information for the task, the agent will call the backend APIs (provided by APP developers) to fetch information, and then generate a response based on the query result. + +There are some drawbacks of this framework. Firstly, TODs rely on publicly accessible APIs or APIs designed for TODs to perform tasks, but such APIs may not exist in real-life APPs, which hinders the application of TODs. Secondly, a system should be customized to recognize the pre-defined API-related slots, which limits the generality. + +::: table* + Action Description + ---------------------------- --------------------------------------------------------------------- -- + `Click`(`item` = $x$) Click the item with index $x$ on the screen. + `Swipe`(`direction` = $x$) Swipe screen towards direction $x$, which includes "up" and "down". + `Input`(`text` = $x$) Input the text $x$ to the smartphone. + `Enter`( ) Press the "Enter" button on the keyboard. + `Clear`( ) Clear the current input box. + `Back`( ) Press the "back" button on the smartphone. + `End`( ) Turn has been finished and it will go to Response Generator module. +::: + +Consider how humans perform tasks on smartphones They don't need a parametric API but finish tasks by interacting with the GUI (graphical user interface), indicating that GUI is a more general interface. Previous studies explore how to translate natural language commands into GUI operations [@mazumder2021flin; @pasupat2018mapping; @xu2021grounding]. These studies focus on single query and step-by-step operations, while in real scenarios the query would be multi-turn interaction and there is no clear instruction about how to execute the task. Etan [@riva2021etna] and SUGILITE [@li2017sugilite] are two systems that support learning GUI operations from demonstrations, but these systems are script-based and are sensitive to the change in GUI and workflow. Duplex on the web [@duplex] can directly operate the website to perform the required task, for example booking a movie ticket. However, it only supports limited websites, and it's more a unified GUI interface than a task-oriented dialogue system that enables general GUI operation. + +To this end, we propose the task of GUI-based task-oriented dialogue system (GUI-TOD). It supports multi-turn conversation and direct GUI operation. All tasks would be performed on the GUI of real APPs, which means we no longer need TOD-specific APIs to communicate with APPs, and it would be possible to apply TOD on any APPs. Since there is no available benchmark published, We collect META-GUI, a dataset with dialogues and GUI traces on real Android APPs. A GUI trace is a series of GUI operations, including screenshots, Android view hierarchies as well as actions. Android view hierarchy is an XML-style file, which organizes the content of GUI through a hierarchical structure. It also contains the types of items on the screen and their bounding boxes. An example is shown in Appendix [10](#view hierarchy){reference-type="ref" reference="view hierarchy"}. When a user requests a task, the system should open the related APP and execute the task through multiple operations on GUI. It requires a comprehensive understanding of GUI structure and interaction logic. An interaction example is shown in Figure [1](#fig:system_example){reference-type="ref" reference="fig:system_example"}. + +We focus on building an agent with general ability to operate GUI, rather than optimize for specific APPs. Our proposed GUI-TOD system leverages both the visual information and textual information on the screen to predict the next action to be executed and generate the system response. Our experiments show that the GUI-TOD outperforms heuristic baselines by a large margin, with an action completion rate of 82.74%. + +Our contributions are followings: + +- We propose a GUI-based task-oriented dialogue system, which can perform tasks on mobile APPs through multiple operations on GUI. + +- We collect META-GUI, a dataset with dialogues and GUI operation traces serving as the benchmark for the proposed system. + +- We conduct thorough experiments on our dataset and validate the importance of multi-modal information and history information. We show that it is a promising task but needs further exploration. + +
+ +
The overview of GUI-based task-oriented dialogue system (GUI-TOD).
+
+ +The overview of GUI-TOD is shown in Figure [2](#fig:gui-tod){reference-type="ref" reference="fig:gui-tod"}. It consists of two sub-modules: Action Executor (AE) and Response Generator (RG). The traditional task-oriented dialogue system [@chen2017survey; @zhang2020recent; @yu2014cognitive] splits the task into natural language understanding (NLU) [@zhu2021few], dialogue manager (DM) [@chen2020schema; @zhu-etal-2020-efficient; @chen2018policy; @chen2019agentgraph; @chen2020distributed], and natural language generation (NLG) [@keskar2019ctrl]. We omit the NLU module and directly send user utterances to AE. The AE module has similar features with DM, it executes the requested task by interacting with the GUI for multiple rounds, while DM accomplishes this by calling TOD-specific APIs. The RG module will generate the system response based on the execution results, which is the same as NLG. The process of executing a task is a series of GUI operations, including click, swipe, etc. The task of AE module is action prediction, which aims at predicting the next action to be performed on GUI, and the RG module focuses on generating system's response after executing a task. A major improvement of GUI-TOD is that it does not rely on a pre-defined domain ontology. Conventionally, the DM module will identify a set of slot-value from the user utterance, which serves as the parameter for backend APIs. However, GUI-TOD handles task-specific slot-values during the execution of tasks. When the APP requires a certain input (for example, entering the time and destination), the system can obtain the information by understanding the current user utterance or generating a response for further asking. Compared with CUED actions [@young2007cued] in traditional TOD, actions in GUI-TOD are GUI-related operations rather than communication actions between user and system. + +Formally, the action prediction task can be defined as: given the GUI trace and dialogue history, predict the next action to be performed. We define the set of actions that can be performed on the APPs in Table [\[action_table\]](#action_table){reference-type="ref" reference="action_table"}. All the actions would take the form of $Action(parameter=*)$. There are seven types of $Action$, including six physical actions: *click, swipe, input, enter, clear, back*, and one virtual action: *end*. The corresponding parameters are listed in Table [\[action_table\]](#action_table){reference-type="ref" reference="action_table"}. The $end$ action is the last action for every GUI trace, which means the end of GUI operations. After an $end$ action is generated, the GUI-TOD would move to the RG module. We denote the *j*th action in turn *i* as $\mathcal{A}_{i,j}=\left( t, p \right)$, where *t* is the action type and *p* is the corresponding parameter. $\mathcal{S}_{i,j}=\left( s, v \right)$ is the *j*th screen in turn *i*, including the screenshot *s* and the view hierarchy *v*. The dialogue in turn *i* is represented as $\mathcal{D}_{i}=\left( U_i, R_i \right)$ where $U_i$ is the $i$th user utterance and $R_i$ is the $i$th system response. The action prediction task is formulated as: $$\begin{equation} + \mathcal{A}_{i,j}=\mathcal{F}\left( \mathcal{S}_{1:i, 1:j}, \mathcal{A}_{1:i,1:j-1}, \mathcal{D}_{1:i-1}, U_i \right), +\end{equation}$$ where $1:i$ means from turn $1$ to $i$, $\mathcal{F}$ is a trainable action model, which we discuss in [4.1](#actionmodel){reference-type="ref" reference="actionmodel"}. The RG module takes the GUI trace and dialogue history as input, then generates a response based on the execution result and context. Denote the set of actions in turn $i$ as $\mathcal{A}_i$, the screens in turn $i$ as $\mathcal{S}_i$, the response generation task is formulated as: $$\begin{equation} + \mathcal{R}_{i}=\mathcal{G}\left( \mathcal{S}_{1:i}, \mathcal{A}_{1:i}, \mathcal{D}_{1:i-1}, U_i \right), +\end{equation}$$ where $\mathcal{G}$ is the response generator model, which we discuss in [4.2](#responsemodel){reference-type="ref" reference="responsemodel"}. + +Our dataset consists of two kinds of data: dialogues and GUI operation traces. In each dialogue, user would ask the agent to complete a certain task through multi-turn interaction. Our tasks involve six different domains: weather, calendar, search, taxi, hotel and restaurant. In this paper, we consider APPs that accomplish the same kind of tasks to be in the same domain. To enhance the diversity of our dataset, we use multiple Apps from the calendar and weather domains. The details of APPs are listed in Appendix [8](#app_data){reference-type="ref" reference="app_data"}. + +
+ +
The distribution of the total number of items versus the clicked one for each item type.
+
+ +We collected our data in two-stage: first we collected GUI traces for existing dialogues, then we collected both dialogues and GUI traces. + +In the first stage, we provided dialogues to annotators and instructed them to perform tasks on real APPs. We started from extracting dialogues from the SMCalFlow dataset [@andreas2020task]. SMCalFlow contains multi-turn task-oriented dialogues, which is known for complex reference phenomenon that requires a comprehensive understanding of context. We extract dialogues from calendar, weather and search domains. Six annotators were recruited to label the GUI traces. We built a web-based annotation system, which was connected to a real Android smartphone (see Appendix [9](#annotation system){reference-type="ref" reference="annotation system"}). Annotators can see the current screen of the smartphone in the system, and control the smartphone by clicking buttons. A dialogue would be shown in the system. Annotators should first read the dialogue, then they were allowed to explore how to finish the task (e.g. check the weather) on smartphone. If the task requirement in the dialogue conflicted with the real-world scenario (for example, creating an event in the past), the annotators could change the content of the dialogue to make the task achievable. After they were ready, they need to use the annotation system to record the actual process of executing the task. Each operation would be recorded, and the screenshot after each operation was also saved together with the view hierarchy. + +In the second stage, we collected dialogues and GUI traces for domains of hotel, restaurant and taxi. Because there are no available dialogues of these domains in previous datasets, we asked annotators to write new dialogues. We selected three experienced annotators from the last stage. Different from the last stage, the annotator was shown a task objective, which was generated randomly from all available conditions in APPs. The annotators should act as user and system alternatively to write dialogues according to the task objectives. To avoid annotators writing short and simple dialogues, we added constraints about the number of turns and the behaviors in dialogue, e.g. adding a condition or changing a condition. An example of the generated target is shown in Appendix [12](#exampleoftarget){reference-type="ref" reference="exampleoftarget"}. After writing dialogues, the annotators should also record the corresponding GUI operation traces for each turn, which is the same as the last stage. + +After annotation, we manually reviewed the data. The checklist includes: whether the recorded GUI traces match the dialogues, whether there are invalid operations due to the system error or misoperation, and whether there are redundant operations in the GUI trace. We manually fixed annotations that only have small mistakes, and discarded the task requiring significant modification. The dialogue level pass rate is about $63.6\%$, and finally we got 1125 dialogues in total. For more information, please refer to Appendix [11](#datareview){reference-type="ref" reference="datareview"}. + +The dialogues collected in the second state were created by three annotators, which lack diversity in expression. Therefore, we published a dialog rewritten task on AMT[^1] (Amazon Mechanical Turk) to polish the dialogues. + +
+ +
The distribution of actions.
+
+ +During GUI trace annotation, some APPs can not obtain valid Android hierarchy. To handle this problem, we used the online Optical Character Recognition (OCR) service, provided by Baidu Cloud [^2], to detect all texts on the image with their corresponding positions and generate a pseudo layout file. + +
+ +
The illustration of our proposed model. There are five parts in this figure: (1) encoder; (2) image feature extraction; (3) multi-modal information fusion; (4) the Action Module; (5) the Response Module.
+
+ +We extract items from screen using the corresponding layout file. An item is a clickable leaf node. Similar to [@zhou2021large], we consider an item to be clickable if its `clickable` attribute is true or its parent node is clickable. An item consists of text content, item type and bounding box. We extract the text content of an item by looking at its `text` property first. If it is empty, we use its `content-desc` attribute, otherwise we would use the `resource-id` property. Based on the extracted items, we can locate the target item for the `click` action by comparing the click position and the bounding boxes of items. + +The total number of dialogues in our dataset is 1125, including 4684 turns. The average number of images for each turn is 5.30, and the average number of words for each utterance is 8. On average, there are 23.80 items for each image, and the item text length is 2.48 words. The distribution of item types is shown in Figure [3](#fig:item_type){reference-type="ref" reference="fig:item_type"}. We also provide an example for each item type in Appendix [13](#exampleofitemtypes){reference-type="ref" reference="exampleofitemtypes"}. It is clear that `TextView` and `ImageView` are the two most frequent type, which indicates that our dataset is informative. + +The distribution of actions is listed in Figure [4](#fig:action_distribution){reference-type="ref" reference="fig:action_distribution"}. The `click` is the most frequent action, while `clear` is the least action for the reason that only a small number of tasks require clearing the current input box. For `click` action, we further compute the type distribution of target items, which is shown in Figure [3](#fig:item_type){reference-type="ref" reference="fig:item_type"}. `TextView` and `Button` type are mostly clicked, while there are 8 item types never been operated. This implies that the item types may supply some hints for predicting the target items. Besides, the average numbers of words for `response` and `input` action are 9 and 3 respectively. + +The overview of our system is illustrated in Figure [5](#fig:armodel){reference-type="ref" reference="fig:armodel"}. It's composed of four components: encoder, image feature extractor, multi-modal information fusion module and the output module. The output module can be the Action Module or the Response Module. + +We call the combination of encoder, image feature extractor, multi-modal information fusion module and the Action Module as Action Model, which is used to predict the next GUI action based on the history. Next, we will describe these modules respectively. For simplify, for the screen history we only consider the last screen here. We will discuss adding more screen histories later. + +The input of encoder consists of two parts: dialog history $\{\mathcal{D}_{1:i-1}, U_i\}=\{w_{1},...,w_n\}$ and texts in the items $\{m_{1,1:l_1}, \dots, m_{k,1:l_k}\}$. Items are extracted from the last screen, $k$ is the number of items and $l_i$ is the length of the $i$th item's text: $$\begin{align} +\begin{split} + X &= \{ w_{1:n};m_{1,1:l_1}, \dots, m_{k,1:l_k}\}, \\ + \textbf{H} &= \text{TransformerEncoder}(X), \\ +\end{split} +\end{align}$$ where $\textbf{H} = \left[\textbf{D};\textbf{M}\right]$ and $\textbf{D}=\{ \textbf{w}_1, \textbf{w}_2, \dots, \textbf{w}_n\}$ represents encoder outputs of the dialogue history, $\textbf{M}=\{ \textbf{m}_{1,1:l_1}; \dots; \textbf{m}_{k,1:l_k} \}$ represents encoder outputs of item texts. + +Given a screenshot and its corresponding layout file, we use Faster R-CNN [@ren2015faster] to extract the feature map. Then we apply ROI pooling based on the bounding box of each item, and get the item-level image features $\mathbf{I}=\{\mathbf{I}_1, ..., \mathbf{I}_k\}$. + +Given the encoder output and the regional image feature extracted above, we concatenate them together. The text features from one item $\textbf{m}_{i,1:l_k}$ are concatenated with the same item feature $\mathbf{I}_i$, and the $\mathbf{w}_{1:n}$ are concatenated with zeros. Then we use a Transformer encoder with $M$ layers to fuse the multi-modal features. For each layer, to enhance the image information, we will concatenate the image features and the output from the last layer again to form the input for the next layer. + +For the Action model, we need to predict the action type and its corresponding parameters. As shown in Table [\[action_table\]](#action_table){reference-type="ref" reference="action_table"}, there are 7 action types with 3 different parameters. We show some examples of parameter predictions in Appendix [14](#parameterpredictions){reference-type="ref" reference="parameterpredictions"}. + +We use the encoder output of the `[CLS]` token for action type prediction. We apply a feed-forward network followed by a Softmax layer to predict the action type: $$\begin{align} +\begin{split} + \mathbf{p}_a &= \text{Softmax}(\text{FFN}_1(\mathbf{E_{[CLS]}})), +\end{split} +\end{align}$$ where $\mathbf{p}_{a}$ is the probability distribution of action, and FFN represents the Feed-Forward Network. + +For the action parameter, we use three different classifiers: + +1\) *Input Text Prediction*  We assume that the input to the APPs must be part of the user utterance, so we formulate the prediction of input text as a span prediction task. We use two classifiers to predict the begin and end positions in the dialogue: $$\begin{equation} + \begin{split} + \mathbf{p}_{ds}=\text{FFN}_2(\mathbf{D}), + \mathbf{p}_{de}=\text{FFN}_3(\mathbf{D}),\\ + \end{split} +\end{equation}$$ where the $\mathbf{p}_{ds}$ and $\mathbf{p}_{ds}$ are the probability of start and end position respectively. + +2\) *Target Item Prediction*  The target item classifier is based on the encoding outputs of items. We first computed the item representation by applying average pooling on the encoding outputs, then we use a feed-forward layer to compute the probability of selecting an item followed by a Softmax layer: $$\begin{equation} + \begin{split} + \bar{\textbf{m}}_i =& \ \text{Avgpooling}(\textbf{m}_{i,1:l_i}) \ \ 1 \leq i \leq k, \\ + \bar{\textbf{m}} =& \left[\bar{\textbf{m}}_1, \dots, \bar{\textbf{m}}_k\right]. \\ + \mathbf{p}_{m} =& \ \text{Softmax}(\text{FFN}_4(\bar{\textbf{m}})), + \end{split} +\end{equation}$$ where $\mathbf{p}_{m}$ is the probability distribution of items. + +3\) *Direction Prediction*  The direction classifier is a two-classes classification layer for the direction *up* and *down*: $$\begin{equation} + \mathbf{p}_d = \text{Softmax}(\text{FFN}_5(\mathbf{E_{[CLS]}})), +\end{equation}$$ where $\mathbf{p}_{d}$ is the probability distribution of swipe direction. + +According to the task definition, besides dialogue histories, we can still use action histories and screen histories. To verify this, we add them to the action model. For action histories, we regard action types as special tokens and add them to the dictionary. We concatenate the most recent $H$ action types $\{t_{1:H}\}$ before the dialogue history as input: $$\begin{equation} + X = \{t_{1:H};w_{1:n};m_{1,1:l_1},\dots, m_{k,1:l_k}\}, +\end{equation}$$ where $X$ stands for the input of Encoder, $t$ represents the action type. + +For screenshot histories, we encode all the screenshot in a recurrent way. Assume $\hat{\mathbf{I}}_i=\left[\mathbf{I}_{i,1},...,\mathbf{I}_{i,k}\right]$ is the image feature for $i$th screenshot, and $\bar{\mathbf{I}}_i$ is the history image feature for time step $i$. We compute $\bar{\mathbf{I}}_{i+1}$ by: $$\begin{equation} +\begin{split} + \bar{\mathbf{I}}_{i+1} = \text{Attn}(\mathbf{W}_1&\hat{\mathbf{I}}_{i+1},\mathbf{W}_2\bar{\mathbf{I}}_{i},\mathbf{W}_3\bar{\mathbf{I}}_{i}), \\ + &1 \leq i \leq H-1, +\end{split} +\end{equation}$$ where $\bar{\mathbf{I}}_{1} = \hat{\mathbf{I}}_{1}$, $H$ is the length of history, $\text{Attn}$ is the attention mechanism [@vaswani2017attention], and $\mathbf{W}_*$ are trainable parameters. We use the $\bar{\mathbf{I}}_{H}$ to replace the image features in Figure [5](#fig:armodel){reference-type="ref" reference="fig:armodel"}. + +The Response Model aims to generate the response to user. We use the Response Module as the output module and the other parts are the same as Action Model. Considering the prediction of response is mainly decided by the execution results and dialogues, we do not use action histories for the Response Model. For the Response Module, we use a Transformer Decoder with N layers: $$\begin{align} +\begin{split} + \textbf{R} = \text{TransformerDecoder}(\left[\textbf{D};\textbf{M}\right]), +\end{split} +\end{align}$$ where $\textbf{R}$ represents the predicted response text. diff --git a/2206.07043/main_diagram/main_diagram.drawio b/2206.07043/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..79160d095ef9c5d5284f8de11d1aa95bfe980055 --- /dev/null +++ b/2206.07043/main_diagram/main_diagram.drawio @@ -0,0 +1 @@ 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 \ No newline at end of file diff --git a/2206.07043/main_diagram/main_diagram.pdf b/2206.07043/main_diagram/main_diagram.pdf new file mode 100644 index 0000000000000000000000000000000000000000..f91961228b227d0854543a9d325bb10a5acf07b0 Binary files /dev/null and b/2206.07043/main_diagram/main_diagram.pdf differ diff --git a/2206.07043/paper_text/intro_method.md b/2206.07043/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..57650e0d0336ffb363c6c50a52b81f68d70908fe --- /dev/null +++ b/2206.07043/paper_text/intro_method.md @@ -0,0 +1,108 @@ +# Introduction + +After revolutionizing the field of machine translation [@sutskever2014sequence; @cho-etal-2014-learning; @bahdanau2014neural], sequence-to-sequence (seq2seq) methods have quickly become the standard approach for not only multilingual but also for *monolingual* sequence transduction / text generation tasks, such as text summarization, style transfer, and grammatical error correction. While delivering significant quality gains, these models, however, are prone to hallucinations [@maynez-etal-2020-faithfulness; @pagnoni-etal-2021-understanding]. The seq2seq task setup (where targets are generated from scratch word by word) overlooks the fact that in many monolingual tasks the source and target sequences have a considerable overlap, hence targets could be reconstructed from the source inputs by applying a set of edit operations. + +Text-editing models attempt to address some of the limitations of seq2seq approaches and there has been recently a surge of interest in applying them to a variety of monolingual tasks including text simplification [@dong-etal-2019-editnts; @mallinson-etal-2020-felix; @agrawal-etal-2021-non], grammatical error correction [@awasthi-etal-2019-parallel; @omelianchuk-etal-2020-gector; @malmi-etal-2019-encode; @stahlberg-kumar-2020-seq2edits; @rothe-etal-2021-simple; @chen-etal-2020-improving-efficiency; @hinson-etal-2020-heterogeneous; @gao2021hierarchical], sentence fusion [@malmi-etal-2019-encode; @mallinson-etal-2020-felix] (see an example in Figure [1](#fig:example){reference-type="ref" reference="fig:example"}), MT automatic post-editing [@gu2019levenshtein; @zietkiewicz2020post; @mallinson-etal-2020-felix], text style transfer [@reid-zhong-2021-lewis; @malmi-etal-2020-unsupervised], data-to-text generation [@kasner-dusek-2020-data], and utterance rewriting [@liu-etal-2020-incomplete; @DBLP:conf/sigir/VoskaridesLRKR20; @Hct2022]. + +
+ +
An example of using a text-editing approach to solve a sentence-fusion task.
+
+ +Text-editing approaches claim to be more accurate or on-par with seq2seq baselines especially in low resource settings, less prone to hallucinations and faster at inference time. These advantages have generated a substantial and continued level of interest in text-editing research. The goal of this tutorial is to provide the first comprehensive overview of the family of text-editing approaches and to offer practical guidelines for applying them to a variety of text-generation tasks. + +The tutorial is intended for researchers and practitioners who are familiar with generic seq2seq text-generation methods, such as Transformer [@vaswani2017attention] and pre-trained language models like BERT [@devlin-etal-2019-bert]. However, prior experience with text-editing models is not required to be able to follow the tutorial. + +We expect the topic to attract people in both academia and industry. The high-sample efficiency and low-computational requirements of text-editing models [@malmi-etal-2019-encode; @mallinson-etal-2020-felix] makes them an attractive baseline, e.g., for researchers developing new text-generation tasks for which large training sets do not yet exist. Moreover, the high-inference speed of text-editing methods, owing to their often non-autoregressive architecture [@awasthi-etal-2019-parallel; @mallinson-etal-2020-felix], makes them suitable for building real-time applications. + +The structure of the tutorial with duration estimates for different sections are shown in Table [1](#tab:structure){reference-type="ref" reference="tab:structure"}. Below we provide brief descriptions for each section. + +::: {#tab:structure} + **Section** **Duration** + ------------------------------------------------- -------------- + Introduction 15 min + $\quad$ What are text-editing models? + $\quad$ Text-editing vs. seq2seq models + Model design 40 min + $\quad$ Example model + model landscape + $\quad$ Edit-operation types + $\quad$ Tagging architecture + $\quad$ Auto-regressiveness + $\quad$ Converting target texts to target edits + Applications 45 min + $\quad$ Overview + $\quad$ Grammatical Error Correction + $\quad$ Text Simplification + $\quad$ Unsupervised Style Transfer + $\quad$ Incomplete Utterance Rewriting + Controllable generation 25 min + $\quad$ Mitigating hallucinations + $\quad$ Controllable dataset generation + Multilingual text editing 25 min + $\quad$ Tokenization + $\quad$ Handling morphology + $\quad$ Practical aspects + Productionization 25 min + $\quad$ Latency + $\quad$ Sample efficiency + Recommendations and future directions 5 min + **Total** 180 min + + : Tutorial structure and duration of each section. +::: + +We first define the family of text-editing methods: Text-editing models are sequence-transduction methods that produce the output text by predicting edit operations which are applied to the inputs. In contrast, the traditional seq2seq methods produce the output from scratch, token by token. We summarize the main pros and cons of these two approaches and provide guidelines for choosing which approach is more suitable for a given task. + +The similarities and differences of a set of popular text-editing methods will be analyzed in terms of the types of edit operations they employ, their tagging architecture, and whether they are auto-regressive or feedforward. We also discuss methods for converting target texts into target edit sequences, a task which often does not have a unique solution. Table [\[tab:methods\]](#tab:methods){reference-type="ref" reference="tab:methods"} provides a summary of the similarities and differences between the methods covered in the tutorial. + +::: table* +::: + +A key criterion for determining whether text-editing models are a good fit for a given application is the average degree of overlap between source and target texts. The higher the overlap, the more input tokens can be reused to generate the target, thus resulting in a simpler edit sequence. We give an overview of applications with a high degree of overlap to which text-editing methods have been applied to. Then we do a deep dive in to the following applications: grammatical error correction, text simplification, unsupervised style transfer, and incomplete utterance rewriting. + +Text-editing models with a restricted vocabulary of phrases to insert [@malmi-etal-2019-encode; @Hct2022] or with linguistically informed suffix-transformation operations [@awasthi-etal-2019-parallel; @omelianchuk-etal-2020-gector] are less prone to different types of hallucination since the models cannot produce arbitrary outputs. Moreover, the restricted vocabulary makes it feasible to manually refine the list of phrases that the model can insert. Another route through which the decomposition of the generation task into explicit edit operations can improve controllability is via biasing of certain types of edits to control how often the model will insert new text [@dong-etal-2019-editnts; @omelianchuk-etal-2020-gector]. Controllable generation with editing models can be useful for generating large synthetic datasets with a desired distribution of errors, which yields improvements in tasks such as grammatical error correction [@stahlberg-kumar-2021-synthetic]. We will provide concrete examples of the aforementioned control measures and their effects. + +Most text-editing models, like text-generation models in general, are evaluated on English, but there are also methods evaluated or specifically developed for other languages, including Chinese [@hinson-etal-2020-heterogeneous; @liu-etal-2020-incomplete], Czech [@naplava-straka-2019-grammatical], German [@mallinson-etal-2020-felix], Russian [@stahlberg-kumar-2020-seq2edits], and Ukrainian [@syvokon2021uagec]. Apart from general tokenization-related challenges discussed in [@mielke2021between], an additional challenge with applying text-editing methods to morphologically rich languages is a potential mismatch between the subword tokens, on which the underlying sequence labeling model operates, and the morphemes or affixes, on which the edits should happen. Possible solutions to this challenge include developing custom inflection operations [@awasthi-etal-2019-parallel; @omelianchuk-etal-2020-gector] or learning them from the data [@straka2021character], and using more fine-grained edit operations, such as character-level edits [@gao2021hierarchical]. + +An additional challenge when building a truly multilingual model---as opposed to one model per language---is to ensure that it is not skewed towards a particular language or a set of languages [@chung-etal-2020-improving] while being computationally efficient. + +We discuss how casting a text-generation problem as a text-editing task often allows the use of significantly faster and more data-efficient model architectures, without sacrificing output quality. We make use of the TensorFlow Profiler[^2] to compare latencies of text-editing and non-text-editing solutions for an example problem, and illustrate where the time savings come from. + +
+ +
Proposed flowchart for deciding when to try a text-editing approach.
+
+ +We provide practical guidelines for when to use (and when not to use) text-editing methods (see Figure [2](#fig:rec){reference-type="ref" reference="fig:rec"} for a summary). We also outline possible future directions which include: ($i$) learned edit operations, ($ii$) studying the effects of different subword segmentation methods since these typically determine the granularity at which the edit operations are applied, ($iii$) text-editing-specific pre-training methods, ($iv$) sampling strategies for text-editing methods, and ($v$) studying the effects of scaling up text-editing methods, a strategy that has been found to be very effective for many other text-generation methods [@brown2020language; @chowdhery2022palm]. + +A significant portion of the tutorial is devoted to discussing multilingual text-editing, including applying text-editing models to morphologically rich languages which presents specific challenges related to larger vocabularies and the need to edit word affixes. The presenters come from both academia and industry, are native speakers of 8 languages based in 4 different countries (Switzerland, Germany, Canada, USA), and are of different seniority levels from a PhD student to a Senior Staff Research Scientist. + +Before the tutorial, we expect the audience to read [@vaswani2017attention] and [@devlin-etal-2019-bert]. For references to text-editing works that will be discussed in the tutorial, see Table [\[tab:methods\]](#tab:methods){reference-type="ref" reference="tab:methods"}. + +50% of the methods that will be discussed in the tutorial (cf. Table [\[tab:methods\]](#tab:methods){reference-type="ref" reference="tab:methods"}) are developed by different subsets of the tutorial instructors. + +is a Senior Research Scientist at Google Switzerland. His research is focused on developing text-generation models for grammatical error correction and text style transfer. He received his PhD from Aalto University, Finland, where he also taught a course on Recent Advances in Natural Language Generation in Spring 2022. + +is a final-year PhD student in CS at McGill University and Mila, Canada. Her research is focused on conditional text generation. She is a co-organizer for the NewSum workshop at EMNLP 2021 and ENLSP workshop at NeurIPS 2021. + +is a Research Engineer at Google Switzerland. His research is focused on low-latency text-to-text generation. He received his PhD from the University of Edinburgh, Scotland. + +is a Research Engineer at Google Switzerland. His current research focuses on multi-lingual NLG. He organized [workshops](http://scai-workshop.github.io) and conducted [tutorials](https://clickmodels.weebly.com/tutorials.html) at conferences such as SIGIR, EMNLP, and IJCAI. Aleksandr received his PhD from University of Amsterdam, The Netherlands. + +is a Research Engineer at Google Switzerland focusing on grammatical error correction and low-latency models. He received his MSc from Jagiellonian University. + +is a Research Engineer at Google Switzerland working on text editing with application to grammatical error correction. He received his PhD from the University of Trento, Italy. + +is a Research Scientist at Google focusing on grammatical error correction and text style models. He received his PhD from Cambridge University, UK. + +is a Senior Research Engineer at Google Switzerland. His work is focused on multi-lingual rewriting of questions in low-latency settings. Sebastian received his PhD in Engineering from the Technical University of Berlin, Germany. + +is a Senior Staff Research Scientist at Google leading a research team working on speech and language algorithms. He received his PhD from the Johns Hopkins University, US. + +is a Staff Research Scientist at Google Switzerland leading an applied research team working on next generation NLG solutions. He received his PhD from University of Trento, Italy. + +Text-generation methods have the potential to generate non-factual [@maynez-etal-2020-faithfulness; @pagnoni-etal-2021-understanding; @kreps2020all] and offensive content [@gehman-etal-2020-realtoxicityprompts]. Furthermore, training these models on uncurated data can lead to the models replicating harmful views presented in the training data [@bender2021dangers]. Text-editing models are also susceptible to these issues, but they have been shown to mitigate some of them. Specifically, they reduce the likelihood of different types of hallucination [@malmi-etal-2019-encode] and their higher sample efficiency [@malmi-etal-2019-encode; @mallinson-etal-2020-felix] enables more careful curation of the training data. The tutorial will discuss the ethical issues related to text generation and provide concrete examples on how text-editing models can help mitigate them. + +[^1]: Website: + +[^2]: diff --git a/2206.08194/main_diagram/main_diagram.drawio b/2206.08194/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..e1fb6e08ecf1ca6a41062680dbba942aa4b374c2 --- /dev/null +++ b/2206.08194/main_diagram/main_diagram.drawio @@ -0,0 +1 @@ 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 \ No newline at end of file diff --git a/2206.08194/main_diagram/main_diagram.pdf b/2206.08194/main_diagram/main_diagram.pdf new file mode 100644 index 0000000000000000000000000000000000000000..81f627f76a4a7a66b236977573f3c5116bc5a774 Binary files /dev/null and b/2206.08194/main_diagram/main_diagram.pdf differ diff --git a/2206.08194/paper_text/intro_method.md b/2206.08194/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..7decfee3550b43478486a6ad219d221267eb5311 --- /dev/null +++ b/2206.08194/paper_text/intro_method.md @@ -0,0 +1,196 @@ +# Introduction + +Due to their low acquisition latency and high precision, rotating LiDAR sensors are among the most prevalent sensors for autonomous vehicles [39]. The acquired sequences of 3D points exhibit a complex structure in which the temporal and spatial dimensions are entangled through the rotation of the sensor around a reference point in motion; see Figure 1. However, this structure is often not reflected in the formatting of open-access LiDAR datasets [3,25,29], which are discrete sequences of range images, or frames, each corresponding to a 360◦ degree arc around the sensor. Consequently, most LiDAR semantic segmentation methods operate on one or several such frames at the same time, in the image [12] or point cloud [51,48,43] format. However, waiting for an entire frame to be acquired introduces an unavoidable latency of more than 100ms on top of the processing time, excluding applications for high-speed or urban driving. In this paper, we address this issue by introducing (i) HelixNet, the largest available LiDAR dataset, and whose fine-grained point information allows for the realistic real-time evaluation of segmentation methods, and (ii) Helix4D, a spatio-temporal transformer designed for the efficient segmentation of LiDAR sequences. + +![](_page_1_Figure_2.jpeg) + +Fig. 1: Online LiDAR Segmentation. The 3D point sequences of rotating LiDAR data of our proposed dataset HelixNet follow a complex helix-like structure in space and time, represented in a by using the vertical axis for both time and elevation. We propose an efficient spatio-temporal transformer to process angular slices of data centered on the sensor's position. The slices are partitioned into voxels, each attending other voxels from past slices to build a large spatio-temporal receptive field b. Our proposed model can segment the LiDAR point stream c with state-of-the-art accuracy and in real-time. + +Our dataset HelixNet, has several key advantages compared to standard datasets such as SemanticKITTI [3], see Table 1. By organizing points with respect to sensor rotation and reporting their precise release times, we can accurately benchmark the real-time readiness of leading state-of-the-art LiDAR sequence segmentation algorithms. Furthermore, the pointwise sensor orientation allows us to split the data into slices of acquisition corresponding to a fraction of the sensor's rotation. These slices can be processed sequentially by our proposed network Helix4D, resulting in a lower acquisition latency and a more realistic scenario for autonomous driving. Based on a spatio-temporal transformer designed explicitly for LiDAR sequences, Helix4D is more than 50 times smaller than the current best semantic segmentation architectures and reaches state-of-the-art performance with significantly reduced latency. + +# Method + +We introduce HelixNet, a new large-scale and open-access LiDAR dataset intended for the evaluation of real-time semantic segmentation algorithms. In contrast to other large-scale datasets, HelixNet includes fine-grained data on sensor rotation and position, as well as point release time. + +General Characteristics. As seen in Figure 2, HelixNet contains 20 sequences of 3D points, each corresponding to 6 to 7 minutes of continuous acquisition, for a total of 129 minutes. Scanning was performed by an HDL-64E Velodyne rotating LiDAR [24] mounted on a mobile platform [36]. As shown in Figure 3, HelixNet covers multiple cities and a wide variety of environments such as a university campus, dense historical centers, and a highway interchange. With a total of 10 billion points across 78 800 frames and 8.85 billion individual labels, HelixNet is the largest densely annotated open-access rotating LiDAR dataset by a factor of 1.7 as shown in Table 1. HelixNet follows the file format of SemanticKITTI [3], allowing researchers to evaluate existing code with minimal effort. + +![](_page_4_Figure_2.jpeg) + +Fig. 3: Extracts from HelixNet. Our proposed dataset contains various urban scenes from motorway to pedestrian plazas and historical centers. In the first row, we represent extracts of 15 to 30s of acquisition colored according to the point release time. In the second row, we represent the point semantic labels. + +We use a 9-classes nomenclature: road (16.4% of all points), other surface (22.0%), building (31.3%), vegetation (8.5%), traffic signs (1.6%), static vehicle (4.9%), moving vehicle (2.1%), pedestrian (0.9%), and acquisition artifact (0.05%). Points without labels correspond to either un-annotated (6.2%) parts of the clouds due to their ambiguity, or point without echos (6.1%). Compared to fine-grained classes such as the ones used by SemanticKITTI [3] or Paris-Lille3D's [38], our focused nomenclature limits class imbalance and makes macro-averaged metrics more stable. + +Each point is associated with the 9 following values: (1-3) Cartesian coordinates in a fixed frame of a reference, (4-6) cylindrical coordinate relative to the sensor at the time of acquisition, (7) intensity, (8) fiber index, and (9) packet output time. As detailed in the next paragraph, the last two features are not typically available in large-scale datasets and cannot be inferred. + +Sensor-Based Timing and Grouping. A rotating LiDAR consists of a set of lasers—or fibers—arranged on a rotating sensor head. The lasers send periodic pulses of light whose return times give the position of the impact points relative to the sensor. In the context of autonomous driving, these sensors are typically deployed on a moving platform and capture 3D points with centimetric accuracy. The sensor releases the data stream as a discrete temporal sequence of packets of 3D points. For an HDL-64E LiDAR, each packet contains 6 × 64 points, corresponding to around 1◦ rotation of the sensor. To represent the real-time + +![](_page_5_Figure_2.jpeg) + +Fig. 4: Sensor Acquisition Geometry. We represent in a the acquisition of a rotating sensor, which is split into 1 ⁄3 turn slices in b. As the Laser emitters position forms an angle of over 17.3 ◦ around the sensor head, taking slices with respect to the sensor rotation θ results in a jagged profile. + +operational setting of autonomous driving, we associate with each point the timestamp of its packet output event, i.e. the instant the packet is available and not the acquisition time of the point. The latency between the acquisition of the first point and the complete transfer of its packet is 278µs. Although small compared to acquisition and inference times, this more rigorous timing constitutes a step towards a more realistic evaluation setting of segmentation algorithms of LiDAR sequences. + +On top of its absolute position, we associate with each individual point its cylindrical coordinates relative to the position of the sensor at the exact time of its acquisition. This differs from other datasets such as SemanticKITTI [3], which gives the relative position of all points but the absolute position of the sensor only once per frame. While sensor movement can be interpolated, the vehicle trajectory might not be linear and the sensor head rotates. For comparison, at 50km/h, the sensor moves more than 1.4m during each rotation. + +LiDAR sequences are typically split into frames containing points that cover a 360◦ degree arc around the sensor. However, the acquisition geometry makes this grouping artificial. Indeed, the fibers (i.e. the individual lasers) do not all face the same direction: they are arranged around the sensor's heads at different angles, with a range of more than 17.3 ◦ . This means that the points within a packet are not vertically aligned but present a jagged profile as seen in Figure 4. In order to obtain frames with straight edges such as those of SemanticKITTI [3], we would have to consider an acquisition over a sensor rotation of 377◦ , adding a further 5ms of latency. Contrary to other datasets, HelixNet contains the index of the emitter of each point and organizes the points with respect to the angle of the sensor itself This allows us to easily build frames or frame portions that are directly consistent with the rotation of the sensor head itself. This is important for measuring the real latency of segmentation methods and, as described in the next section, contributes to the efficiency of our proposed network. + +We consider a sequence of 3D points acquired by a rotating LiDAR on a mobile platform, which we split into chronologically ordered slices of acquisition. As represented in Figure 5, we process each slice with a U-Net architecture [37] with cylindrical convolutions [51]. At the lowest resolution, a spatio-temporal transformer network connects neighboring voxels in space and time, resulting in a large receptive field. We first describe the construction of slices, then our cylindrical U-Net, and finally the transformer module. + +Instead of processing the data frame-by-frame, we propose to split the sequence into slices covering a fixed portion of the sensor rotation, resulting in a shorter acquisition time and a lower latency. Each point i of the sequence is characterized by the angular position $\theta_i$ of the sensor head at its exact time of acquisition. The points are sorted in chronological acquisition order i.e. $\theta_i \leq \theta_j$ if i < j. We partition the sequence into groups of contiguous points called slices, acquired during a portion $\Delta\theta \in ]0, 2\pi]$ of a full rotation of the sensor itself. Choosing $\Delta\theta = 2\pi$ corresponds to the classic frame-by-frame setting and implies an acquisition latency of 104ms in HelixNet or SemanticKITTI [3]. A slice size of $\Delta\theta = 2\pi/5$ leads to an acquisition latency of 21ms, which is more conducive to real-time processing of driving data. + +Inspired by the Cylinder3D model [51], we first discretize each slice along a fine cylindrical partition grid (1). Each point i is associated with a descriptor $x_i^{\text{point}}$ based on its intensity, relative position with respect to the sensor in Cartesian and cylindrical coordinates, and its offset with respect to the center of its voxels in grid (1). We compute the point feature $f_i^{\text{point}}$ by applying a shared Multi-Layer Perceptron (MLP) $\mathcal{E}^{\text{point}}$ to $x_i^{\text{point}}$ for all points i in the slice. The resulting $f_i^{\text{point}}$ are then maxpooled with respect to the voxels of grid (1) to serve as input to a convolutional encoder $\mathcal{E}^{\text{grid}}$ . The network $\mathcal{E}^{\text{grid}}$ is composed of sparse cylindrical convolutions [18] and strided convolutions for downsampling. $\mathcal{E}^{\text{grid}}$ produces a set of L sparse feature maps $f^{\text{grid}(1)}, \dots, f^{\text{grid}(L)}$ with decreasing resolutions: + +$$f_i^{\text{point}} = \mathcal{E}^{\text{point}} \left( x_i^{\text{point}} \right)$$ + (1) + +$$f^{\mathsf{grid}(1)}, \cdots, f^{\mathsf{grid}(L)} = \mathcal{E}^{\mathsf{grid}} \left( \mathsf{maxpool} \left( f^{\mathsf{point}} \right) \right) ,$$ + (2) + +where maxpool is performed with respect to grid (1). At the lowest resolution grid (L), we apply the transformer-based module $\mathcal{T}$ presented in the next subsection to the feature map $f^{\text{grid}(L)}$ to obtain the coarse cylindrical map $g^{\text{grid}(L)}$ : + +$$g^{\operatorname{grid}(L)} = \mathcal{T}\left(f^{\operatorname{grid}(L)}\right)$$ + (3) + +![](_page_7_Figure_2.jpeg) + +Fig. 5: Helix4D Architecture. A point sequence is split into angular slices, whose points are encoded by E point and pooled along a fine-grained cylindrical partition. A convolutional encoder E grid yields feature maps at lower resolutions. We apply W consecutive spatio-temporal transformer blocks T w on the coarse voxels, with attention spanning across current and past slices. The resulting features are up-sampled to full resolution with a convolutional decoder D grid using the encoder's maps at intermediate resolutions through skip connections. Finally, the grid features are allocated to the points, which are classified by D point . + +The decoder Dgrid combines cylindrical convolutions and strided transposed convolutions to map g grid(L) to a feature map g grid(1) at the highest resolution, and uses the maps f grid(L−1) , · · · , f grid(1) through residual skip connections. We concatenate for each point i the descriptor g grid(1)(i) of its voxel in grid (1) and its point feature f point i . Finally, the point decoder Dpoint associates a vector of class scores c point i with each point i: + +$$g^{\mathsf{grid}(1)} = \mathcal{D}^{\mathsf{grid}}\left(g^{\mathsf{grid}(L)}, f^{\mathsf{grid}(L-1)}, \cdots, f^{\mathsf{grid}(1)}\right)$$ + (4) + +$$c_i^{\mathsf{point}} = \mathcal{D}^{\mathsf{point}}\left(\left[g^{\mathsf{grid}(1)}(i), f_i^{\mathsf{point}}\right]\right) ,$$ + (5) + +where [ · ] is the channelwise concatenation operator. The network is supervised by the cross-entropy and Lov´asz-softmax [4] losses directly on the point prediction, without class weights. + +Our approach differs from Cylinder3D [51] by relying on simple 3×3×3 sparse cylindrical convolutions instead of asymmetrical convolutions and dimensionbased context modeling. Furthermore, we do not use voxel-wise supervision. + +Our simplified architecture results in a lighter computational and memory load, but can still learn rich spatio-temporal features thanks to the addition of the transformer module described below. + +We denote by V the set of non-empty voxels at the lowest resolution grid (L) for all slices of the considered sequence. We associate with each voxel v of V a + +feature $f_v^{\text{voxel}}$ defined as the value of $f^{\text{grid}(L)}$ at v. We remark that $f^{\text{voxel}}$ can be ordered as a non-strictly ordered time sequence, and propose to successively apply W independent transformer blocks $\mathcal{T}^1, \dots, \mathcal{T}^W$ whose architecture is described below. We denote by $g^{\text{voxel}}$ the resulting spatio-temporal voxel representation: + +$$g^{\text{voxel}} = \mathcal{T}^W \circ \dots \circ \mathcal{T}^1(f^{\text{voxel}}) \ .$$ + (6) + +We associate each voxel v of $\mathcal{V}$ with the absolute position $(X_v, Y_v, Z_v)$ of its center, the release time $T_v$ of its first point, and the index $I_v$ of the sensor rotation of its corresponding slice. In order to use a sparse attention scheme, we define for each voxel v a spatio-temporal mask M(v) characterized by a radius R and a set of rotation offsets $P \subset \mathbb{N}$ : + +$$M(v) = \{ u \mid ||(X_v, Y_v, Z_v) - (X_u, Y_u, Z_u)|| < R, I_v - I_u \in P \} .$$ + (7) + +In the context of autonomous driving, we choose R = 6m and $P = \{0, 5, 10\}$ . With a standard rotation speed of 10Hz, this corresponds to considering slices 0.5 and 1 seconds in the past along with the current one. See Figure 6 for an illustration of the receptive field and attention maps. + +Simplified Transformer Block. We now define a single transformer block $\mathcal{T}^w$ with H heads operating on a sequence of voxel features $f^{\text{voxel}}$ of dimension D. For each head h and each voxel v, we apply the following operations: + +- (i) A single linear layer $\mathcal{L}^h$ generates both a key $k_v^h$ of dimension K and a value $\operatorname{val}_{v}^{h}$ of dimension D/H. +- (ii) For all voxels u in the mask M(v), we define the compatibility score $y_{u,v}^h$ as the cross-product between keys and with a learned relative positional encoding $PE^h(u, v)$ . +- (iii) The cross-voxel attention $a_{u,v}^h$ is obtained with a scaled softmax. +- (iv) The values valuh of voxels in M(v) are averaged into a vector $\tilde{f}_v^h$ using their respective cross-voxel attention as weights. +- (v) The vectors $\hat{f}_{v}^{h}$ are concatenated channelwise across heads and added to the input of the block to define its output. + +These operations can be summarized as follows: + +$$k_v^h, \operatorname{val}_v^h = \mathcal{L}^h \left( f_v^{\mathsf{voxel}} \right)$$ + (8) + +$$k_v^h, \operatorname{val}_v^h = \mathcal{L}^h \left( f_v^{\text{voxel}} \right)$$ + (8) + $y_{u,v}^h = \left( k_v^h \right)^{\mathsf{T}} \left( k_u^h + \operatorname{PE}^h(u, v) \right)$ for $u \in M(v)$ (9) + +$$\left\{a_{u,v}^{h}\right\}_{u\in M(v)} = \operatorname{softmax}\left(\left\{y_{u,v}^{h}\right\}_{u\in M(v)}/\sqrt{K}\right)$$ +(10) + +$$\tilde{f}_{v}^{h} = \sum_{u \in M(v)} a_{u,v}^{h} \operatorname{val}_{u}^{h}$$ + +$$\mathcal{T}^{w}(f^{\text{voxel}})_{v} = f_{v}^{\text{voxel}} + [\tilde{f}_{v}^{1}, \cdots, \tilde{f}_{v}^{H}].$$ +(11) + +$$\mathcal{T}^w(f^{\text{voxel}})_v = f_v^{\text{voxel}} + [\tilde{f}_v^1, \cdots, \tilde{f}_v^H] . \tag{12}$$ + +Our design is similar to the classical transformer architecture but uses keys as queries to save memory and computation. We also do not use feed-forward + +![](_page_9_Figure_2.jpeg) + +Fig. 6: **Spatio-Temporal Attention.** We represent the spatio-temporal mask and attention score of one head of the transformer for two different voxels. The network gathers information from different frame offsets P as the sensor moves. + +networks after averaging the values: the only learnable part of a block $\mathcal{T}^w$ is its linear layers $\mathcal{L}^h$ and its relative positional encoding PEh. + +Since $g^{\text{voxel}}$ only requires information about the voxels of the current and past slices, it can be computed sequentially for all slices in the order in which the sensor releases them. For a given slice, the voxel map $g^{\text{grid}(L)}$ for non-empty voxels is given by the values of $g^{\text{voxel}}$ , and set to zero otherwise. To save computation at inference time, we store in memory the keys, values, and absolute positions of the voxels in past slices with a fixed buffer of $\max(P)$ rotations. This allows us to allocate a large spatio-temporal receptive field to each voxel without supplementary computations. + +Relative Positional Encoding. We propose to learn relative positional vectors $\operatorname{PE}^h(u,v)$ that encode the spatio-temporal offset $(X_u,Y_u,Z_u,T_u)-(X_v,Y_v,Z_v,T_v)$ between voxels u and v for each transformer block w independently. Inspired by the work of Wu et al. [47], we first discretize the offsets along each dimension $d \in \{X,Y,Z,T\}$ with $B_d$ irregular bins. For each dimension d and head d, we learn d weight vectors of size d. We define the functions $\operatorname{PE}^h_d: \mathbb{R} \mapsto \mathbb{R}^K$ that map the d-dimension of an offset to the vector associated with its corresponding bin. The positional encoding between two voxels d and d is the sum of the vectors corresponding to their discretized offsets in each dimension: + +$$PE^{h}(u,v) = PE_{X}^{h}(X_{u} - X_{v}) + PE_{Y}^{h}(Y_{u} - Y_{v}) + PE_{Z}^{h}(Z_{u} - Z_{v}) + PE_{T}^{h}(T_{u} - T_{v}).$$ +(13) + +Relative positional encoding vectors are used directly in the calculation of the compatibility score, as given in (9). Additional details on positional encoding are given in the supplementary material. + +We evaluate the performance and inference time of our approach and other state-of-the-art methods in both online and frame-by-frame settings. We use our proposed dataset HelixNet and the standard SemanticKITTI dataset. + +Online Evaluation Setting We aim at evaluating the real-time readiness of rotating LiDAR semantic segmentation algorithms in the context of autonomous driving. The total latency of a model is determined by its inference speed and also the time it takes to acquire its input. Operating on full frames requires at least 104ms of acquisition, which is incompatible with realistic autonomous driving scenarios. Instead, we propose an online evaluation setting using the slices defined in Sec. 4.1. By default, we use a slice size of a fifth turn of the sensor head: ∆θ = 2π/5, corresponding to 21ms of acquisition. + +Slices are processed sequentially. We define the inference latency of a segmentation method as the average time between the release of the last point of a slice and its segmentation. To meet the real-time requirement, inference must be faster than the acquisition of a slice. Slower processing would cause the classification to continuously fall behind. Although thinner slices directly reduce acquisition latency, they also make the real-time requirement more strict: as a full turn must be processed in less than 104ms, a fifth turn must be in at most 21ms. + +Adapting SemanticKITTI. SemanticKITTI [3,16] contains 43 552 frames along 22 sequences of LiDAR scans densely annotated with 19 classes. In contrast to HelixNet, SemanticKITTI is not formatted with respect to the sensor rotation and only gives the acquisition time and sensor position once per frame. To measure the latency, we make the following approximation: (i) the fibers are assumed to be vertically aligned, meaning that the angle of the points is the same as the sensor's; (ii) we interpolate the acquisition time of points between frames from their angular positions; (iii) we use the acquisition time as release time. To obtain the absolute positions of the voxels, we assume that the sensor jumps between the positions given by the camera poses for each frame. In our open-source implementation, we provide an adapted dataloader allowing methods already running on SemanticKITTI to be evaluated in the online setting with minimal adaptation. + +Adapting Competing Methods. To evaluate the semantic segmentation performance and latency of other segmentation algorithms in the online setting, we process the point clouds corresponding to each slice independently and sequentially. This approach restricts the spatial receptive field to the extent of the slices. However, as the sensor moves, it is not straightforward to add past slices whose relative positions may no longer be valid. By explicitly modeling the spatiotemporal offset between voxels, Helix4D does not suffer from this limitation. + +We selected five segmentation algorithms with open-source implementations and trained models for SemanticKITTI. SalsaNeXt [12] uses range images, Polar-Net [48] and panoptic PolarNet [50] a bird's eye view polar grid, SPVNAS [43] a regular grid, and Cylinder3D [51] a cylindrical grid. We do not consider methods that stack frames as their structure and resulting latency is incompatible with the online setting. When using SemanticKITTI, we evaluate the provided pretrained models on the validation set. On HelixNet, we retrain the models from scratch + +Table 2: **Semantic Segmentation Results.** Performance of Helix4D and competing approaches on HelixNet and on the validation set of SemanticKITTI\*, in the frame-by-frame and online setting. We report the mean Intersection-over-Union (mIoU) and the inference time in ms. Methods meeting the real-time requirement are indicated with $\checkmark$ and those who do not with $\checkmark$ . $\star$ SemanticKITTI is denoted as SK. Measuring the latency on this dataset requires making non-realistic approximations about the fiber position. + +| Method | Size | Full frame $lacktriangle$ | | $104 \mathrm{ms}$ | ¹⁄₅ frame ▶ | | $21 \mathrm{ms}$ | | +|--------------------|---------------|---------------------------|------|-------------------|-------------|------|------------------|--| +| | $\times 10^6$ | HelixNet | SK* | Inf. (ms) | HelixNet | SK* | Inf. (ms) | | +| SalsaNeXt [12] | 6.7 | 69.4 | 55.8 | 23 🗸 | 68.2 | 55.6 | 10 🗸 | | +| PolarNet [48] | 13.6 | 73.6 | 58.2 | 49 🗸 | 72.2 | 56.9 | 36 🗶 | | +| Pan. PolarNet [50] | 13.7 | | 64.5 | 50 🗸 | _ | 60.3 | 44 🗶 | | +| SPVNAS [43] | 10.8 | 73.4 | 64.7 | 73 🗸 | 69.9 | 57.8 | 44 🗶 | | +| Cylinder3D [51] | 55.9 | 76.6 | 66.9 | 108 🗡 | 75.0 | 65.3 | 54 🗶 | | +| Helix4D (Ours) | 1.0 | 79.4 | 66.7 | $45 \checkmark$ | 78.7 | 66.8 | 19 🗸 | | + +Table 3: **HelixNet Semantic Segmentation Scores.** We report the IoU for each class of HelixNet evaluated in the online setting with slices of 72°. + +| | λ | .s. | aurface | 78 × 8, | tion ac | signs | vehicle | g vehicle | rian
Artifac | χ. | +|-----------------|------|--------|---------|---------|---------|--------|---------|-------------|-----------------|------| +| Method | Road | Office | Build | Jeger | Traffic | Static | Month | Pedes | Artife | Avg | +| SalsaNeXt [12] | 84.4 | 76.1 | 88.7 | 70.7 | 61.4 | 58.6 | 35.1 | 68.5 | 69.7 | 68.2 | +| PolarNet [48] | 86.2 | 77.9 | 91.2 | 77.9 | 63.2 | 64.8 | 35.4 | 68.1 | 84.8 | 72.2 | +| SPVNAS [43] | 80.5 | 77.1 | 93.0 | 81.8 | 68.0 | 60.9 | 36.9 | 71.7 | 59.0 | 69.9 | +| Cylinder3D [51] | 85.3 | 78.4 | 93.5 | 83.9 | 66.2 | 63.3 | 35.7 | 77.7 | 90.9 | 75.0 | +| Helix4D (Ours) | 87.8 | 82.5 | 94.0 | 84.4 | 68.9 | 72.3 | 46.4 | 78.8 | 93.3 | 78.7 | + +using the procedure of their official repository. We removed all test-time augmentations that resulted in prohibitive inference time. All methods are evaluated on the same workstation using a NVIDIA TESLA V100 32Go GPU. + +Analysis. In Table 2, we report performance in frame-by-frame and online setting with slices of $72^{\circ}$ , for Helix4D and competing methods, for HelixNet and SemanticKITTI. We observe that Helix4D yields state-of-the-art accuracy, with mIoU scores only matched by Cylinder3D [51]. However, Cylinder3D is 50 times larger in terms of parameters and twice slower, not meeting the real-time requirement even in the full frame setting. As reported in Table 3, distinguishing moving vehicles in HelixNet is particularly difficult. Our approach even largely outperforms Panoptic PolarNet despite this method using instance annotation as supervision, preventing us from evaluating on HelixNet. Helix4D yields significantly improved scores thanks to its larger spatio-temporal receptive fields: 14m and 1000ms vs. 8m and 21ms for Cylinder3D for a fifth rotation. + +![](_page_12_Figure_2.jpeg) + +Fig. 7: Influence of Slice Size. We plot the processing time (left, in ms) and precision (right, in mIoU) of different methods with respect to the considered size of slices, estimated on the validation set of SemanticKITTI [3]. Methods whose inference time is slower than the acquisition time of the slice (red shaded area) do not meet the real time requirement. + +Table 4: Ablation Study. We report the speed and accuracy of several modification of our Helix4D on the validation set of SemanticKITTI. + +| Method | Size
×103 | Full Frame
mIoU | 104ms
Inf. (ms) | ⁄5 Frame
1
mIoU | 21ms
Inf. (ms) | +|----------------------------------|--------------|--------------------|--------------------|-----------------------|-------------------| +| Helix4D | 985 | 66.7 | 45 ✓ | 66.8 | 19 ✓ | +| (a) Asymmetric Convolutions 1171 | | 66.6 | 56 ✓ | 66.6 | 31 ✗ | +| (b) Cylindrical U-Net | 985 | 58.6 | 22 ✓ | 60.2 | 16 ✓ | +| (c) Slice-by-Slice | 985 | 62.9 | 29 ✓ | 62.6 | 19 ✓ | +| (d) w. Queries | 993 | 65.2 | 45 ✓ | 64.8 | 20 ✓ | +| (e) w/o. Positional Encoding | 983 | 64.3 | 41 ✓ | 64.1 | 18 ✓ | +| (f) Helix4D Tiny | 306 | 65.3 | 45 ✓ | 64.9 | 17 ✓ | + +In the online setting, only two approaches meet the real-time requirement: SalsaNeXt [12] and Helix4D. Our approach outperforms SalsaNeXt by over 10 mIoU points in both the full frame and the on-line settings. In short, Helix4D is as accurate as the largest and slowest models with an inference speed comparable to that of the fastest and less accurate models. The total latency (acquisition plus inference time) of our model evaluated online is 40ms (21 + 19ms), and reaches the same performance as Cylinder3D evaluated on full frame with a latency of 212ms (104 + 108ms), an acceleration of more than 5 folds. + +In Figure 7, we report the inference time and mIoU for different slice sizes. Due to various overheads, the inference time appears in an affine relationship with the size of slices, making the real-time requirement stricter for smaller slices. Due to its very design, the performance of Helix4D is not affected by the slice size. In contrast, competing methods perform worse with smaller slices. + +Ablation Study We assess on SemanticKITTI the impact of different design choices by evaluating several alterations of our method, reported in Table 4. + +- (a) Asymmetric Convolutions: we replace the 3 × 3 × 3 convolutions in our U-Net with the convolution design proposed by Cylinder3D [51]. We did not observe a significant change in performance and an increase in run-time of 50%, failing the real-time requirement for slices of 72◦ . +- (b) Cylindrical U-Net: we replace the transformer by a 1 × 1 × 1 convolution on the voxels of the lowest resolution. We observe a slight decrease in run-time and a significant drop of over 6 mIoU points. This result shows that the transformer is able to learn meaningful spatio-temporal features at low resolution. +- (c) Slice-by-Slice: we restrict the mask M(v) of each voxel to its current slice. This reduction in the temporal receptive field results in a drop of 4 mIoU points, without any appreciable acceleration. +- (d) w. Queries: we modify our simplified transformer to associate a query for each voxel along with keys and values, and use key-queries compatibilities. This does not affect the run-time and slightly decreases the performance. +- (e) w/o. Positional Encoding: we remove the relative positional encoding PE in the calculation of compatibilities in equation (9). This leads to a slightly decreased run time, but decreases performance by more than 2.5 points. This illustrates the advantage of explicitly modeling the spatio-temporal voxel offsets. +- (f) Helix4D Tiny: we replace the learned pooling in our U-Net with maxpools and use narrower feature maps for a total of 306k parameters. This method only performs two points under Helix4D with a third of its parameters. diff --git a/2209.05861/main_diagram/main_diagram.drawio b/2209.05861/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..80169dfcf1f0a48396d26b2e886c8f7dee803737 --- /dev/null +++ b/2209.05861/main_diagram/main_diagram.drawio @@ -0,0 +1,575 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/2209.05861/paper_text/intro_method.md b/2209.05861/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..279105e0a240dc958e666a041e03b04fd6dc9021 --- /dev/null +++ b/2209.05861/paper_text/intro_method.md @@ -0,0 +1,78 @@ +# Introduction + +Overview: Sponsored search is a form of online advertising that enables advertisers to display their ads along with organic results on web search engines. It is a major source of revenue for search engines and helps advertisers attract targeted user traffic. To participate in sponsored search, advertisers can bid on keywords that are + +Bhargav Dodla\* dodla.bhargav@gmail.com Indian Institute of Technology Madras Chennai, India + +> Amit Singh siamit@microsoft.com Microsoft Bengaluru, India + +relevant to their products or services using different match types. For instance, an exact match bid keyword will only match search queries that have the same search intent as the keyword. Conversely, phrase match will match queries that contain or include the meaning of the keyword. Matching search queries with relevant keywords, also referred to as query rewriting, is a complex and challenging task for various reasons. First, search queries and bid keywords are often short and ambiguous, making it difficult to precisely infer the user's or advertiser's intent. For example, the query *"currys 912 black"* is vague, but the user is actually looking for the HP 912 ink cartridge in black color from the retailer Currys. Second, search queries and keywords can span a wide range of topics, domains, languages, and countries, adding to the complexity of the task. Third, a search query must be matched with all possible high-quality exact/phrase keywords, not just one, from a large collection of bid keywords. Lastly, the matching must be done in real-time and be computationally efficient to handle all search traffic. + +Existing methods for query rewriting matching can be broadly categorized into two main groups: information retrieval (IR) and generative or NLG based retrieval. IR methods use various techniques to learn sparse or dense representations for queries and keywords based on bag-of-words, static pretrained features, or deep learning models [\[1,](#page-6-0) [3,](#page-6-1) [4,](#page-6-2) [6,](#page-6-3) [16,](#page-6-4) [20–](#page-6-5)[22,](#page-6-6) [25\]](#page-6-7). Among them, recent Dense Retrieval (DR) methods that leverage effective negative mining strategies such as ANCE [\[29\]](#page-6-8), RocketQA [\[24\]](#page-6-9), NGAME [\[7\]](#page-6-10) have been shown to achieve state-of-the-art performance on various retrieval benchmarks [\[2\]](#page-6-11). On the other hand, NLG-based methods for query rewriting such as CLOVER [\[18\]](#page-6-12), ProphetNet-Ads [\[23\]](#page-6-13) use generative models to directly transform queries into keywords. This involves training language models to generate query rewrites and then constraining their generation space during inference to the set of bid keywords. + +Given these two distinct approaches to the same problem, we perform a comprehensive study comparing their performance. Our results show NLG and DR methods retrieve a similar number of high-quality keywords per query, but many of them are different and non-overlapping. Specifically, about 40% of the high-quality keywords retrieved by DR and NLG are unique to each method and weren't retrieved by the other. We analyze the reasons for these and find they stem from the structural differences between the approaches. For instance, NLG methods treat query rewriting as a + +\*Work done during internship at Microsoft + + + +| Query | Generative Retrieval | Dense Retrieval | CLOVER-Unity | +|----------------------------|-----------------------------------|-----------------------------------|------------------------------------------------| +| | black 912 | currys kettles | hp 912 ink cartridges black | +| currys 912 black | currys currys | currys electrical kettles | currys ink cartridges | +| | black curry | kettles at currys | black hp printer cartridges | +| | student properties nottingham | rentals in nottinghamshire uk | nottingham student studios | +| student houses for rent | student accom nottingham | rentals nottingham united kingdom | student flats in nottingham | +| nottingham | rent house rental | 3 bedroom house nottingham | student lettings nottingham united kingdom | +| | 250 honda | honda | honda 250 cc motorcycle | +| 06 honda 250 | honda motor | hondas 250 | 06 honda cbr 250 | +| | honda250 | car for lease | honda 2006 bike | +| | biglietti per la cappella sistina | sistina roma | cappella sistina excursions | +| | (tickets for the sistine chapel) | (sistine rome) | (sistine chapel excursions) | +| cappella sistina biglietti | capella sistina | cappella sistina museum | cappella sistina vatican city | +| (sistine chapel tickets) | (sistine chapel) | (sistine chapel museum) | (sistine chapel vatican city) | +| | la cappella | cappella sistina orari e prezzi | musei vaticani e cappella sistina prenotazione | +| | (the chapel) | (sistine chapel times and prices) | (vatican museums and sistine chapel booking) | + +Table 1: Comparison of keywords retrieved by CLOVERv2, NGAME and our proposed CLOVER-Unity approach for four queries + +token-level generation task with a token-level loss and thus are better at learning word-level relationships such as synonymy and hypernymy. As shown in example 2 of Table [1,](#page-1-0) NLG models can better identify interchangeable tokens, such as *houses* and *properties* or *accom*, which do not alter the query/keyword intent. However, we also find that NLG methods, especially the non-autoregressive ones, struggle to retrieve longer keywords that may have a few additional tokens. This is because non-autoregressive NLG models have to predict all tokens of the keyword in parallel during inference. In contrast, DR methods do not depend on keyword token length during inference and thus do not have this limitation. In the same example in Table [1,](#page-1-0) DR methods retrieve keywords that contain additional tokens such as *united kingdom*, which NLG models miss. + +With NLG and DR models retrieving substantially different highquality keywords for the same query, it leads to a natural question of whether we can combine the strengths of both approaches. To this end, we propose CLOVER-Unity, a unified retriever that blends the advantages of NLG and DR methods using a shared encoder. Unlike standard multi-task learning, where a shared model solves more than one related task, CLOVER-Unity solves the same task (query rewriting) using two different methods: NLG and DR. Through offline experiments, we show that the NLG and DR components of CLOVER-Unity consistently outperform the individually trained NLG and DR models by an average of 19.2% and 7.4%, respectively, in terms of good keyword density. We attribute these improvements to training the Unity model with both the contrastive DR loss and the token-level NLG loss. For instance, we show that the DR component of CLOVER-Unity is better at identifying token-level relationships due to the additional token-level objective function. Further, we show that a single CLOVER-Unity model with one forward pass retrieves 9.8% more good keywords than the ensemble of two separate DR and NLG models while using 45.8% lower GPU compute. We also conduct extensive online experiments on Microsoft Bing in 140+ countries and achieve significant gains in revenue, clicks, impressions, and coverage, demonstrating the practical utility of CLOVER-Unity for query rewriting in sponsored search. To the best of our knowledge, we are the first to present such a unified + +![](_page_1_Figure_5.jpeg) + +Figure 1: Training process of CLOVER-Unity using contrastive DR and log-likelihood NLG loss functions + +framework for query rewriting. We also share our practical lessons and optimization tricks for deploying such unified models in production. Finally, we demonstrate the versatility of CLOVER-Unity by applying it to problems beyond sponsored search. We evaluate it on public datasets from the extreme classification repository [\[2\]](#page-6-11) and show that CLOVER-Unity outperforms the leading extreme classification algorithms in all metrics. + +# Method + +In this section, we describe the various components of our proposed approach for query rewriting. Our goal is to match any given search query with all its relevant exact bid keywords and phrase bid keywords in real-time. We first present details of the NLG and DR baselines in [2.1](#page-2-0) and [2.2.](#page-2-1) Then, we introduce CLOVER-Unity, which combines the benefits of both methods in a single model. + +NLG models treat query rewriting as a token-level prediction task, *i.e.*, they predict relevant keyword tokens $\{k_t\}_{t=1}^m$ individually instead of retrieving them directly. However, predicting tokens one at a time in an autoregressive (AR) fashion is very slow as it requires m sequential forward passes over the model (m is the keyword length). Moreover, when performing beam search with beam size B, each forward pass requires O(B) compute. As we show in section 3.2, existing AR models like CLOVER [18], ProphetNetAds [23], and even efficient variants with deep encoder and shallow decoder [14] have an order of magnitude higher inference cost than dense retrieval approaches. To overcome these issues, we propose CLOVERv2, which uses an encoder-based non-autoregressive (NAR) model with AR trie decoding for efficient inference. Specifically, we first assume the keyword tokens $K_t$ are conditionally independent given the query, i.e., the probability distribution can be factorized as: + +$$p^{NAR}(K_1,\ldots,K_m|Q,\theta) = \prod_{t=1}^m p^{NAR}(K_t|Q,\theta)$$ + +This decomposition allows us to predict the token distributions independently in parallel, using only one forward pass through an encoder model such as BERT [8] with a language modeling head. + +**Decoding:** We construct a trie $T_K$ of valid bid keywords K to constrain the generation space to the closed set K. We developed an optimized C++ implementation of the trie data structure that is highly memory efficient and can be hosted even on commodity CPU machines. For example, a trie of 1 billion keywords (45 GB of raw text) requires only 6.5 GB of CPU RAM. The trie plays a crucial role in decoding; it functions as an autoregressive language model to guide the generation process. In particular, we proceed in a left-to-right manner, where at each time step t, we consider only the children of the partially generated prefix $k_{< t}$ in $T_K$ as possible next tokens. Essentially, decoding a NAR model with trie is equivalent to an AR language model with (unnormalized) probability distribution: + +$$\tilde{p}(K_t = k_t | k_{< t}, q, \mathcal{K}) = \begin{cases} p^{NAR}(K_t = k_t | q), & \text{if } k_t \in \text{child}_{T_{\mathcal{K}}}(k_{< t}) \\ 0, & \text{otherwise} \end{cases}$$ + +We note that NAR models with AR trie decoding (i) do not suffer from issues such as token repetition or mode mixing observed in vanilla NAR [10], and (ii) achieve comparable or better performance than AR models while being more efficient (section 3.2). + +We adopt a siamese architecture with shared parameters to learn dense representations for queries and keywords. To train the encoder effectively, we need to carefully select a small set of irrelevant keywords for each query. We cannot use all irrelevant keywords for a query because the number of training points and bid keywords is very large (millions or billions). Moreover, randomly choosing negative keywords from a uniform distribution or from the same batch can result in uninformative negatives and slow convergence [29]. To address these issues, we use NGAME [7], a negative mining technique that selects hard negative samples within the training batches. NGAME has been shown to outperform other approaches such as ANCE [29], RocketQA [24], and TAS [11]. In our application, we omit the classifier layer in NGAME because bid keywords $\mathcal K$ change + +over time. After training, we index the keyword embeddings on SSD using the DiskANN algorithm [13], which requires minimal CPU RAM online. During inference, we retrieve the keywords that are approximately the top k nearest neighbors of the query embedding. + +We now introduce CLOVER-unity, which uses a variant of the standard multi-task learning setup where we solve one task with two different methods. In particular, we use a single encoder to perform dense and generative retrieval simultaneously, as shown in Figure 1. Given a query $q = \{q_1, \ldots, q_n\}$ , we prepend a special [CLS] token and encode the query using a transformer encoder $\mathcal{E}_{\theta}$ . The resulting sequence of hidden states, $h^q = \{h^q_c, h^q_1, \ldots, h^q_n\}$ , is used to compute the DR and NLG representations of the query. The first hidden state, $h^q_c$ , serves as the DR representation, while the remaining hidden states, $h^q_t \in \mathbb{R}^d$ , are projected to the output vocabulary space to obtain the NAR logits, $\log \tilde{p}(K_t|q,\theta) = \mathbf{W}h^q_t$ where $\mathbf{W} \in \mathbb{R}^{V \times d}$ is a learnable weight matrix, V, and d are vocabulary and hidden sizes. We apply the same process to keywords, obtaining a DR representation $h^c_c$ for each keyword. + +**Training:** We train CLOVER-Unity on a supervised dataset $\mathcal{D} = \{(q^{(i)}, k^{(i)})\}_{i=1}^L$ of high-quality query keyword pairs. For each query $q^{(i)}$ , we obtain hard negative keywords $l^{(i)}$ using the NGAME negative mining technique. Our objective is to minimize a weighted combination of the triplet margin loss for DR and negative log-likelihood for NLG, as defined by the following equation: + +$$\mathcal{L}(\theta, \mathcal{D}) = \frac{1}{L} \sum_{q,k \in \mathcal{D}} \left( \left[ (h_c^l)^T h_c^q - (h_c^k)^T h_c^q + \gamma \right]_+ \right.$$ +$$\left. - \alpha \sum_{t=1}^m \log p(K_t = k_t | q, \theta) \right)$$ + +where $h_c^q$ , $h_c^k$ , and $h_c^l$ are the DR representations of the query, relevant, and irrelevant keywords, respectively. + +**Inference:** During inference, we obtain the query representation $h_c^q$ and the NAR probabilities $p(K_t|q,\theta)$ through a single pass over the encoder $\mathcal{E}_\theta$ . We then simultaneously retrieve bid keywords using (i) DiskANN search on the query vector $h_c^q$ , and (ii) trie-based beam search on the predicted probabilities $p(K_t|q,\theta)$ . However, we observed that using standard beam search for decoding NLG models tends to produce very similar keywords with common prefixes and low diversity. To address this issue, we propose permutation decoding, which leverages a property of NAR models: they do not have an explicit order of generation. Therefore, we decode NAR models in multiple orders (e.g., left-to-right and right-to-left) and rank the top B results based on cumulative log probability, which remains unchanged regardless of the decoding order. diff --git a/2209.08244/main_diagram/main_diagram.drawio b/2209.08244/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..8eade43c11561ea07b42b32220131090f079c1bc --- /dev/null +++ b/2209.08244/main_diagram/main_diagram.drawio @@ -0,0 +1 @@ +7V1tc9o4EP41zPRuhozfZMNHIEnvZq6ddDpzbT91DAjQxVjENkm4X396tS2/gJNgx1zVacFayWujXT2PtF65A3u2ff4Y+bvNJ7yEwcAyls8D+3pgWZbhuuSLSg5cYhoe4JJ1hJZClgm+on+hbCike7SEsdIwwThI0E4VLnAYwkWiyPwowk9qsxUO1Kvu/LW4opEJvi78AJaafUPLZCOkppFr/gdE64249AiIiq2fNuaCeOMv8VPuWvbNwJ5FGCf8aPs8gwHtPdkvaINuH7azzzerOH7eWd8m/m045NpvX3JK+hMiGCbnVW1x1Y9+sBf9tcMBWhzoeQmM/AThkBzjFbXGml6e/HQw+2CSj99oI1ob4X24FBX3vMINyI1O451P6+PkIKzhPuxpb00XOMDRwJ7Qk9dz/4NBz7Zm5LPyiF6JnMXMHybDlb9FwYGfvsUhJpdZQLVJzPyQNjB2z9l1ydGafZM7Jf+Kfg9Iz5EPbtO0JHsSsL4kkmt6TG8O0M4DpMdPtTXTttKKr1JjZWq4xdIa8sE6n379NhQGAcwMpM10HrHSmpekiYFiZMDNDNgdEWWIK8suyc2YXpJdBPK7M5nepw1R95UbgwifSM8y+SbZBrlmSz/eKOdRizHgYCKXy1AQzJiXUGGIQ1Y7zXUAjBL43KiL06q0dz9CvIVJxPsgp8S1M3MYh0w+zomfGIpkyikeyroNB5JMX+48P87k69z1i7ZnxdQB8kLVLUW7kv/yoUdwgI8+4fAcwdKBmFqP2kCxnLCbsBqvTg3EzJM3DtVlW8bUmkzoeE8ifA9lDTOaPcVELUoofYyM9E5OYpnAPGHlDIYFtmUGNFKqcvgpgqgcieNPOdQHos0mB/iynS+IJrNNBqaD1CIvwFb7BdgqzUYHKrdYHm0tjbYabTXaarTtDdo6cnrfG7R1Sp0Ol2QhIIo4SjZ4jUM/uMmk08U+ekytktmIdlx2wl8Y70STf2CSHIRp/H2CVbvBZ5R8zx3/oKqugChdPwvNrHCQhZD89u/5Qu4sWsxOYyV5XtFHXmbnGO+jBTy9LEj8aA2P6RMUR3v6qNdEMCC496guyc7uAuBshGtrwtWEqwlXE25vCNftHeG6mnDPR7h2Q8IFvSLc0dkI90g80ZqamnQ16WrS1aTbNemOTPs06bpeh6Q71qR7PtIFDUl31CvSNcusKwco7a9Kbizwl+kcITD63HPhBwNv+mngXf9EA3BDjjkPe9eclkP4wA/qGtOGRb4n3/6WokA4j3ecVwXr78MAxrFgfPJvh34SLUNE1fAWG4p09BkxGbxrCvQpIvHfXINIZJwnqvv6AVqH5HhBXAgSRJlSNEDkJ0xExRYtl2zYRJB0mD9nqqgz7jAKE2ZMQLrqmuoigyMW7lmHVHlYk6KcV3uDlLu4kragbGyr6wfglKGMUVJnUGaOa/14iR4r3ZiacygsSP1YGLHsyep4WPkLVc8GBo+QWj1/qhHL3AUpM5XqRUpOmZDQ1Nzy/fo7mBcFqNH49KrG56d9kKDhRMyaJwH56aGf0EnqnZx1/5lNvdIBgoq3QGSl2yoNpFTMbHF8eL2A6Z0WXdwygMrWxqjk4rZV4eJOWy4uL5Z3cYJwV1dXHNYuDa3atJ5pSGJLJ1tl8zkV1rPasp53kXOtor1iMr1JJjSRKrMpk90i2h1i1rWULRaBH8dowYWiialM4a4MK5238RsanZrHsdIdjBAxDHXj65dP2k9O5kYNJ3PSXc43mxOn3tERl/dnqxCxGxUcld+qOK3gq+l9vAF85BBUwScRM6ovw3SdHg90hpeOxuhojI7G1EVjbm9HM3f8ztGYYoaXSwvv+ghE3sArIVandWmI1RCrIbY/EFtM6+oBxNoXuQrracQ73YFzcpVknXuV9DYvcN5EtDqdSxOtJlpNtP0h2mI6Vw+IFmiiPSPRWk2J1ukX0XovJlqd2KXpV9Ovpt9Lot9iYlcl/Xaa2GWNNP2ekX6dpvTr9Yp+7frcrsvJiVmtRgsywt85J+bL8BdPhik8LBmbJYTrNhlGKtbJMKetV0qGcW2vZL5Ok2Gsy9zu01I2zLnpymtIV9Lk7SevFP0PWKqOlpNXnPJe3lr6qF04FhkEVDHIZ7LAKGcm18yLz3FB+LBHAZpHaL89cR2Vmy4NHztLbTYNz1Oc1XErtmlUoaXdGte9P1r2Avnax6mRpZreA1egEVJV6AJOQZerauI43h7mVcTeNOZpzKvBPDXZ3bUrIhjdYt4Z9iW5tfFXFjQeeLMvfIcRUTe75zuOLGNIl3tc/nsmZ81/ki8UrpIDVxBA/g13MQrY3uR8BFrvJmr8+MoqeN+o4VaL1rxPzk872hVnEne7Z9YjXja9E2VrFtFvWpqle9ea+dY5Q6yQ/Sn5EKmxXXtsLwf5OOu41ZRNS42zAjmvz6OU6UrK7iQOIVm+ByvZbPX6I1dzKtJ6pW6/GByPtCp4kHccUJwUZlO+/LzQVOeFAnWySWELWzmaPjuVY74nwVvQnyfoqmMZr3Is83/oWE3T3/rmWW4tuVXNodsgPIsR3DFC6/5RYn94zgFqBKKK59wug+2gfnN1Vx5ja485NoPum8e49RPorjzGLLxqQvtOTfaCm86Ye+M99bGrty/+/6ZvAyEHcxws48OWfPHyjr8g5IaX8I4+7sVR6G8hkYTwidbyv2ydv6aL/1fpov/nRKbr5HtNdEjhVRTqKi7tGHbZpat8ur0XHL5/Ss6vEsQ3FdO7jveGIH4hs0uGJTsK4oP+vDVCRxfe9qpNt19rwPdHIx1dOE/Of888S5Jq7zyrY8hyRDk3s5+OAH3nY189rukLDGU0oC8OZ/bU4TqGsstzuMZ51X3zuPfPx9EQ1+prvWT4qi8Op58FXajDNc7F7czjBmIzV26dmm3jsm/+Aw== \ No newline at end of file diff --git a/2209.08244/main_diagram/main_diagram.pdf b/2209.08244/main_diagram/main_diagram.pdf new file mode 100644 index 0000000000000000000000000000000000000000..ad391827063aa28ecdbc9b3f30f344aea13a5a18 Binary files /dev/null and b/2209.08244/main_diagram/main_diagram.pdf differ diff --git a/2209.08244/paper_text/intro_method.md b/2209.08244/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..3014bc81ce4fa0a1e0e2fb3ae0e563079bdbd258 --- /dev/null +++ b/2209.08244/paper_text/intro_method.md @@ -0,0 +1,162 @@ +# Introduction + +Cooperative multi-agent reinforcement learning (MARL) is a well-abstracted model for a broad range of real applications, including logistics [@li2019cooperative], traffic signal control [@xu2021hierarchically], power dispatch [@wang2021multiagent], and inventory management [@feng2022multiagent]. In cooperative MARL, centralized training with decentralized execution (CTDE) is a popular learning paradigm, where the information of all agents can be gathered and used in training. Many CTDE methods [@lowe2017multi; @foerster2018COMA; @sunehag2018vdn; @rashid2018qmix; @wang2020qplex; @zhang2021fop; @yu2021surprising] have been proposed and shown great potential to solve cooperative multi-agent tasks. + +Another paradigm is decentralized learning, where each agent learns its policy based on only local information. Decentralized learning is less investigated but desirable in many scenarios where the information of other agents is not available, and for better robustness, scalability, and security [@Zhang2019multi]. However, *fully decentralized learning* of agent policies (*i.e.*, without communication) is still an open challenge in cooperative MARL. + +The most straightforward way for decentralized learning is directly applying independent learning at each agent [@tan1993multi], which however induces the well-known non-stationarity problem for all agents [@Zhang2019multi] and may lead to the learning instability and a non-convergent joint policy, though the performance varies as shown in empirical studies [@rashid2018qmix; @de2020independent; @papoudakis2021benchmarking; @yu2021surprising]. + +In the paper, we directly tackle the non-stationarity problem in the simplest and fundamental way, *i.e.*, fixing the policies of other agents while one agent is learning. Following this principle, we propose *multi-agent alternate Q-learning* (**MA2QL**), a *minimalist* approach to fully decentralized cooperative multi-agent reinforcement learning, where agents take turns to update their policies by Q-learning. MA2QL is theoretically grounded and we prove that when each agent guarantees an $\varepsilon$-convergence at each turn, their joint policy converges to a Nash equilibrium. In practice, MA2QL only requires the minimal changes to independent Q-learning (IQL) [@tan1993multi; @tampuu2015multiagent] and also independent DDPG [@lillicrap2015continuous] for continuous action, *i.e.*, simply swapping the order of two lines of codes as follows. + +
+
+
+

all agents interact in the environment agent i updates by Q-learning terminate

+
+
+
+
+

all agents interact in the environment agent i updates by Q-learning terminate

+
+
+
MA2QL
+
+ +We evaluate MA2QL on a didactic game to empirically verify its convergence, and multi-agent particle environments [@lowe2017multi], multi-agent MuJoCo [@peng2021facmac], and StarCraft multi-agent challenge [@samvelyan2019starcraft] to verify its performance in discrete and continuous action spaces, fully and partially observable environments. We find that MA2QL consistently outperforms IQL, despite such minimal changes. The effectiveness of MA2QL suggests that simpler approaches may have been left underexplored for fully decentralized cooperative multi-agent reinforcement learning. + +# Method + +**Dec-POMDP.** Decentralized partially observable Markov decision process (Dec-POMDP) is a general model for cooperative MARL. A Dec-POMDP is a tuple $M=\left\{S,A,P,Y,O,I,n,r,\gamma\right\}$. $S$ is the state space, $n$ is the number of agents, $\gamma \in [0,1)$ is the discount factor, and $I = \{1,2\cdots n\}$ is the set of all agents. $A = A_1 \times A_2 \times \cdots \times A_n$ represents the joint action space where $A_i$ is the individual action space for agent $i$. $P(s^{\prime} |s,\bm{a} ): S \times A \times S \to [0,1]$ is the transition function, and $r(s,\bm{a} ): S \times A \to \mathbb{R}$ is the reward function of state $s$ and joint action $\bm{a}$. $Y$ is the observation space, and $O(s,i):S \times I \to Y$ is a mapping from state to observation for each agent. The objective of Dec-POMDP is to maximize $J({\bm{\pi}}) = \mathbb{E}_{\bm{\pi}}\left[ \sum_{t = 0}^\infty \gamma^t r(s_t,\bm{a}_t ) \right],$ and thus we need to find the optimal joint policy ${\bm{\pi}}^{*} = \arg\max_{{\bm{\pi}}} J({\bm{\pi}})$. To settle the partial observable problem, history $\tau_i \in \mathcal{T}_i: (Y \times A_i)^*$ is often used to replace observation $o_i \in Y$. Each agent $i$ has an individual policy $\pi_i(a_i|\tau_i)$ and the joint policy $\bm{\pi}$ is the product of each $\pi_i$. Though individual policy is learned as $\pi_i(a_i|\tau_i)$ in practice, we will use $\pi_i(a_i|s)$ in analysis and proofs for simplicity. + +**Dec-MARL.** Although decentralized cooperative multi-agent reinforcement learning (Dec-MARL) has been previously investigated [@zhang2018fully; @de2020independent], the setting varies across these studies. In this paper, we consider Dec-MARL as a *fully* decentralized solution to Dec-POMDP, where each agent learns its policy/Q-function from its own action individually ***without communication or parameter-sharing***. Therefore, in Dec-MARL, each agent $i$ actually learns in the environment with transition function $P_i(s'|s,a_i) = \mathbb{E}_{a_{-i} \sim \pi_{-i}}[P(s'|s,a_i,a_{-i})]$ and reward function $r_i(s,a_i) = \mathbb{E}_{a_{-i} \sim \pi_{-i}} [r(s,a_i,a_{-i})]$, where $\pi_{-i}$ and $a_{-i}$ respectively denote the joint policy and joint action of all agents expect $i$. As other agents are also learning (*i.e.*, $\pi_{-i}$ is changing), from the perspective of each individual agent, the environment is non-stationary. This is the non-stationarity problem, the main challenge in Dec-MARL. + +**IQL.** Independent Q-learning (IQL) is a straightforward method for Dec-MARL, where each agent $i$ learns a Q-function $Q(s,a_i)$ by Q-learning. However, as all agents learn simultaneously, there is no theoretical guarantee on the convergence due to non-stationarity, to the best of our knowledge. In practice, IQL is often taken as a simple baseline in favor of more elaborate MARL approaches, such as value-based CTDE methods [@rashid2018qmix; @son2019qtran]. However, much less attention has been paid to IQL itself for Dec-MARL. + +To address the non-stationarity problem in Dec-MARL, a fundamental way is simply to make the environment stationary during the learning of each agent. Following this principle, we let agents learn by turns; in each turn, one agent performs policy iteration while fixing the policies of other agents. This procedure is referred to as *multi-agent alternate policy iteration*. As illustrated in Figure [1](#fig:api){reference-type="ref" reference="fig:api"}, multi-agent alternate policy iteration differs from policy iteration in single-agent RL. In single-agent RL, policy iteration is performed on the same MDP. However, here, for each agent, policy iteration at a different round is performed on a different MDP. As $\pi_{-i}$ is fixed at each turn, $P_i(s'|s,a_i)$ and $r_i(s,a_i)$ are stationary and we can easily have the following lemma. + +::: {#lemma-policy-iteration .lemma} +**Lemma 1** (multi-agent alternate policy iteration). *If all agents take turns to perform policy iteration, their joint policy sequence $\{\bm{\pi}\}$ monotonically improves and converges to a Nash equilibrium.* +::: + +::: proof +*Proof.* In each turn, as the policies of other agents are fixed, the agent $i$ has the following update rule for policy evaluation, $$\begin{equation} + Q_{\pi_i}(s,a_i) \leftarrow r_i(s,a_i) + \gamma \mathbb{E}_{s'\sim P_i,a_i' \sim \pi_i}[Q_{\pi_i}(s',a_i')]. +\end{equation}$$ We can have the convergence of policy evaluation in each turn by the standard results [@sutton2018reinforcement]. Moreover, as $\pi_{-i}$ is fixed, it is straightforward to have $$\begin{equation} + Q_{\pi_i}(s,a_i) = \mathbb{E}_{a_{-i} \sim \pi_{-i}}[Q_{\bm{\pi}}(s,a,a_i)]. +\end{equation}$$ + +Then, the agent $i$ performs policy improvement by $$\begin{equation} + \label{cond1} + \pi^{\operatorname{new}}_i(s) = \arg\max_{a_i} \mathbb{E}_{\pi^{\operatorname{old}}_{-i} }\left[ Q_{\bm{\pi}^{\operatorname{old}}}(s,a_i,a_{-i} ) \right]. +\end{equation}$$ As the policies of other agents are fixed (i.e., $\pi^{\operatorname{new}}_{-i}=\pi^{\operatorname{old}}_{-i}$), we have $$\begin{equation} +\begin{aligned} + V_{\bm{\pi}^{\operatorname{old}}}(s) & = \mathbb{E}_{\bm{\pi}^{\operatorname{old}}}[Q_{\bm{\pi}^{\operatorname{old}}}(s,a_i,a_{-i})] = \mathbb{E}_{\pi_i^{\operatorname{old}}}\mathbb{E}_{\pi_{-i}^{\operatorname{old}}}[Q_{\bm{\pi}^{\operatorname{old}}}(s,a_i,a_{-i})] \\ + & \le \mathbb{E}_{\pi_i^{\operatorname{new}}}\mathbb{E}_{\pi_{-i}^{\operatorname{old}}}[Q_{\bm{\pi}^{\operatorname{old}}}(s,a_i,a_{-i}) ] = \mathbb{E}_{\pi_i^{\operatorname{new}}}\mathbb{E}_{\pi_{-i}^{\operatorname{new}}}[Q_{\bm{\pi}^{\operatorname{old}}}(s,a_i,a_{-i}) ] \\ + & = \mathbb{E}_{\bm{\pi}^{\operatorname{new}}}[Q_{\bm{\pi}^{\operatorname{old}}}(s,a_i,a_{-i}) ] = \mathbb{E}_{\bm{\pi}^{\operatorname{new}}}[r(s,a_i,a_{-i}) + \gamma V_{\bm{\pi}^{\operatorname{old}}}(s^\prime) ]\\ + & \le \cdots \le V_{\bm{\pi}^{\operatorname{new}}}(s), +\end{aligned} +\end{equation}$$ where the first inequality is from [\[cond1\]](#cond1){reference-type="eqref" reference="cond1"}. This proves that the policy improvement of agent $i$ in each turn also improves the joint policy. Thus, as agents perform policy iteration by turn, the joint policy sequence $\{\bm{\pi}\}$ improves monotonically, and $\{\bm{\pi}\}$ will converge to a Nash equilibrium since no agents can improve the joint policy unilaterally at convergence. ◻ +::: + +
+ +
Illustration of multi-agent alternate policy iteration (upper panel) and multi-agent alternate Q-iteration (lower panel) of three agents. As essentially the MDP differs at the different turn of each agent, policy iteration/Q-iteration of each agent iterates over different MDPs.
+
+ +Lemma [1](#lemma-policy-iteration){reference-type="ref" reference="lemma-policy-iteration"} immediately indicates an approach for Dec-MARL with convergence guarantee and also tells us that if we find the optimal policy for agent $i$ in each round $k$ given the other agents' policies $\pi_{-i}^k$, then the joint policy will obtain the largest improvement. This result can be formulated as following, $$\begin{equation} +\label{eq:pimax} +\begin{aligned} + &\pi^{*,k}_i = \arg\max_{\pi_i} \mathbb{E}_{\pi^{k}_{-i}}\left[ Q_{\pi_i,\pi_{-i}^{k}}(s,a_i,a_{-i} ) \right] \\ + &V_{\pi_i,{\pi}_{-i}^{k}}(s) \le V_{\pi^{*,k}_i,{\pi}_{-i}^k}(s) \quad \forall \pi_i, \forall s. +\end{aligned} +\end{equation}$$ We could obtain this $\pi^{*,k}_i$ by policy iteration with many *on-policy* iterations. However, such a method will face the issue of sample inefficiency which may be amplified in MARL settings. We will use Q-iteration to settle this problem in the next section. + +To address the problem of multi-agent alternate policy iteration, we propose *multi-agent alternate Q-iteration*, which is sufficiently truncated for fast learning but still has the same theoretical guarantee. Further, based on multi-agent alternate Q-iteration, we derive *multi-agent alternate Q-learning*, which makes the minimal change to IQL to form a simple yet effective value-based decentralized learning method for cooperative MARL. + +Instead of policy iteration, we let agents perform Q-iteration by turns as depicted in Figure [1](#fig:api){reference-type="ref" reference="fig:api"}. Let $\mathcal{M}^k_i=\{P_i^k,r^k_i\}$ denote the MDP of agent $i$ in round $k$, where we have $\mathcal{M}^{k}_i \neq \mathcal{M}^{k-1}_i$ unless $\pi_{-i}$ has converged, and $Q^{t,k}_{i}(s,a_i)$ denote the Q-function of agent $i$ with $t$ updates in the round $k$. We define the Q-iteration as follows, $$\begin{align} +\label{q-learning} + Q^{t+1,k}_{i}(s,a_i) \leftarrow r^k_i(s,a_i) + \gamma \mathbb{E}_{s^\prime \sim P_i^k} \left[ \max_{a_i^\prime} Q^{t,k}_{i}(s^\prime,a^\prime_i) \right]. +\end{align}$$ Then, the sequence $\{Q^{t,k}_{i}\}$ converges to $Q^{*,k}_{i}$ with respect to the MDP $\mathcal{M}_i^k=\{P_i^k,r_i^k\}$, and we have the following lemma. + +::: {#lemma-varepsilon .lemma} +**Lemma 2** ($\varepsilon$-convergent Q-iteration). *By iteratively applying Q-iteration [\[q-learning\]](#q-learning){reference-type="eqref" reference="q-learning"} at each agent $i$ for each turn, for any $\varepsilon>0$, we have $$\begin{equation} +\begin{aligned} + \big\|{Q^{t,k}_{i} - Q^{*,k}_{i}}\big\|_\infty \le \varepsilon, \quad \text{when \,\,} t \ge \frac{\log\left( (1- \gamma) \varepsilon \right) - \log (2R + 2\varepsilon)}{\log \gamma}, +\end{aligned} +\end{equation}$$ where $R = \frac{r_{\max}}{1-\gamma}$ and $r_{\max} = \max_{s,\bm{a}}r(s,\bm{a})$.* +::: + +::: proof +*Proof.* From the definition of $Q^{t,k}_{i}$ [\[q-learning\]](#q-learning){reference-type="eqref" reference="q-learning"}, we have $$\begin{equation} +\begin{aligned} + \big\|Q^{t+1,k}_{i} - Q^{t,k}_{i}\big\|_\infty & = \big\|\gamma \mathbb{E}_{s'\sim P_i^k}[\operatorname{max}_{a_i^\prime} Q^{t,k}_i(s^\prime,a_i^\prime) - \operatorname{max}_{a_i^\prime} Q^{t-1,k}_i(s^\prime,a_i^\prime)] \big\|_\infty \\ + & \le \gamma \big\|Q^{t,k}_i - Q^{t-1,k}_i \big\|_\infty \le \gamma^t \big\|Q^{1,k}_i - Q^{0,k}_i \big\|_\infty. +\end{aligned} +\end{equation}$$ Then for any integer $m \ge 1$, we have $$\begin{equation} +\begin{aligned} + \big\|Q^{t+m,k}_{i} - Q^{t,k}_{i}\big\|_\infty & \le \big\|Q^{t+m,k}_{i} - Q^{t+m-1,k}_{i}\big\|_\infty + \cdots + \big\|Q^{t+1,k}_{i} - Q^{t,k}_{i}\big\|_\infty \\ + & \le \gamma^t \frac{1 - \gamma^m}{1-\gamma} \big\| Q^{1,k}_i - Q^{0,k}_i\big\|_\infty. +\end{aligned} +\end{equation}$$ Let $m \to \infty$, and we have $$\begin{equation} +\label{td1} + \begin{aligned} + \big\| Q^{*,k}_{i} - Q^{t,k}_{i} \big\|_\infty & \le \frac{\gamma^t}{1-\gamma}\big\| Q^{1,k}_i - Q^{0,k}_i \big\|_\infty \\ + & \le \frac{\gamma^t}{1-\gamma} \max_{s,a_i} \big| r^k_i(s,a_i) + \gamma \mathbb{E}_{s'\sim P_i^k}[\operatorname{max}_{a_i^\prime} Q^{0,k}_i(s^\prime,a_i^\prime)] - Q^{0,k}_i(s,a_i) \big|. + \end{aligned} +\end{equation}$$ *As all agents update by turns*, we have $Q^{0,k}_i = Q^{t^{k-1}_{i},k-1}_{i}$, where $t^{k-1}_{i}$ is the number of Q-iteration for agent $i$ in the $k-1$ round. Therefore, we have $$\begin{equation} + \big\|{Q^{0,k}_i - Q^{*,k-1}_{i}\big\|_\infty = \big\| Q^{t^{k-1}_{i},k-1}_{i-1} - Q^{*,k-1}_{i}} \big\|_\infty \le \varepsilon. + \label{induct-prop} +\end{equation}$$ With this property, we have $$\begin{align} + & \big| r^k_i(s,a_i) + \gamma \mathbb{E}_{s' \sim P_i^k}[\operatorname{max}_{a_i^\prime} Q^{0,k}_i(s^\prime,a_i^\prime)] - Q^{0,k}_i(s,a_i) \big| \nonumber \\ + & = \big| r^k_i(s,a_i) + \gamma \mathbb{E}_{s' \sim P_i^k}[\operatorname{max}_{a_i^\prime} Q^{0,k}_i(s^\prime,a_i^\prime)] - Q^{*,k-1}_{i}(s,a_i) + Q^{*,k-1}_{i}(s,a_i) - Q^{0,k}_i(s,a_i) \big| \nonumber \\ + & \le \big| r^k_i - r^{k-1}_{i} \big| + \gamma \big| \mathbb{E}_{s' \sim P_i^k}[\operatorname{max}_{a_i^\prime} Q^{0,k}_i(s^\prime,a_i^\prime)] - \mathbb{E}_{s' \sim P_i^{k-1}}[\operatorname{max}_{a_i^\prime} Q^{*,k-1}_{i}(s^\prime,a_i^\prime)] \big| \nonumber \\ + & \quad + \big|Q^{*,k-1}_{i}(s,a_i) - Q^{0,k}_i(s,a_i) \big| \le 2r_{\max} + (\frac{2\gamma r_{\max}}{1-\gamma} + \varepsilon) +\varepsilon = 2R + 2\varepsilon \label{td2}, +\end{align}$$ where the second term in the last inequality is from $\|Q^{*,k-1}_{i}\|_\infty \le \frac{r_{\max}}{1-\gamma}$, $\|Q^{0,k}_{i}\|_\infty \le \|Q^{*,k-1}_{i}\|_\infty + \varepsilon$, and [\[induct-prop\]](#induct-prop){reference-type="eqref" reference="induct-prop"}. Finally, by combining [\[td1\]](#td1){reference-type="eqref" reference="td1"} and [\[td2\]](#td2){reference-type="eqref" reference="td2"}, we have $$\begin{equation} + \big\|Q^{*,k}_{i} - Q^{t,k}_{i}\big\|_\infty \le \frac{\gamma^t}{1-\gamma}(2R+2\varepsilon). +\end{equation}$$ We need $\|Q^{*,k}_{i} - Q^{t,k}_{i}\|_\infty \le \varepsilon$, which can be guaranteed by $t \ge \frac{\log\left( (1- \gamma) \varepsilon \right) - \log (2R +2\varepsilon)}{\log \gamma}$. ◻ +::: + +::: {#corollary-t .corollary} +**Corollary 1**. *For any $\varepsilon > 0$, if we take sufficient Q-iteration $t^k_i$, *i.e.*, $Q^k_i = Q^{t_i^k,k}_i$, then we have $$\begin{align*} + \big\|Q^k_i - Q^{*,k}_i\big\|_\infty \le \varepsilon \quad \forall k, i. +\end{align*}$$* +::: + +With Lemma [1](#lemma-policy-iteration){reference-type="ref" reference="lemma-policy-iteration"}, Lemma [2](#lemma-varepsilon){reference-type="ref" reference="lemma-varepsilon"}, and Corollary [1](#corollary-t){reference-type="ref" reference="corollary-t"}, we have the following theorem. + +::: {#theorem-q-iteration .theorem} +**Theorem 1** (multi-agent alternate Q-iteration). *Suppose that $Q^*_i(s,\cdot)$ has the unique maximum for all states and all agents. If all agents in turn take Q-iteration to $\|Q^k_i - Q^{*,k}_i\|_\infty \le \varepsilon$, then their joint policy sequence $\{\boldsymbol{\pi}^k\}$ converges to a Nash equilibrium, where $\pi_i^k(s) = \arg\max_{a_i}Q_i^k(s,a_i)$.* +::: + +::: proof +*Proof.* First, from Lemma [1](#lemma-policy-iteration){reference-type="ref" reference="lemma-policy-iteration"}, we know $Q^{*,k}_i$ also induces a joint policy improvement, thus $Q^{*,k}_i$ converges to $Q^{*}_i$. Let $\pi_i^*(s)=\arg\max_{a_i}Q^*_i(s,a_i)$, then $\bm{\pi}^*$ is the joint policy of a Nash equilibrium. + +Then, we define $\Delta$ as $$\begin{equation} + \Delta = \min_{s,i} \max_{a_i \not = \pi^*_i(s)} |Q^*_i(s,\pi^*_i(s)) - Q^*_i(s,a_i) |. +\end{equation}$$ From the assumption we know that $\Delta > 0$. We take $\varepsilon = \frac{\Delta}{6}$, and from Lemma [2](#lemma-varepsilon){reference-type="ref" reference="lemma-varepsilon"}, we know there exists $k_0$ such that $$\begin{equation} + \big\|Q^*_i - Q^{*,k}_i\big\|_\infty \le \varepsilon \quad \forall k \ge k_0. +\end{equation}$$ For $k\ge k_0$ and any action $a_i \not = \pi^*_i(s)$, we have $$\begin{equation} + \begin{aligned} + Q^k_i(s,& \pi^*_i(s)) - Q^k_i(s,a_i) \\ + & = Q^k_i(s,\pi^*_i(s)) - Q^{*,k}_i(s,\pi^*_i(s)) + Q^{*,k}_i(s,\pi^*_i(s)) - Q^{*}_i(s,\pi^*_i(s)) \\ + & \quad + Q^{*}_i(s,\pi^*_i(s)) - Q^{*}_i(s,a_i) + Q^{*}_i(s,a_i) - Q^{*,k}_i(s,a_i) + Q^{*,k}_i(s,a_i) - Q^k_i(s,a_i) \\ + & \ge Q^{*}_i(s,\pi^*_i(s)) - Q^{*}_i(s,a_i) - |Q^k_i(s,a_i) - Q^{*,k}_i(s,a_i)| - |Q^{*,k}_i(s,a_i) - Q^{*}_i(s,a_i)|\\ + & \quad - |Q^{*}_i(s,\pi^*_i(s)) - Q^{*,k}_i(s,\pi^*_i(s))| - |Q^{*,k}_i(s,\pi^*_i(s)) -Q^k_i(s,\pi^*_i(s))| \\ + & = \Delta - 4\varepsilon = \Delta/3 > 0, + \end{aligned} +\end{equation}$$ which means $\pi^k_i(s) = \arg\max_{a_i} Q^k_i(s,a_i) = \arg\max_{a_i} Q^*_i(s,a_i) = \pi^*_i(s)$. Thus, $Q_i^{k}$ of each agent $i$ induces $\pi_i^*$ and all together induce $\bm{\pi}^*$, which the joint policy of a Nash equilibrium. ◻ +::: + +Theorem [1](#theorem-q-iteration){reference-type="ref" reference="theorem-q-iteration"} assumes that for each agent, $Q_i^*$ has the unique maximum for all states. Although this may not hold in general, in practice we can easily settle this by introducing a positive random noise to the reward function. Suppose the random noise is bounded by $\delta$, then we can easily derive that the performance drop of optimizing environmental reward plus noise is bounded by $\delta/(1-\gamma)$. As we can make $\delta$ arbitrarily small, the bound is tight. + +From Theorem [1](#theorem-q-iteration){reference-type="ref" reference="theorem-q-iteration"}, we know that if each agent $i$ guarantees $\varepsilon$-convergence to $Q_i^{*,k}$ in each round $k$, multi-agent alternate Q-iteration also guarantees a Nash equilibrium of the joint policy. This immediately suggests a simple, practical decentralized learning method, namely multi-agent alternate Q-learning (MA2QL). + +For learning Q-table or Q-network, MA2QL makes the minimal changes to IQL. + +- For learning Q-tables, all agents in turn update their Q-tables. At a round $k$ of an agent $i$, all agents interact in the environment, and the agent $i$ updates its Q-table a few times using the collected transitions $\left$. + +- For learning Q-networks, all agents in turn update their Q-networks. At a round of an agent $i$, all agents interact in the environment and each agent $j$ stores the collected transitions $\left$ into its replay buffer, and the agent $i$ updates its Q-network using sampled mini-batches from its replay buffer. + +There is a slight difference between learning Q-table and Q-network. Strictly following multi-agent alternate Q-iteration, Q-table is updated by transitions sampled from the current MDP. On the other hand, Q-network is updated by minibatches sampled from the replay buffer. If the replay buffer only contains the experiences sampled from current MDP, learning Q-network also strictly follows multi-agent alternate Q-iteration. However, in practice, we slightly deviate from that and allow the replay buffer to contain transitions of past MDPs, following IQL [@sunehag2018vdn; @rashid2018qmix; @papoudakis2021benchmarking] for sample efficiency, the convergence may not be theoretically guaranteed though. + +MA2QL and IQL can be simply summarized and highlighted as ***MA2QL agents take turns to update Q-functions by Q-learning, whereas IQL agents simultaneously update Q-functions by Q-learning.*** diff --git a/2210.08933/main_diagram/main_diagram.drawio b/2210.08933/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..0740e5beaaafbdc0a2ef29e46146cb056d81f4ca --- /dev/null +++ b/2210.08933/main_diagram/main_diagram.drawio @@ -0,0 +1 @@ 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 \ No newline at end of file diff --git a/2210.08933/main_diagram/main_diagram.pdf b/2210.08933/main_diagram/main_diagram.pdf new file mode 100644 index 0000000000000000000000000000000000000000..bd04167e308d02c56d0063f9827d4ffdc1aba09b Binary files /dev/null and b/2210.08933/main_diagram/main_diagram.pdf differ diff --git a/2210.08933/paper_text/intro_method.md b/2210.08933/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..9b206082ea06b6d95bd82a605a76c2788a9879ab --- /dev/null +++ b/2210.08933/paper_text/intro_method.md @@ -0,0 +1,103 @@ +# Introduction + +Among existing generative models, GAN [\(Goodfellow et al.,](#page-9-0) [2014\)](#page-9-0) suffers from the instability issue [\(Salimans et al.,](#page-11-0) [2016\)](#page-11-0), subjecting to mode collapse [\(Metz et al.,](#page-10-0) [2017\)](#page-10-0); VAE [\(Kingma &](#page-10-1) [Welling,](#page-10-1) [2014\)](#page-10-1) has to rely on surrogate objectives to approximate maximum likelihood training and Flow-based models [\(Dinh et al.,](#page-9-1) [2017\)](#page-9-1) has to use specialized architectures to construct reversible transform. Diffusion models [\(Ho et al.,](#page-10-2) [2020;](#page-10-2) [Nichol & Dhariwal,](#page-10-3) [2021\)](#page-10-3) have circumvented several of these limitations and emerged as a new paradigm for generative models, theoretically underpinned by non-equilibrium thermodynamics [\(Sohl-Dickstein et al.,](#page-11-1) [2015\)](#page-11-1) and score-matching network [\(Song & Ermon,](#page-11-2) [2019\)](#page-11-2). To date, the major breakthroughs are in domains using continuous signals, such as vision [\(Saharia et al.,](#page-11-3) [2022a;](#page-11-3)[b;](#page-11-4) [Ramesh et al.,](#page-11-5) [2022\)](#page-11-5) and audio [\(Kong et al.,](#page-10-4) [2020\)](#page-10-4). However, extending continuous diffusion models to natural language remains an open challenge due to the inherently discrete nature of texts. + +On the basis of unconditional generation in continuous space which is illustrated in Figure [1\(](#page-1-0)a), existing efforts [\(Hoogeboom et al.,](#page-10-5) [2021;](#page-10-5) [Austin et al.,](#page-9-2) [2021\)](#page-9-2) start customizing diffusion models to text in discrete space on unconditional language modeling (i.e., free text generation). Diffusion-LM [\(Li](#page-10-6) [et al.,](#page-10-6) [2022\)](#page-10-6), as in Figure [1\(](#page-1-0)b), models texts in continuous space and proposes to use an extra-trained classifier as guidance (i.e., the condition signal x) to impose subtle changes (usually complex, finegrained constraints) on generated sentences. Nonetheless, these models do not naturally generalize to conditional language modeling (i.e., the model assigns probabilities p(w|x) to sequences of words w given x). In the more general sequence-to-sequence (SEQ2SEQ) setting where the condition x is also a sequence of words, applying Diffusion-LM can be difficult. The reason is that classifiers are attributes-oriented, and we can not train hundreds-of-thousands classifiers to model the semantic meaning between conditions and generated sentences. + +SEQ2SEQ is an essential setting in NLP that covers a wide range of important tasks such as openended sentence generation, dialogue, paraphrasing, and text style transfer. In this paper, we propose + +1Code is available at + +![](_page_1_Figure_1.jpeg) + +Figure 1: The demonstration of unconditional, classifier-guided, and classifier-free diffusion models. + +DIFFUSEQ, depicted in Figure 1(c), a classifier-free diffusion model that supports SEQ2SEQ text generation tasks. By modeling the conditional probability of the target sentence $\mathbf{w}$ given context $\mathbf{x}$ using one single model, one advantage of DIFFUSEQ is that this paradigm allows a complete model to fit data distribution and utilize conditional guidance, rather than depending on a separate classifier. + +Different from canonical generation approaches in an autoregressive (AR) left-to-right manner (Radford et al., 2019), DIFFUSEQ generates text tokens parallelly in the non-autoregressive (NAR) way. To corroborate the effectiveness of our DIFFUSEQ, we conduct experiments on four SEQ2SEQ tasks. Compared to AR and NAR models, which suffer from the "degeneration" problem (Holtzman et al., 2019) and rely on decoding strategies, DIFFUSEQ can achieve considerable sentence-level diversity without sacrificing the quality (see § 4.2). + +To sum up, we make a series of technical and conceptual contributions: (a) we are the first to deploy the diffusion model on SEQ2SEQ text generation, and our proposed DIFFUSEQ as a conditional language model is trained end-to-end in a classifier-free manner; (b) we establish a theoretical connection among AR, NAR and DIFFUSEQ models, and justify DIFFUSEQ as an extension of iterative-NAR models; (c) with strong empirical evidence, we demonstrate the great potential of diffusion models in complex conditional language generation tasks. + +# Method + +**Preliminary.** A diffusion model typically contains forward and reverse processes. Given a data point sampled from a real-world data distribution $\mathbf{z}_0 \sim q(\mathbf{z})$ , the forward process gradually corrupts $\mathbf{z}_0$ into a standard Gaussian noise $\mathbf{z}_T \sim \mathcal{N}(0, \mathbf{I})$ . For each forward step $t \in [1, 2, ..., T]$ , the perturbation is controlled by $q(\mathbf{z}_t | \mathbf{z}_{t-1}) = \mathcal{N}(\mathbf{z}_t; \sqrt{1 - \beta_t} \mathbf{z}_{t-1}, \beta_t \mathbf{I})$ , with $\beta_t \in (0, 1)$ as different variance scales. Once the forward process is completed, the reverse denoising process tries to gradually reconstruct the original data $\mathbf{z}_0$ via sampling from $\mathbf{z}_T$ by learning a diffusion model $f_{\theta}$ . + +**Problem Statement.** Many recent efforts have been devoted to adapting diffusion models to discrete texts (See § 5). However, they all focus on unconditional sequence modeling. In this paper, we target the sequence-to-sequence text generation tasks. In particular, given a m-length source sequence $\mathbf{w}^x = \{w_1^x, ..., w_m^x\}$ , we aim to learn a diffusion model that can produce a n-length target sequence $\mathbf{w}^y = \{w_1^y, ..., w_n^y\}$ conditioning on the source sequence. + +We propose DIFFUSEQ to extend vanilla diffusion models to learn conditional text generation (as shown in Figure 2), concerning the model architecture and the training objective. + +Forward Process with Partial Noising. In the beginning of forward process, we follow Diffusion-LM (Li et al., 2022) to design an embedding function $EMB(\mathbf{w})$ to map the discrete text $\mathbf{w}$ into a continuous space. In particular, given a pair of sequence $\mathbf{w}^x$ and $\mathbf{w}^y$ , DIFFUSEQ learns a unified + +![](_page_2_Figure_1.jpeg) + +Figure 2: The diffusion process of our conditional diffusion language model DIFFUSEQ. Given the source $\mathbf{w}^x$ and the target $\mathbf{w}^y$ , we pair-wisely transform them into continuous space $\mathbf{z}_0$ . The partial Gaussian noise is iteratively added on the target space of $\mathbf{z}_t$ . + +feature space of $\mathbf{w}^x$ and $\mathbf{w}^y$ by embedding transformation and concatenation as $\mathrm{EMB}(\mathbf{w}^{x \oplus y}) = [\mathrm{EMB}(w_1^x),...,\mathrm{EMB}(w_m^x),\mathrm{EMB}(w_1^y),...,\mathrm{EMB}(w_n^y)] \in \mathbb{R}^{(m+n) \times d}$ . The transformation allows us to adapt discrete textual input into the standard forward process, by extending the original forward chain to a new Markov transition $q_\phi(\mathbf{z}_0|\mathbf{w}^{x \oplus y}) = \mathcal{N}(\mathrm{EMB}(\mathbf{w}^{x \oplus y}),\beta_0\mathbf{I})$ . + +We denote $\mathbf{z}_t = \mathbf{x}_t \oplus \mathbf{y}_t$ to simplify the wordings, where $\mathbf{x}_t$ and $\mathbf{y}_t$ represent parts of $\mathbf{z}_t$ that belong to $\mathbf{w}^x$ and $\mathbf{w}^y$ , respectively. For each forward step $q(\mathbf{z}_t|\mathbf{z}_{t-1})$ , we gradually inject noise into last step's hidden state $\mathbf{z}_{t-1}$ to obtain $\mathbf{z}_t$ . Unlike conventional diffusion models that corrupt the whole $\mathbf{z}_t$ (both $\mathbf{x}_t$ and $\mathbf{y}_t$ ) without distinction, we only impose noising on $\mathbf{y}_t$ . This modification (termed **partial noising**) allows us to adapt diffusion models for conditional language modeling. + +Reverse Process with Conditional Denoising. The ultimate goal of the reverse process is to recover the original $\mathbf{z}_0$ by denoising $\mathbf{z}_t$ : $p_{\theta}(\mathbf{z}_{0:T}) := p(\mathbf{z}_T) \prod_{t=1}^T p_{\theta}(\mathbf{z}_{t-1}|\mathbf{z}_t)$ . We model the learning process $p_{\theta}(\mathbf{z}_{t-1}|\mathbf{z}_t) = \mathcal{N}(\mathbf{z}_{t-1}; \mu_{\theta}(\mathbf{z}_t, t), \sigma_{\theta}(\mathbf{z}_t, t))$ using the proposed diffusion model DIFFUSEQ: $f_{\theta}(\mathbf{z}_t, t)$ , where the $\mu_{\theta}(\cdot)$ and $\sigma_{\theta}(\cdot)$ is the parameterization of the predicted mean and standard deviation of $q(\mathbf{z}_{t-1}|\mathbf{z}_t)$ in forward process, derived using Bayes' rule. The detailed derivations are in Appendix A. With the partial nosing strategy adopted in the forward process, we can impose the input as the condition when denoising as shown in Figure 1. The proposed conditional denoising is classifier-free by nature: we do not require extra-trained classifiers to control the denoising process. + +Specifically, we use a transformer architecture to model $f_{\theta}$ , which spontaneously models the semantic relation between $\mathbf{x}_t$ and $\mathbf{y}_t$ . We compute the variational lower bound ( $\mathcal{L}_{VLB}$ ) following the original diffusion process. $\mathcal{L}_{round}$ corresponds to rounding operation in Figure 2. + +$$\mathcal{L}_{\text{VLB}} = \mathbb{E}_{q(\mathbf{z}_{1:T}|\mathbf{z}_{0})} \left[ \underbrace{\log \frac{q(\mathbf{z}_{T}|\mathbf{z}_{0})}{p_{\theta}(\mathbf{z}_{T})}}_{\mathcal{L}_{T}} + \sum_{t=2}^{T} \underbrace{\log \frac{q(\mathbf{z}_{t-1}|\mathbf{z}_{0}, \mathbf{z}_{t})}{p_{\theta}(\mathbf{z}_{t-1}|\mathbf{z}_{t})}}_{\mathcal{L}_{t-1}} + \underbrace{\log \frac{q_{\phi}(\mathbf{z}_{0}|\mathbf{w}^{x \oplus y})}{p_{\theta}(\mathbf{z}_{0}|\mathbf{z}_{1})}}_{\mathcal{L}_{0}} - \underbrace{\log p_{\theta}(\mathbf{w}^{x \oplus y}|\mathbf{z}_{0})}_{\mathcal{L}_{\text{grand}}} \right].$$ + +$$(1)$$ + +We further simplify the training objective as follows (details in Appendix A): + +$$\min_{\theta} \mathcal{L}_{\text{VLB}} = \min_{\theta} \left[ \sum_{t=2}^{T} ||\mathbf{z}_0 - f_{\theta}(\mathbf{z}_t, t)||^2 + ||\text{EMB}(\mathbf{w}^{x \oplus y}) - f_{\theta}(\mathbf{z}_1, 1)||^2 - \log p_{\theta}(\mathbf{w}^{x \oplus y} | \mathbf{z}_0) \right] +\rightarrow \min_{\theta} \left[ \sum_{t=2}^{T} ||\mathbf{y}_0 - \tilde{f}_{\theta}(\mathbf{z}_t, t)||^2 + ||\text{EMB}(\mathbf{w}^y) - \tilde{f}_{\theta}(\mathbf{z}_1, 1)||^2 + \mathcal{R}(||\mathbf{z}_0||^2) \right], \tag{2}$$ + +here we use $\tilde{f}_{\theta}(\mathbf{z}_t, t)$ to denote the fractions of recovered $\mathbf{z}_0$ corresponding to $\mathbf{y}_0$ . Note that although in the first term, we only compute the loss w.r.t $\mathbf{y}_0$ , due to the attention mechanism in the transformer, the reconstruction of $\mathbf{y}_0$ also takes $\mathbf{x}_0$ into account, thus the gradients from the first term + +will also affect the learning of $\mathbf{x}_0$ . The mathematically equivalent regularization term $\mathcal{R}(||\mathbf{z}_0||^2)$ ) regularize the embedding learning. We further share the embedding function between source and target sequences, enabling the training of two different feature spaces jointly. This sets DIFFUSEQ away from existing solutions in vision such as GLIDE (Nichol et al., 2022). + +**Training and Inference Methods.** In our preliminary experiments, we find that the high diversity in NLP datasets and long diffusion steps often result in insufficient training. We hypothesize the reason is that sampling step t uniformly causes unnecessary noise in the $\mathcal{L}_{VLB}$ objective. We hence employ importance sampling (Nichol & Dhariwal, 2021) to address this problem. + +$$\mathcal{L}_{\text{VLB}} = \mathbb{E}_{t \sim p_t} \left[ \frac{\mathcal{L}_t}{p_t} \right], \ p_t \propto \sqrt{\mathbb{E}[\mathcal{L}_t^2]}, \ \sum_{t=0}^{T-1} p_t = 1.$$ + (3) + +Intuitively, the importance-weighted sampling algorithm will spend more steps on diffusion steps with larger $\mathcal{L}_t$ , and vice versa. + +To conduct SEQ2SEQ generation given the condition EMB( $\mathbf{w}^x$ ), we randomly sample $\mathbf{y}_T \sim \mathcal{N}(0,I)$ and concatenate $\mathbf{y}_T$ with EMB( $\mathbf{w}^x$ ) to obtain $\mathbf{z}_T$ . We can now repeat the reverse process until we arrive at $\mathbf{z}_0$ . At each sampling step, an anchoring function is executed towards reparameterized $\mathbf{z}_t$ . Specifically, the anchoring function: (a) operates rounding on $\mathbf{z}_t$ to map it back to word embedding space following Li et al. (2022); (b) replaces the part of recovered $\mathbf{z}_{t-1}$ that belongs to $\mathbf{w}^x$ with the original $\mathbf{x}_0$ , considering that this part is recovered from corrupted $\mathbf{z}_t$ via $f_\theta$ and not strictly equals to $\mathbf{x}_0$ . Note that (b) is designed for DIFFUSEQ. + +To improve the quality of generation, we apply the widely used Minimum Bayes Risk (MBR) decoding strategy (Koehn, 2004). We first generate a set of candidate samples $\mathcal S$ from different random seeds of DIFFUSEQ and select the best output sequence that achieves the minimum expected risk under a meaningful loss function (e.g. BLEU or other cheaper metrics like precision). In practice, we use the negative BLEU score in our implementation. + +Connections to AR, Iter-NAR, and Fully-NAR Models. To better understand the behavior of DIFFUSEQ, we give the theoretical connection to autoregressive (AR), iterative non-autoregressive (iter-NAR), and fully non-autoregressive (fully-NAR) models. We argue that DIFFUSEQ can be seen as an extension of iter-NAR model. Detailed graphical learning discrepancies of these four cases are discussed in Appendix B for reference. + +AR models learn $p(\mathbf{w}_{1:n}^y|\mathbf{w}^x)$ by autoregressive decomposition based on left-context: + + +$$p_{\text{AR}}(\mathbf{w}_{1:n}^{y}|\mathbf{w}^{x}) = \underbrace{p(w_{1}^{y}|\mathbf{w}^{x})}_{\text{initial prediction}} \underbrace{\prod_{i=1,\dots,n-1} p(w_{i+1}^{y}|\mathbf{w}_{1:i}^{y},\mathbf{w}^{x})}_{\text{progressive left-context prediction}},$$ +(4) + +while fully-NAR models (Gu et al., 2018; Qian et al., 2021) learn the conditional probability given independent assumption for fast inference: + + +$$p_{\text{fully-NAR}}(\mathbf{w}_{1:n}^y|\mathbf{w}^x) = \prod_{i=1,\dots,n} p(w_i^y|\mathbf{w}^x).$$ + (5) + +To make a better analogy to AR and NAR models, we use a lossless way to formulate iterative NAR models (Gu et al., 2019; Ghazvininejad et al., 2019) by introducing a series of intermediate sequences $\mathbf{w}_{1:K-1}^y$ , $\mathbf{w}_K^y = \mathbf{w}^y$ with K editable iterations: + + +$$p_{\text{iter-NAR}}(\mathbf{w}_{1:n}^{y}|\mathbf{w}^{x}) = \sum_{\mathbf{w}_{1}^{y}, \dots, \mathbf{w}_{K-1}^{y}} \underbrace{\prod_{i=1\dots n} p(w_{1,i}^{y}|\mathbf{w}^{x})}_{\text{initial prediction}} \underbrace{\prod_{k=1\dots K-1} \prod_{i=1\dots n} p(w_{k+1,i}^{y}|\mathbf{w}_{k,1:n}^{y}, \mathbf{w}^{x})}_{\text{progressive full-context prediction}}.$$ +(6) + +Previous study (Huang et al., 2022) shows that there is a gap called *conditional total correlation* between AR Eq. (4) and fully-NAR Eq. (5) learning paradigms, because of lossy decomposition of NAR models. However, when comparing iter-NAR Eq. (6) with AR Eq. (4) models, they both can be factorized into an initial prediction term and a progressive prediction process based on different context (i.e. left-context in AR and full-context in iter-NAR), and the discrepancy pointed out by + +Huang et al. (2022) is therefore closed in iter-NAR assuming sufficient steps. By showing DIF-FUSEQ is an extension of the iter-NAR model, we offer a justification that it will not suffer from the conditional total correlation for the same reason. + +A straight-forward way to formulate pure continuous diffusion models is to introduce a series of Gaussian noise-corrupted features along with diffusion steps: $\mathbf{y}_{1:T-1}, \mathbf{y}_0 = \mathbf{y}, \mathbf{y}_T \sim \mathcal{N}(0, \mathbf{I})$ . + + +$$p_{\text{diffusion}}(\mathbf{w}^y|\mathbf{w}^x) = \int_{\mathbf{y}_T, \dots, \mathbf{y}_0} \underbrace{p(\mathbf{w}^y|\mathbf{y}_0, \mathbf{w}^x)}_{\text{final prediction}} \underbrace{\prod_{t=T, \dots, 1} p(\mathbf{y}_{t-1}|\mathbf{y}_t, \mathbf{w}^x)}_{\text{progressive full-context diffusion}}, \tag{7}$$ + +where $p(\mathbf{y}_{t-1}|\mathbf{y}_t, \mathbf{w}^x)$ describes the diffusion step on continuous representations $\mathbf{y}$ . The rounding operation in DIFFUSEQ maps the continuous vectors $\mathbf{y}$ to discrete $\mathbf{w}^y$ for each time step t, we in addition introduce this into Eq. (7): + +$$p_{\text{DIFFUSEQ}}(\mathbf{w}^y|\mathbf{w}^x) = \sum_{\mathbf{w}_T^y, \dots, \mathbf{w}_1^y} \int_{\mathbf{y}_T, \dots, \mathbf{y}_0} p(\mathbf{w}^y|\mathbf{y}_0, \mathbf{w}^x) \prod_{t=T, \dots, 1} p(\mathbf{w}_t^y|\mathbf{y}_t, \mathbf{w}^x) p(\mathbf{y}_{t-1}|\mathbf{w}_t^y)$$ +(8) +$$= \sum_{\mathbf{w}_T^y, \dots, \mathbf{w}_1^y} \int_{\mathbf{y}_T, \dots, \mathbf{y}_0} p(\mathbf{w}_T^y|\mathbf{y}_T, \mathbf{w}^x) \prod_{t=T-1, \dots, 0} p(\mathbf{y}_t|\mathbf{w}_{t+1}^y) p(\mathbf{w}_t^y|\mathbf{y}_t, \mathbf{w}^x).$$ +(9) + +By rearranging Eq. (8) into Eq. (9), we can see DIFFUSEQ can be seen as a more generalized form of iter-NAR Eq. (6) before marginalizing out $\{y_T, \ldots, y_0\}$ , despite the different initialization of $y_T^2$ . A more detailed derivation is shown in Appendix C. diff --git a/2210.13005/main_diagram/main_diagram.drawio b/2210.13005/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..78a5666809b52a02253e263b0d6c9aaf11ff9100 --- /dev/null +++ b/2210.13005/main_diagram/main_diagram.drawio @@ -0,0 +1 @@ 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 \ No newline at end of file diff --git a/2210.13005/main_diagram/main_diagram.pdf b/2210.13005/main_diagram/main_diagram.pdf new file mode 100644 index 0000000000000000000000000000000000000000..98664c3674678b951a3c6ca4626802a50f0d7e85 Binary files /dev/null and b/2210.13005/main_diagram/main_diagram.pdf differ diff --git a/2210.13005/paper_text/intro_method.md b/2210.13005/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..5bf58a836c2360bed7ce31f1ccfbb96256aa826a --- /dev/null +++ b/2210.13005/paper_text/intro_method.md @@ -0,0 +1,115 @@ +# Introduction + +Real-world problem scenarios are flooded with sequential event data consisting of chronologically arrived events which reflect certain behaviors, activities or responses. A typical example is user activity prediction [@ye2013s; @feng2015personalized; @zhou2019deep; @hidasi2018recurrent; @sun2019bert4rec; @zhou2020s3] that aims to harness user's recent activities to estimate future ones, which could help downstream target advertisement in online platforms (e.g., e-commerce or social networks). Predicting future events also plays a central role in some real situations such as clinical treatment [@beunza2019comparison] for promoting social welfare. + +A common nature of (sequential) event prediction lies in the different time intervals within which training and testing data are generated. Namely, models trained with data collected at one time are supposed to predict next event in the future [@zhao2020event; @peska2020off; @rossetti2016contrasting] where the underlying data-generating distributions may have gone through variation due to environmental changes. However, most existing approaches [@hidasi2018recurrent; @granroth2016happens; @qiao2018pairwise; @kang2018self; @sun2019bert4rec; @rendle2010factorizing; @he2016fusing; @he2017translation] overlook this issue in both problem formulation and empirical evaluation, which may leave the real problems under-resolved with model mis-specification, and result in over-estimation for model performance on real data. + +As a concrete example in sequential recommenders [@fang2020deep], the data-generating distribution for user clicking behaviors over items is highly dependent on user preferences that are normally correlated with some external factors, like the dynamic fashion trends in different years [@gong2021aesthetics] or seasonal influences on item popularity [@stormer2007improving]. These highly time-sensitive external factors would induce distinct user preferences leading to different behavioral data as time goes by. This issue could also partially explain the common phenomenon of model performance drop after adaptation from offline to online environments [@rossetti2016contrasting; @peska2020off; @huzhang2021aliexpress]. + +Handling distribution shift in sequential event data poses several non-trivial challenges. *First*, the temporal shift requires models' capability for out-of-distribution (OoD) generalization [@ood-classic-1], i.e., extrapolating from training environments to new unseen environments in remote future. Prior art that focus on model learning and evaluation over in-distribution data may yield sub-optimal results on OoD testing instances. *Second*, as mentioned before, there exist external factors that impact the generation of events. These external factors, which we term as *contexts*, are unobserved in practice. To eliminate their effects, one may also need the latent distributions that characterize how the contexts affect events generation. Unfortunately, such information is often inaccessible due to constrained data collection, which requires the models to learn from pure observed sequences. + +
+

[]

+

+
A toy example in recommendation. Prior art would spuriously correlate non-causal items (‘ice cream’ and ‘beach T-shirt’) and produce undesired results under a new environment in the future. Our approach endeavors to alleviate the issue via counteracting the effects of contexts (seasons).
+
+ +To resolve these difficulties, in this paper, we adopt a generative perspective to investigate the *temporal distribution shift* problem and propose a new variational context adjustment approach with instantiations to solve the issue for sequential event prediction. + +**A generative perspective on temporal distribution shift.** We use proof-of-concept Structural Causal Models (SCMs) to characterize the dependency among contexts, historical sequences and labels (i.e., next event type) in terms of both data generation and model prediction. We show that contexts essentially act as a confounder, which leads the model to leverage spurious correlations and fail to generalize to data from new distributions. See Fig. [1](#fig_motiv){reference-type="ref" reference="fig_motiv"} for a toy example to illustrate the issue. + +**Variational context adjustment.** We propose a variational context adjustment approach to resolve the issue. The resulting new learning objective has two-fold effects: 1) helping to uncover the underlying relations between latent contexts and sequences in a data-driven way, and 2) canceling out the confounding effect of contexts to facilitate the model to explore the true causality of interest (as illustrated in Fig. [1](#fig_motiv){reference-type="ref" reference="fig_motiv"}). We also propose a mixture of variational posteriors to approximate context prior via randomly simulating pseudo input sequences. + +**Instantiation by a flexible framework.** We propose a framework named as CaseQ to instantiate ingredients in the objective, which could combine with most off-the-shelf sequence backbone models. To accommodate temporal patterns under different environments, we devise a novel hierarchical branching structure for learning context-specific representations of sequences. It could dynamically evolves its architecture to adapt for variable contexts. + +**Empirical results.** We carry out comprehensive experiments on three sequential event prediction tasks with valuation protocols designed for testing model performance under temporal distribution shift. Specifically, when we enlarge time gap between training and testing data, CaseQ can alleviate performance drop by $47.77\%$ w.r.t. Normalized Discounted Cumulative Gain (NDCG) and $35.73\%$ w.r.t. Hit Ratio (HR) for sequential recommendation, which shows its robustness against temporal distribution shift. + +# Method + +We denote $\mathcal X=\{1,2,\cdots,M\}$ as the space of event types, and assume each event is assigned with an event type $x_i\in \mathcal X$. Events occurred in chronological order consist of a sequence $\mathcal S =\{x_1, x_2, \cdots, x_{|\mathcal S|}\}$. As mentioned before, the data distribution is normally affected by time-dependent external factors, i.e., *context* $c$. We use $S$, $Y$, $\hat Y$ and $C$ to denote the random variables of historical event sequence $\mathcal S$, ground-truth next event $y$, predicted next event $\hat y$ and context $c$, respectively. The data distribution can be characterized as $P(S, Y| C) = P(S| C) P(Y| S, C)$. + +Given training data $\{(\mathcal S_i, y_i)\}_{i=1}^N$ generated from data distributions with $(\mathcal S_i, y_i)\sim P(S, Y| C = c_{tr}^{(i)})$, where $c_{tr}^{(i)}$ denotes the specific context when the $i$-th training sample is generated, we are to learn a prediction model $\hat y_i = f(\mathcal S_i; \theta)$ that can generalize to testing data $\{(\mathcal S_j, y_j)\}_{j=1}^{N'}$ from new distributions with $(\mathcal S_j, y_j)\sim P(S, Y| C = c_{te}^{(j)})$, where $c_{te}^{(j)}$ denotes the specific context when the $j$-th testing sample is generated. The distribution shift stems from different contexts that change over time, which we call *temporal distribution shift* in this paper. + +
+ +
Structural causal model for sequence learning.
+
+ +Most of existing approaches target maximizing the likelihood $P_\theta(y|\mathcal S)$ as the objective for optimization. Here we use $P_\theta(\cdot)$ to denote the distribution induced by prediction model $f_\theta$. Based on the definitions, we can build two Structural Causal Models (SCMs) that interpret the causal relations among 1) $C$, $S$, $Y$ (given by data-generating process) and 2) $C$, $S$ and $\hat{Y}$ (given by model learning), as shown in Fig. [2](#fig_diag){reference-type="ref" reference="fig_diag"}(a) and (b). + +For Fig. [2](#fig_diag){reference-type="ref" reference="fig_diag"}(a), the three causal relations are given by the definitions for data generation, i.e., $P(S, Y|C) = P(S| C) P(Y|S, C)$. We next illustrate the rationales behind the other two causal relations $S\rightarrow \hat Y$ and $C\rightarrow \hat Y$. + +$\boldsymbol{S\rightarrow \hat Y}$: This relation is induced by the prediction model $\hat y = f(\mathcal S;\theta)$ that takes a historical event sequence $\mathcal S$ as input and outputs the prediction for next event $\hat y$. The relation from $\mathcal S$ to $\hat y$ is deterministic given fixed model parameters $\theta$. + +$\boldsymbol{C\rightarrow \hat Y}$: This relation is implicitly embodied in the learning process that optimizes the model parameters with a given dataset collected at one time. By our definition, the training dataset is generated from a latent distribution affected by context $c_{tr}$ and the MLE algorithm yields $$\begin{equation} + \theta^* = \arg\min_{\theta}\mathbb E_{(\mathcal S,y)\sim P(S,Y|C=c_{tr})} [l\left(f(\mathcal S; \theta), y\right)], +\end{equation}$$ where $l(\cdot, \cdot)$ denotes a certain loss function, e.g., cross-entropy. This indicates that the learned model parameters $\theta^*$ is dependent on the distribution of $c_{tr}$. Also, due to the fact $\hat y = f(\mathcal S;\theta)$, we conclude the proof for relation from $C$ to $\hat Y$. In short, intuitive explanations for such a causal relation lie in two facts: 1) $C$ affects the generation of data used for model training, and 2) $\hat Y$ is the output of the trained model given input sequence $S$. + +The key observation is that $C$ acts as the confounder in both Fig. [2](#fig_diag){reference-type="ref" reference="fig_diag"}(a) and (b), which could play a crucial role leading to undesirable testing performance of MLE-based approaches once the distribution is shifted. As implied by the causal relations in Fig. [2](#fig_diag){reference-type="ref" reference="fig_diag"}(a), there exists partial information in $S$ that is predictive for $Y$ yet highly sensitive to $C$, a usually fast-changing variable in real-world scenarios. As a result, the correlation between $S$ and $Y$ in previous contexts may become spurious in future ones. This also explains the failure of MLE-based models for generalizing to data from new distributions, according to the similar causal pattern in Fig. [2](#fig_diag){reference-type="ref" reference="fig_diag"}(b). + +We can reuse the toy example in Fig. [1](#fig_motiv){reference-type="ref" reference="fig_motiv"} as an intuitive interpretation for the failure. The 'summer' season (a context) acts as a confounder that correlates buying 'ice cream' (a historical event) and buying 'T-shirt' (a label), between which exists no obvious causal relation. However, the model would 'memorize' their correlation and tend to improperly recommend 'T-shirt' in the 'winter' season (new context). Whereas in fact the user purchases ice cream in winter most possibly because he/she is a dessert lover, in which case recommending other desserts would be a better decision. + +To address the confounding effect of $C$ and endow the model with robustness to temporal distribution shift, we propose to target model learning with $P_\theta(Y|do(S))$ instead of the conventional $P_\theta(Y|S)$. As shown in Fig. [\[fig_context\]](#fig_context){reference-type="ref" reference="fig_context"}(c), the $do$-operator cuts off the arrow (i.e., causal relation) coming from $C$ to $S$, which essentially simulates an ideal data-generating process where sequences are generated independently from contexts. This operation blocks the backdoor path $S\leftarrow C\rightarrow \hat Y$ that spuriously correlates $S$ and $Y$, and enables the model to learn the desired causal relation $S\rightarrow Y$ which is invariant to environmental change. + +An ideal way to compute $P_\theta(Y|do(S))$ is to carry out randomized controlled trial (RCT) [@pearl2016causal] by recollecting data from a prohibitively large quantity of random samples under any possible context, which is infeasible since we could neither control the environment nor collect data in the future. Fortunately, there exists a statistical estimation of $P_\theta(Y|do(S))$ by leveraging backdoor adjustment [@pearl2016causal], wherein the confounder $C$ is stratified into discrete pieces $\mathcal C = \{c_i\}_{i=1}^{|\mathcal C|}$. By using basic rules induced by the $do$-operator (see derivation in Appendix [6](#ap_backdoor){reference-type="ref" reference="ap_backdoor"}), we have: $$\begin{equation} + \label{eqn_backdoor} + P_\theta(Y|do(S)) = \sum_{i=1}^{|\mathcal C|} P_\theta(Y|S, C = c_i) P(C = c_i). +\end{equation}$$ Intuitively, the backdoor adjustment approximates an ideal situation where $c_i$ is enumerated according to the context prior $P(C)$ independent from input sequences, which serves to counteract the effect of $C$ on the generation of $S$. Nevertheless, optimizing with Eq. [\[eqn_backdoor\]](#eqn_backdoor){reference-type="eqref" reference="eqn_backdoor"} is intractable since $c_i$ is usually unobserved or even undefined, and its prior distribution $P(C)$ is also unknown. Furthermore, computing Eq. [\[eqn_backdoor\]](#eqn_backdoor){reference-type="eqref" reference="eqn_backdoor"} requires awareness of the relation between contexts and generated sequences, which is implicit in the data-generating process behind the data. + +To address the difficulty for targeting Eq. [\[eqn_backdoor\]](#eqn_backdoor){reference-type="eqref" reference="eqn_backdoor"}, we introduce a variational distribution $Q(C|S)$ as the estimation for latent contexts given input sequences. By treating $C$ as a latent variable and using variational inference technique, we can obtain the following tractable evidence lower bound (ELBO) as learning objective (see derivation in Appendix [7](#ap_variational){reference-type="ref" reference="ap_variational"}): $$\begin{equation} + \label{eqn_obj} +\begin{split} + &\log P_\theta(Y|do(S=\mathcal S)) \\ + \geq & \mathbb{E}_{c\sim Q(C|S=\mathcal S)} \left [ \log P_\theta(Y|S=\mathcal S, C=c) \right ] - \mathcal{D}_{KL}\left(Q(C|S=\mathcal S) \| P(C)\right), +\end{split} +\end{equation}$$ where the last step is given by Jensen's Inequality and the equality holds if and only if $Q(C|S)$ exactly fits the true posterior $P(C|S, Y)$, which suggests it successfully uncovers the latent context from observed data. The first component in Eq. [\[eqn_obj\]](#eqn_obj){reference-type="eqref" reference="eqn_obj"} is the negative reconstruction error (i.e. prediction error). The second component is the Kullback--Leibler (KL) divergence of the variational distribution and context prior distribution. Similar with [@makhzani2015adversarial], we could re-write the second term as a summation of the entropy of $Q(C|S)$, i.e., $\mathbb H[Q(C|S=\mathcal S)]$ and the cross-entropy between $Q(C|S)$ and prior distribution, i.e., $\mathbb H[Q(C),P(C)]$. The former enforces high variance of contexts for each event sequence, and the latter aligns the aggregated posterior with the prior. + +The distribution $P(C)$ characterizes the true context prior in real world. It serves to regularize $Q(C|S)$ through the KL term in Eq. [\[eqn_obj\]](#eqn_obj){reference-type="eqref" reference="eqn_obj"}. Choosing an appropriate $P(C)$ is important yet challenging. One straightforward solution adopted by some previous works [@dinh2015nice; @zhang2020causal] is to use a pre-defined prior such as uniform distribution. However, such a simplistic distribution may not reflect the true context prior, and potentially lead to over-regularization [@burda2015importance]. An alternative way is to estimate the prior by computing the average of the variational posterior [@hoffman2016elbo], i.e., $P(C) \approx 1/N \sum_{i=1}^N Q(C|S=\mathcal S_i)$. However, such a method is computationally expensive [@tomczak2018vae] and would often result in biased estimation given a limited quantity of training data collected within a certain time interval. Inspired by [@tomczak2018vae] using mixture of Gaussian as a flexible and learnable prior, we propose to use a mixture of pseudo variational posteriors as an estimation: $$\begin{equation} +\label{eqn-prior} + \hat P(C) = \frac{1}{R} \sum_{j=1}^R Q(C|S=\mathcal S'_j). +\end{equation}$$ where $\mathcal S'_j$ is a randomly generated pseudo event sequence and we set $R\ll N$ to reduce the computational cost. Note that the variational distribution $Q(C|S)$ is given by the prediction model (see details in Section [3](#sec-framework){reference-type="ref" reference="sec-framework"}). Therefore, the prior estimation by Eq. [\[eqn-prior\]](#eqn-prior){reference-type="eqref" reference="eqn-prior"} is a general reflection of how the model favors each context given uninformative inputs. Also, the estimated prior is learned with the model in a fully data-driven manner. + +We next parameterize the components in Eq. [\[eqn_obj\]](#eqn_obj){reference-type="eqref" reference="eqn_obj"}, i.e., $P_\theta(Y|S, C)$ and $Q(C|S)$, and propose a flexible framework named as CaseQ by taking 'cas' from 'causal' and 'seq' from 'sequence'. + +Given an event sequence $\mathcal S = \{x_1, x_2, \cdots, x_t\}$ as input, we use an embedding to represent each type of event. We consider a global embedding matrix $\mathbf H_x \in \mathbb R^{M\times d}$ to map each type of event into a $d$-dimensional embedding space: $$\begin{equation} + \mathbf h_m^1 = OneHot(x_m)^{\top} \mathbf H_x, +\end{equation}$$ where $M$ denotes the number of event types, $OneHot(\cdot): \mathbb Z^+ \rightarrow \{0,1\}^M$ transforms an event into a one-hot column vector, and $\mathbf h_m^1 \in \mathbb R^d$ denotes the initial representation for an event $x_m$. + +To encode input sequences and accommodate the effect of $C$ on prediction, we consider an encoder $\Phi(\cdot)$ conditioned on a specific context. It takes a sequence of event embeddings and estimated context representation $\mathbf c_{(t)}$ at time step $t$ (given by the branching unit detailed next) as input, and outputs a sequence of hidden states: $$\begin{equation} +\begin{split} + [\mathbf h_1^2, \mathbf h_2^2, \cdots, \mathbf h_t^2] &= \Phi([\mathbf h_1^1, \mathbf h_2^1, \cdots, \mathbf h_t^1], \mathbf c_{(t)}; \Theta), +\end{split} +\end{equation}$$ where $\mathbf h^2_m$ is the $m$-th hidden state for the event sequence. Assume that there are $K$ types of contexts, i.e., $\mathcal C = \{\mathbf c_k\}_{k=1}^{K}$ and $c_{(t)}\in \mathcal C$. Each type of context $c_k$ is represented as a $K$-dimensional one-hot column vector $\mathbf c_k \in \{0,1\}^K$ whose $k$-th dimension is $1$. Based on the SCM in Fig. [2](#fig_diag){reference-type="ref" reference="fig_diag"}, the context would change the underlying mechanism of how the next event is generated in real-world. Therefore, we consider context-specific *inference units* $\{\Phi_k(\cdot\;;\theta_k)\}_{k=1}^{K}$ as sub-networks of the encoder $\Phi(\cdot)$ to learn context-aware sequence representations, where $\Theta = \{\theta_k\}_{k=1}^{K}$. Then, we have $$\begin{equation} + \label{eqn_encoder} + \Phi([\mathbf h_1^1, \mathbf h_2^1, \cdots, \mathbf h_t^1], \mathbf c_{(t)}; \Theta) = \sum_{k=1}^K \mathbf c_{(t)}[k] \cdot \Phi_k([\mathbf h_1^1, \cdots, \mathbf h_t^1]; \theta_k), +\end{equation}$$ where $\mathbf c_{(t)}[k]$ returns the $k$-th entry of the vector $\mathbf c_{(t)}$. The inference unit can be specified as arbitrary off-the-shelf sequence models such as recurrent neural network (RNN) or self-attention (SA) [@vaswani2017attention]. + +We next introduce a *branching unit* $\Psi(\cdot)$ to parameterize the variational posterior $Q(C|S)$ that aims to obtain $\mathbf c_{(t)}$ given the sequence. At each time step, it takes the sequence representation as input and outputs a probability vector $\mathbf q_t \in [0, 1]^K$, whose $k$-th entry represents the probability of the corresponding context $c_k$, i.e., $$\begin{equation} +\begin{split} + \mathbf q_t = \Psi(\mathbf h^1_t; \Omega),\; \text{where}\; \sum_{k=1}^{K} \mathbf q_t[k] = 1. +\end{split} +\end{equation}$$ We use a $d'$-dimensional *context embedding* $\mathbf w_k\in \mathbb R^{d'}$ to accommodate the information of each context $\mathbf c_k$ by an embedding matrix $\mathbf H_c \in \mathbb R^{K\times {d'}}$, where $\mathbf w_k = \mathbf {c_{k}}^{\top} \mathbf H_c$ and $d' = d (d + 1)$. Then, we split each context embedding into several fix-sized parameters $\mathbf W_k\in \mathbb R^{d\times d}$, $\mathbf a_k\in \mathbb R^{d}$: $$\begin{equation} +\begin{split} + \mathbf W_k = \mathbf w_k[:d^2].reshape(d, d), \; \mathbf a_k = \mathbf w_k [d^2: d (d+1)]. +\end{split} +\end{equation}$$ The attribution score $s_{tk}$ that measures how likely a sequence up to time step $t$ belongs to context $c_k$ can be calculated via $$\begin{equation} + s_{tk} = \langle\mathbf a_k, Tanh(\mathbf W_k \mathbf h_t)\rangle. +\end{equation}$$ Namely, we first project the sequence representation into a new $d$-dimensional space and then use dot product to measure the similarity. The variational distribution $Q(C|S=\mathcal S)$ can be specified as a probability vector $\mathbf q_t$ that is computed by using Softmax function over $s_{tk}$, i.e., $\mathbf q_t = Softmax([s_{tk}]_{k = 1}^{K}; \tau)$, where $\tau$ controls the confidence level. To implement the random sampling procedure $c\sim Q(C|S)$ in Eq. [\[eqn_obj\]](#eqn_obj){reference-type="eqref" reference="eqn_obj"}, one can also apply the categorical reparameterization trick that uses differentiable samples from Gumbel-Softmax [@jang2017categorical] distribution, i.e., $$\begin{equation} + \mathbf q_t[k] = \frac{\exp \left( (s_{tk}+g)/\tau\right)}{\sum_{k=1}^{K}\exp \left((s_{tk}+g)/\tau\right)}, \quad g\sim Gumbel(0,1). +\end{equation}$$ Then, we can revise the right term of Eq. [\[eqn_encoder\]](#eqn_encoder){reference-type="eqref" reference="eqn_encoder"} as $\sum_{k=1}^K \mathbf q_t[k] \Phi_k([\mathbf h_1^1, \mathbf h_2^1, \cdots, \mathbf h_t^1]; \theta_k)$. + +The aforementioned single-layered architecture requires an independent inference unit for each type of context, i.e., $|\mathcal C|=K$, which has several drawbacks: 1) It is impractical to accommodate a large amount of context types due to limited computational resources; 2) Real-world contexts are not entirely isolated, and thus independently parameterizing them may lead to over-fitting and undesired generalization performance given limited training data. To address these limitations, we further devise a hierarchical branching structure for CaseQ framework. Instead of using a one-hot vector, we denote a context as a 0-1 matrix, i.e., $$\begin{equation} + \mathbf c_k = stack([\mathbf c^1_k, \mathbf c^2_k, \cdots, \mathbf c^{D}_k]) \in \{0,1\}^{D\times K}, +\end{equation}$$ where $D$ denotes the number of layers and $K$ denotes the number of parallel inference units in each layer. Each row in $\mathbf c_k$ is a one-hot vector indicating a certain inference unit in a layer, and each type of context corresponds to a certain combination of inference units. The entire network could then be represented as the stack of multiple parallel inference units and branching units: $$\begin{equation} +\begin{split} + [\mathbf h_1^{l+1}, \mathbf h_2^{l+1}, \cdots, \mathbf h_t^{l+1}] &= \sum_{k=1}^{K} \mathbf q^l_t[k] \;\Phi_k^l([\mathbf h_1^l, \mathbf h_2^l, \cdots, \mathbf h_t^l]; \theta_k) , +\end{split} +\end{equation}$$ where $\mathbf q^l_t = \Psi^l(\mathbf h^l_t)$, and $l\in\{1,2,\cdots,D\}$. The final prediction result at each time step can be obtained by ranking the relevance scores of current hidden states and event embeddings: $$\begin{equation} + \hat y_t = \mbox{argmax}_{m\in \{1,\cdots,M\}}(\mathbf H_x \cdot \mathbf h^{D+1}_t)[m]. +\end{equation}$$ The above computation acts as a realization of $P_\theta(Y|S=\mathcal S, C=c)$ and $c\sim Q(C|S=\mathcal S)$, required by the first term in Eq. [\[eqn_obj\]](#eqn_obj){reference-type="eqref" reference="eqn_obj"}. Besides, the KL term in Eq. [\[eqn_obj\]](#eqn_obj){reference-type="eqref" reference="eqn_obj"} also requires density estimation from $Q(C|S=\mathcal S)$. To achieve this, we use the continued Kronecker product of the probability vector $\mathbf q_t^l$ in each layer to represent the final variational posterior: $$\begin{equation} + \mathbf q_t = Flatten\Big( \mathop{\bigotimes}\limits_{l=1}^{D}\mathbf q_t^l \Big)\in \mathbb R^{K^D}, +\end{equation}$$ where $\mathbf q_t^l\in \mathbb{R}^K$ and $\otimes$ denotes Kronecker product. The produced variational posterior is a $K^D$-dimensional vector, where each entry is the probability of a certain path (i.e., a combination of inference units across layers) in the network. This design has some merits. First, we could accommodate $|\mathcal C|=K^D$ types of contexts that grow exponentially w.r.t. model depth $D$. This helps to endow the model with larger capacity for learning diversified contexts. Second, different contexts may associate with each other by sharing the same inference unit in some layers, which plays as an effective *inductive bias* that guides the model to generalize to new environments (see more details and justifications in Appendix [8](#ap_just){reference-type="ref" reference="ap_just"}. + +According to Eq. [\[eqn_obj\]](#eqn_obj){reference-type="eqref" reference="eqn_obj"}, the final loss function can be written as $$\begin{equation} + \sum_{\mathcal S} \sum_{t=1}^{|\mathcal S|-1} \left[ l(y_t, \hat y_t) + \alpha \mathcal D_{KL}\left(\mathbf q_t \| \frac{1}{R} \sum_{j=1}^R \mathbf q(\mathcal S'_j)\right) \right]. +\end{equation}$$ where $\alpha$ is a weight parameter for balance, $\mathbf q(\mathcal S'_j)$ denotes the produced variational posterior distribution with a pseudo sequence $\mathcal S'_j$ as input, and $l(\cdot, \cdot)$ can be arbitrary loss functions (e.g., cross-entropy) depending on specific tasks. diff --git a/2302.06058/main_diagram/main_diagram.drawio b/2302.06058/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..7a0ad18c3cb9cc6071ea14dbd082a79515841ce3 --- /dev/null +++ b/2302.06058/main_diagram/main_diagram.drawio @@ -0,0 +1 @@ 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 \ No newline at end of file diff --git a/2302.06058/paper_text/intro_method.md b/2302.06058/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..c9bb0ce5b9f06f47875f733396764716f5ee5ecd --- /dev/null +++ b/2302.06058/paper_text/intro_method.md @@ -0,0 +1,117 @@ +# Introduction + +The past decade has witnessed thriving deep neural networks (DNNs) in various machine learning applications [@he2016deep; @he2017mask; @girshick2014rich]. In large part, the prosperity is driven by increasing parameters and computations, which however, make DNN models too cumbersome to be deployed on resource-constrained edge devices such as cell phones and Internet-of-Things (IoT) devices. Therefore, the research community is sorely in need of technical renovation to compress the DNNs [@hubara2016binarized; @tan2019efficientnet; @lin2020hrank]. + +
+
+ +
Vanilla N:M Mask
+
+
+ +
Transposable N:M Mask
+
+
Comparison between vanilla N:M mask and transposable N:M mask (2:4 case). The vanilla N:M mask  generates sparse weights with N:M property in rows, leading to forward acceleration but remaining dense backward propagation as the weight transposition operation impairs N:M blocks. The transposable N:M mask  generates sparse weights that have N:M property in both rows and columns, leading to forward & backward acceleration. Both methods consider only one sparse mask.
+
+ +
+ +
Framework of the proposed Bi-direction Masks (Bi-Mask). It separately builds two N:M sparse masks in the forward and backward direction, thus enabling training acceleration in both directions. During backward propagation, Bi-Mask performs an efficient row permutation to make the sparse weights have more eligible N:M weight blocks before generating the backward mask.
+
+ +By removing redundant network weights [@lecun1989optimal; @han2015learning; @he2017channel], network sparsity has emerged as a piece of modern equipment to obtain a lightweight sparse model. Through removing individual weights at arbitrary positions, fine-grained sparsity is demonstrated to reach a high sparse ratio with performance guarantee [@han2015learning; @evci2020rigging]. Unfortunately, the resulting unstructured sparse weights hardly produce acceleration on off-the-shelf hardware. Coarse-grained sparsity is more hardware friendly as it typically removes an entire weight block [@ji2018tetris; @meng2020pruning] or convolution filter [@liu2019metapruning; @lin2020hrank]. In comparison with fine-grained sparsity, the compressed model gains noticeable speedup, yet suffers more performance degradation. Therefore, it is a challenging yet valuable issue to simultaneously retain model performance of DNN models and achieve hardware acceleration. + +Luckily, recent N:M fine-grained sparsity has provided a promising solution. By requiring at most N non-zero elements out of every M contiguous weights, N:M sparsity includes the performance advantage of fine-grained sparsity as well as practical acceleration thanks to the hardware innovation of N:M sparse tensor core [@ronny2020nvidia; @fang2022algorithm]. Nvidia [@nvidia2020a100] has presented the ASP (APEX's Automatic Sparsity) paradigm that achieves 2:4 sparsity within three steps, unfolded as training a dense network, applying 2:4 fine-grained sparsity using magnitude-based pruning [@han2015learning], and re-training the sparse network. Despite the satisfying performance, ASP exhibits drawbacks in its tedious training cost as it contains dense network training and N:M sparse retraining. This largely prohibits the application of N:M sparsity when confronting with scarce training resources. + +The above issue has been partially addressed by directly training an N:M sparse network from scratch [@zhou2021learning]. Yet, the sparse tensor core is only utilized to accelerate the forward multiplication during training. As illustrated in Fig. [\[fig:vanilla\]](#fig:vanilla){reference-type="ref" reference="fig:vanilla"}, the weight transposition operation in the backward impairs N:M blocks and thus fails to support acceleration in gradient calculation. To mitigate this,  [@hubara2021accelerated] proposed N:M transposable mask, where a binary mask that indicates whether to remove weights is required to have N:M property along the rows and columns. Therefore, after performing transposition, it still satisfies the N:M format as shown in Fig. [\[fig:transposable\]](#fig:transposable){reference-type="ref" reference="fig:transposable"}. Unfortunately, the transposable requirement is observed to have more performance degradation, which is presumably caused by less flexibility of the sparsity pattern [@hubara2021accelerated]. In Sec. [3.2](#trans){reference-type="ref" reference="trans"}, we further show severe performance degradation at a higher sparse level such as 1:8 and 1:16. We therefore reflect on this: how can we address the efficiency of N:M sparse training without a compromise on performance? + +In this paper, we attempt to answer the above question by introducing a novel method of Bi-directional Masks (Bi-Mask) that performs surprisingly well without any N:M transposable constraint. Fig. [4](#fig:framework){reference-type="ref" reference="fig:framework"} illustrates framework of our Bi-Mask. In particular, along the forward and backward directions, two separate binary masks are constructed according to the weight magnitude [@han2015learning]. As a contrast, we require the forward mask to follow N:M property in its rows while in columns for the backward mask. By this way, we concurrently enable forward & backward acceleration from the N:M sparse tensor core. Also, the bi-directional masks benefit performance from more flexible sparsity pattern. Nevertheless, they also bring about deficiency of gradient gap since the backward mask modifies the gradient of forward loss. Given this issue, an efficient row permutation is further introduced before enforcing the backward mask. In detail, we first change row order of weight matrix and then pick up the permutation with the most eligible N:M weight blocks from a dozen of candidates. By changing column order of output gradient accordingly, we succeed in retaining the same outputs between with/without row permutation, and at the same time well reducing the gradient gap between uni-directional/bi-directional mask(s). + +[]{#tab:merit_comparison label="tab:merit_comparison"} + +Our simple design of Bi-Mask turns out to achieve remarkable results. Besides forward & backward training acceleration, Bi-Mask well improves the performance of transposable mask (T-Mask) across different N:M patterns, benchmarks, and networks. For example, Bi-Mask achieves 71.5% Top-1 accuracy when training 1:16 sparse ResNet-50 on ImageNet, surpassing T-mask by 5.3%. More surprisingly, our approach achieves comparable or even better results than vanilla N:M methods, where the backward propagation can not be accelerated. For example, our Bi-Mask exceeds Top-1 accuracy of SR-STE [@zhou2021learning] by 0.4% when training 2:4 sparse ResNet-50 on ImageNet. Table [\[tab:merit_comparison\]](#tab:merit_comparison){reference-type="ref" reference="tab:merit_comparison"} provides advantage comparison between different mask methods. + +# Method + +We first introduce some basic knowledge about the N:M fine-grained sparsity. Let $\mathbf{W}\in \mathbb{R}^{I \times J}$ be the parameter matrix from a specific network layer. Considering the input tensor $\mathbf{X}$, the forward propagation represented with form of matrix multiplication can be formulated as: $$\begin{equation} +\label{eq:forward} + \mathbf{Y}= \mathbf{W}* \mathbf{X}, +\end{equation}$$ where $\mathbf{Y}$ is the output tensor and $*$ is the matrix multiplication operation. N:M sparsity forces at most N out of M consecutive weights in the rows of $\mathbf{W}$ to have non-zero values. The sparsity can be achieved via a binary matrix $\mathbf{B}\in \{0,1\}^{I\times J}$ where a block of every M contiguous elements contains at most N as: $$\begin{equation} +\label{eq:contraint_forward} +{\Vert\mathbf{B}_{i, j:j+\text{M}}\Vert}_0 \le \text{N}, +\end{equation}$$ in which $i = 1, 2, 3, ..., I$ and $j =1, \text{M}, 2\text{M}, ..., J$. Then, the sparse forward propagation can be formulated as: $$\begin{equation} +\label{eq:sparse_forward} + \mathbf{Y}= (\mathbf{B}\odot \mathbf{W}) * \mathbf{X}, +\end{equation}$$ where $\odot$ denotes the element-wise multiplication. Since $\mathbf{B}\odot \mathbf{W}$ meets N:M requirement, the matrix multiplication with $\mathbf{X}$ can be efficiently implemented by the N:M sparse tensor core, as illustrated in Fig. [\[fig:vanilla\]](#fig:vanilla){reference-type="ref" reference="fig:vanilla"}. + +N:M sparse training starts from randomly-initialized networks [@zhou2021learning; @zhang2022learning], thus avoiding heavy burden of pre-training a dense model [@nvidia2020a100]. We base our study on the popular SR-STE  [@zhou2021learning] for N:M sparse training, simply illustrated for ease of understanding in the following. During forward propagation, it adapts the binary mask $\mathbf{B}$ at each iteration as: $$\begin{equation} +\label{eq:weight_projection} +\mathbf{B}_{i, j+m} = \left\{ \begin{array}{ll} + 0, \; \textrm{if} \; |\mathbf{W}_{i, j+m}| < \text{Top}(|\mathbf{W}_{i, j:j+\text{M}}|, \text{N}),\\ + 1, \; \text{otherwise}, + \end{array} \right. +\end{equation}$$ where $1 \leq m \leq \text{M}$, $|\cdot|$ represents the absolute function, and $\text{Top}(|\mathbf{W}_{i, j:j+\text{M}}|, \text{N})$ returns the N-$th$ largest value within $|\mathbf{W}_{i, j:j+\text{M}}|$. Therefore, we obtain the forward binary mask according to the weight magnitude in each block. During backward propagation, the gradients of $\mathbf{B}\odot \mathbf{W}$ are directly passed to $\mathbf{W}$ according to the straight-through-estimator (STE) [@bengio2013estimating]. + +The above sparse mask is indeed uni-directional towards forward propagation. By forming N:M blocks in rows of the mask, Eq. ([\[eq:sparse_forward\]](#eq:sparse_forward){reference-type="ref" reference="eq:sparse_forward"}) permits forward acceleration from the N:M sparse tensor core between the weights and inputs. Unfortunately, such a vanilla mask crashes backward acceleration due simply to the transposition operation. To explain, the gradient in the backward propagation is computed as: $$\begin{equation} +\label{eq:backward} + g(\mathbf{X}) = (\mathbf{B}\odot \mathbf{W})^T * g(\mathbf{Y}), +\end{equation}$$ where $g(\cdot)$ denotes the gradient with respect to its input. The above equation requires $(\mathbf{B}\odot \mathbf{W})^T$ to have N:M blocks in rows for accelerating multiplication with $g(\mathbf{Y})$, however, it is in columns on account of the transposition operation. Thus, the backward propagation remains dense and fails to be accelerated, as illustrated in Fig. [\[fig:vanilla\]](#fig:vanilla){reference-type="ref" reference="fig:vanilla"}. + +To address this issue,  [@hubara2021accelerated] presented transposable N:M mask that is required to satisfy row-wise and column-wise N:M blocks such that the transposition also undertakes an important mission of N:M property in rows. Consequently, the binary mask $\mathbf{B}$ is constrained as: $$\begin{equation} +\label{eq:constraint_transposable} +{\Vert\mathbf{B}_{i, j:j+\text{M}}\Vert}_0 \le \text{N}, + \quad {\Vert\mathbf{B}_{k:k+\text{M}, l}\Vert}_0 \le \text{N}, +\end{equation}$$ where $i = 1, 2, 3, ..., I$, $j = 1, \text{M}, 2\text{M}, ..., J$, $k = 1, M, 2M, ..., I$, and $l = 1, 2, 3, ..., J$. Besides,  [@hubara2021accelerated] further introduced a $2$-approximation algorithm to reduce complexity of finding the transposable mask. + +
+
+ +
Flexibility Comparison
+
+
+ +
Performance Comparison
+
+
Comparison between vanilla mask and transposable mask including (a) flexibility measured by mask diversity  and (b) performance of training sparse ResNet-50  on ImageNet .
+
+ +Here we rethink the transposable pursuit for N:M sparse training. Although it enables backward acceleration, the flexibility of sparse networks is greatly restricted, which comes at the cost of performance degradation. We first report the flexibility comparison between vanilla mask and transposible mask under different N:M cases. Fig. [\[fig:mask-diversity\]](#fig:mask-diversity){reference-type="ref" reference="fig:mask-diversity"} measures the flexibility using mask diversity that calculates the number of all possible masks under a given N:M mask case [@hubara2021accelerated]. We can see a drastic flexibility degradation, in particular in cases of a small N or M. As a consensus [@gale2019state; @nvidia2020a100], more restrictions on sparse patterns incur worse performance of sparse networks. For example, unstructured sparsity [@han2015learning] that removes arbitrary weights generally performs much better than structured sparsity [@li2016pruning] that removes entire filter weights. Consequently, severe performance occurs in transposable mask in comparison with the vanilla method, as we experimentally verify in Fig. [\[fig:performance\]](#fig:performance){reference-type="ref" reference="fig:performance"}, notably very poor 1:8 and 1:16. The uni-directional masks, either vanilla or transposable, do not accomplish N:M backward acceleration without a compromise on performance. Therefore, in what follows, we address this issue from the perspective of bi-directional masks. + +In this section, we formally present our Bi-directional N:M masks (Bi-Mask). As its name suggests, our Bi-Mask disentangles the forward & backward weight sparsity by involving two different masks during N:M sparse training. Concretely speaking, in the forward direction, we count on the vanilla N:M mask $\mathbf{B}$ from Eq. ([\[eq:contraint_forward\]](#eq:contraint_forward){reference-type="ref" reference="eq:contraint_forward"}) that calls for N:M in rows to ensure the forward acceleration and results in better flexibility than the transposable N:M mask as we report in Fig. [\[fig:mask-diversity\]](#fig:mask-diversity){reference-type="ref" reference="fig:mask-diversity"}. Very differently, we additionally build another mask $\bar{\mathbf{B}} \in \{0, 1\}^{I \times J}$ in the backward direction with the N:M requirement on its columns as: $$\begin{equation} + {\Vert\bar{\mathbf{B}}_{k:k+\text{M}, l}\Vert}_0 \le \text{N}, +\end{equation}$$ in which $k=1, \text{M}, 2\text{M}, ..., I$, and $l = 1, 2, 3,..., J$. In this fashion, the backward acceleration is supported as well without a compromise on the flexibility of backward mask, and the backward gradient $g(\mathbf{X})$ in Eq. ([\[eq:backward\]](#eq:backward){reference-type="ref" reference="eq:backward"}) is represented by the following approximation: $$\begin{equation} +\label{eq:gradient_calculation} + \bar{g}(\mathbf{X}) = (\bar{\mathbf{B}} \odot \mathbf{W})^T * g(\mathbf{Y}). + \vspace{-0.2cm} +\end{equation}$$ + +Nevertheless, the forward $\mathbf{B}$ requires gradient of $g(\mathbf{X})$ for our Bi-Mask, which yields a gradient gap between practical bi-directional gradient $\bar{g}(\mathbf{X})$ and ideal uni-directional gradient $g(\mathbf{X})$. To solve this issue, we adapt the backward mask $\bar{\mathbf{B}}$ to the magnitudes of masked weights during sparse training as follows: $$\begin{equation} +\label{eq:backward_mask} +%\scriptsize +\bar{\mathbf{B}}_{k+m, l} = \left\{ \begin{array}{ll} + 0, \; \textrm{if} \; |(\mathbf{B}\odot \mathbf{W})_{k+m, l}| \\ \qquad < \text{Top}(|(\mathbf{B}\odot \mathbf{W})_{k:k+\text{M}, l}|, \text{N}),\\ +\mathbf{B}_{k+m, l}, \; \text{otherwise}, + \end{array} \right. +\end{equation}$$ where $k = 1, \text{M}, 2\text{M}, ..., I$, $l = 1, 2, ..., J$, and $1 \le k \le \text{M}$. For a deeper analysis, it is easy to understand that $\mathbf{B}_{k+m, l} = 0$ is a fully not necessary condition of $\mathbf{B}_{k+m, l} = 0$. That is, the event $\mathbf{B}_{k+m, l} = 0$ will produce the event $\mathbf{B}_{k+m, l} = 0$, but is not the only way for $\mathbf{B}_{k+m, l} = 0$ to occur. + +The rationale behind Eq. ([\[eq:backward_mask\]](#eq:backward_mask){reference-type="ref" reference="eq:backward_mask"}) is two-fold: 1) It maximizes the similarity of forward and backward masks by setting $\bar{\mathbf{B}}_{k+m, l} = \mathbf{B}_{k+m, l}$ if the magnitude of $\mathbf{W}_{k+m,l}$ is beyond the top-N largest. 2) Applying our backward mask does not affect the updating of these weights with zero forward masks since $\mathbf{B}_{k+m, l} = 0$ always results in $\bar{\mathbf{B}}_{k+m, l} = 0$. Unfortunately, it is a possibility that $\bar{\mathbf{B}}_{k+m, l} = 0$ does not necessarily result from $\mathbf{B}_{k+m, l} = 0$, in which case gradients of some non-zero masked weights are mistakenly eliminated, incurring performance degradation. + +::: algorithm +Return $\mathbf{W}\odot \mathbf{B}$. +::: + +To decrease this possibility, we continue a row permutation method along the row dimension of $\mathbf{B}\odot \mathbf{W}$. Our major motivations are also two-fold: 1) We can see from Eq. ([\[eq:backward_mask\]](#eq:backward_mask){reference-type="ref" reference="eq:backward_mask"}) that the resulting mask block $\bar{\mathbf{B}}_{k:k+\text{M}, l}$ would exactly match with ${\mathbf{B}}_{k:k+\text{M}, l}$ if $(\mathbf{B}\odot \mathbf{W})_{k:k+\text{M}, l}$ has N:M sparsity, and no gradient gap would occur. 2) Performing row permutation of $\mathbf{B}\odot \mathbf{W}$ improves the number of eligible N:M blocks as illustrated in Fig. [4](#fig:framework){reference-type="ref" reference="fig:framework"}. Importantly, it does not violate the gradient computation. Denoting $P \in \mathbb{N}^{I}$ as a permutation of $\{1,2,3,...,I\}$, the backward gradient $\bar{g}(\mathbf{X})$ in Eq. ([\[eq:gradient_calculation\]](#eq:gradient_calculation){reference-type="ref" reference="eq:gradient_calculation"}) can be equally computed as: $$\begin{equation} +\label{eq:backward_final} + \bar{g}(\mathbf{X}) = \big(\bar{\mathbf{B}} \odot (\mathbf{W}_{P, :})\big)^T * \big(g(\mathbf{Y})_{:, P}\big), +\end{equation}$$ where the backward mask $\bar{\mathbf{B}}$ is computed based on the permutated $(\mathbf{B}\odot \mathbf{W})_{P, :}$ accordingly: $$\begin{equation} +\label{eq:backward_mask_final} +\bar{\mathbf{B}}_{k+m, l} = \left\{ \begin{array}{ll} + 0, \; \textrm{if} \; \Big|\big((\mathbf{B}\odot \mathbf{W})_{P, :}\big)_{k+m, l}\Big| \\ \quad \quad \; < \text{Top}(\Big|\big((\mathbf{B}\odot \mathbf{W})_{P, :}\big)_{k:k+\text{M}, l}\Big|, \text{N}),\\ +(\mathbf{B}_{P,:})_{k+m, l}, \; \text{otherwise}. + \end{array} \right. +% +\end{equation}$$ + +Therefore, we only need to find a permutation $P$ that results in more eligible N:M blocks in each column of $(\mathbf{B}\odot \mathbf{W})_{P,:}$. More N:M blocks decrease the possibility of eliminating gradients of non-zero masked weights. To avoid the cumbersome $I!$ possible permutations at each training iteration, we update a good permutation at a regularly spaced interval of every $\Delta T$ training iterations, and at each interval pick up the one that leads to the most eligible N:M blocks from randomly generating $K$ permutation candidates. + +In Sec. [4.4](#sec:ablation){reference-type="ref" reference="sec:ablation"}, we analyze that the permutation candidate number $K=100$ already returns good performance. Compared with the aforementioned $2$-approximation algorithm for the transposable N:M mask [@hubara2021accelerated], our method brings negligible runtime burden as we experimentally reported in Sec. [4.3](#imagenet){reference-type="ref" reference="imagenet"}. Our algorithm presented in this paper is outlined in Alg. [\[alg:bimask\]](#alg:bimask){reference-type="ref" reference="alg:bimask"}. diff --git a/2304.04205/main_diagram/main_diagram.drawio b/2304.04205/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..2accd8a3e0fcd151bc84261619c13db9932adc25 --- /dev/null +++ b/2304.04205/main_diagram/main_diagram.drawio @@ -0,0 +1 @@ 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 \ No newline at end of file diff --git a/2304.04205/main_diagram/main_diagram.pdf b/2304.04205/main_diagram/main_diagram.pdf new file mode 100644 index 0000000000000000000000000000000000000000..243fd1a5fbc3857edd1fd74ddea06542698a5437 Binary files /dev/null and b/2304.04205/main_diagram/main_diagram.pdf differ diff --git a/2304.04205/paper_text/intro_method.md b/2304.04205/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..c070fb7b86f4b5b7daf7c40fcb98564d95865600 --- /dev/null +++ b/2304.04205/paper_text/intro_method.md @@ -0,0 +1,181 @@ +# Introduction + +Recently, person re-identification (ReID) for pedestrian matching in non-overlapping camera views has experienced fast development. However, ReID is still challenging when people appear both in the daytime and in low-light situations where only infrared cameras can clearly capture their appearances, raising the task of visible-infrared ReID (VI-ReID). Many remarkable works [@sysumm01; @gan; @duallevel; @xmodal; @twostream; @sdl] have been witnessed in the field of VI-ReID. For realistic scenarios, discovering *rich and diverse* modality-shared semantic concepts usually helps to improve the effectiveness of VI-ReID [@nuances; @paenet]. So far, *diverse* modality-shared feature learning remains challenging. + +
+ +
An illustration of our motivation on VI-ReID. It is assumed that body shape information and identity-related modality-shared information (presented in dashed box) are partially overlapped with each other. To make extracted features more diverse, we propose shape-erased feature learning paradigm that decomposes the representation into shape-related feature and shape-erased one. Learning shape-erased feature drives the model to discover richer modality-shared semantic concepts other than body shape.
+
+ +Among the cues for VI-ReID, we can identify pedestrians by their body shapes in many situations, for it contains modality-invariant information and also robust to light changes. Nevertheless, body shape is not the only or a sufficient semantic concept that interprets the identity of a person. It may be hard in some situations to tell the difference only depending on the body shape, but we can still distinguish them by other semantic concepts, such as their belongings, hairstyles or face structures. Inspired by this, we illustrate an information theoretic measure between visible and infrared modality as a Venn diagram on the left of the dashed line in Fig. [1](#fig:1){reference-type="ref" reference="fig:1"}. It is assumed that body shape (presented in red) and identity-related modality-shared information (presented in dashed box) are partially overlapped with each other. Note that *partially* is also due to there exists identity-unrelated information contained in body shape map, , human pose. This partially overlapped assumption indicates that the target information for VI-ReID, which is identity-related and modality-shared, can be divided into two independent components that are related and unrelated to body shape. + +Based on the above observation and assumption, to dig more diverse modality-shared cues for VI-ReID, we expect to erase the body-shape-related semantic concepts in the features to force the VI-ReID model to extract more and other modality-shared features for identification. As illustrated on the right of the dashed line in Fig. [1](#fig:1){reference-type="ref" reference="fig:1"}, the *shape-erased* feature is decorrelated from the *shape-related* feature to simultaneously discover shape-unrelated knowledge, while *shape-related* feature can be explicitly guided by some given body shape prior, which is easy to obtain by existing pre-trained human parsing models [@humanparsing]. In this way, both *shape-related* and *shape-erased* features are explicitly quantified while the discriminative nature of the two features can be independently maintained. + +Specifically, we propose shape-erased feature learning paradigm that introduces orthogonality into representation to satisfy a relaxation of independent constraint. The representation is then decomposed into two sub-representations lying in two orthogonal subspaces for *shape-related* and *shape-erased* feature learning, respectively. By learning and covering most discriminative body shape feature in one subspace, the *shape-erased* feature is forced to discover other modality-shared discriminative semantic concepts in the the other subspace as *shape-related* feature is constrained in its orthogonal complement. Under the above assumptions, we formulate this shape-erased feature learning paradigm from a mutual information perspective, and demonstrate that jointly learning *shape-erased* and *shape-related* objectives achieves a conditional mutual information maximization between *shape-erased* feature and identity discarding body shape information, thus enhancing the diversity of the learned representation explicitly. We finally design a Shape-Guided dIverse fEature Learning (SGIEL) framework that jointly optimizes *shape-related* and *shape-erased* objectives to learn modality-shared and discriminative integrated representation. The contributions of our work are summarized as follows: + +- We propose a shape-erased feature learning paradigm for VI-ReID that decorrelates *shape-erased* feature from *shape-related* one by orthogonal decomposition. *Shape-related* feature in one subspace is guided by body shape prior while *shape-erased* feature is constrained in its orthogonal complement to discover more and other modality-shared discriminative semantic concepts, thus enhancing the diversity of the learned representation explicitly. + +- Based on the proposed shape-erased feature learning paradigm, we design a Shape-Guided dIverse fEature Learning framework that jointly optimizes *shape-related* and *shape-erased* objectives to learn modality-shared and discriminative integrated representation. + +- Extensive experiments on SYSU-MM01, RegDB, and HITSZ-VCM datasets demonstrate the effectiveness of our method. + +# Method + +Consider random variables $X^{(i)}$ and $Y$ representing data and label of VI-ReID, where $i\!=\!1$ for visible modality and $i\!=\!2$ for infrared modality. The observed values of $X^{(i)}$ and $Y$ are used to build a dataset $D\! =\!\{D^{(i)}\}_{i=1}^2$, where $D^{(i)}\!=\!\{x_j^{(i)},y_j\}_{j=1}^{N_i}$. Samples of each modality are collected from the same group of $C$ persons, but the number of each identity's samples for each modality may arbitrary. Let $f$ and $g$ denote image encoder and classifier, the goal of VI-ReID is to learn an $f$ to extract representation $z^{(i)}=f(x^{(i)}) \in \mathbb{R}^n$ invariant to different modalities and different camera views. + +We borrowed pre-trained Self-Correction Human Parsing (SCHP) model proposed in [@humanparsing] to segment body shape from background. Given a pixel of an image, we directly summed the probabilities of being a part of the head, torso, or limbs, predicted by SCHP, to create the body-shape map. Specifically, for each sample $x^{(i)}$ from dataset $D$, either visible or infrared, we used SCHP to produce its paired body shape map $x^{(s)}$ with the same image size and label, , it is a one-to-one mapping between $D$ and its corresponding body shape data. Let $f_{s}$ and $g_{s}$ denote body shape map encoder and classifier, the latent representation of $x^{(s)}$ is $z^{(s)} =f_{s}(x^{(s)}) \in \mathbb{R}^m,$ $m\!<\!n$. + +In this section, we first explain the key independent assumption for explicitly quantifying *shape-related* and *shape-erased* features, and a relaxation to approximate it. Based on this relaxed independent constraint, we introduce the proposed Shape-Erased Feature Learning. + +
+ +
Shape-Guided Diverse Feature Learning. We utilize one shared backbone for visible, infrared images, and their body shape maps, while only BN layers are view-specific; “cat” refers to concatenating z(i) along batch dimension, where i = 1(2) for visible(infrared); In shape-erased feature learning paradigm, by regularizing P to be semi-orthogonal (Eq. [eq:ortho]), we decompose z(i) into shape-related zsr(i) and shape-erased zse(i). zsr(i) is learned to imitate and cover discriminative body shape features in subspace P (Eq. [eq:mse] and [eq:srce]), while zse(i) is decorrelated to mine other modality-shared discriminative features in subspace P⟂ (Eq. [eq:sc]).
+
+ +We first formulate the main design of our shape-erased feature learning paradigm as a graphical model illustrated in Fig. [2](#fig:graph){reference-type="ref" reference="fig:graph"}. It is assumed that modality-shared *shape-related* feature, $Z_{sr}^{(i)}$, and modality-shared *shape-erased* feature, $Z_{se}^{(i)}$, are independent from each other, and derived from an integrated representation $Z^{(i)}$ extracted in $X^{(i)}$, , $Z^{(i)}\rightarrow Z_{sr}^{(i)},\ Z^{(i)}\rightarrow Z_{se}^{(i)}$. The independence between the two components, $Z_{sr}^{(i)} \perp \!\!\! \perp Z_{se}^{(i)}$, is necessary for learning any two features simultaneously without affecting each other. We formulate this independence as the following Eq. [\[eq:partition\]](#eq:partition){reference-type="eqref" reference="eq:partition"}, $$\begin{equation} + \begin{split} + I(Z_{sr}^{(i)};Z_{se}^{(i)})&=0, + % I(Z_{sr}^{(i)};X^{(i)}) + I(Z_{se}^{(i)};X^{(i)}) &= I(Z^{(i)};X^{(i)}). + \end{split} +\label{eq:partition} +\end{equation}$$ where $I(\cdot;\cdot)$ denotes mutual information. As the mutual information estimation is complex and time-consuming, we relax the independence as an orthogonal constraint, and perform orthogonal decomposition to achieve the relaxed version of Eq. [\[eq:partition\]](#eq:partition){reference-type="eqref" reference="eq:partition"} as: $$\begin{equation} + \begin{split} + z_{sr}^{(i)}&= P^Tz^{(i)}, \\ + z_{se}^{(i)}&=(I_n-PP^T) z^{(i)}, + \end{split} +\label{eq:orthodecompose} +\end{equation}$$ where $P \in \mathbb{R}^{n\times m} (m`{=html}1x & CM-NAS [@cm-nas] & CVPR'21 & 61.99 & 92.87 & 97.25 & 60.02 & 67.01 & 97.02 & 99.32 & 72.95\ +$\thickapprox$``{=html}1x & CAJL [@caj] & ICCV'21 & 69.88 & 95.71 & 98.46 & 66.89 & 76.30 & 97.90 & 99.50 & 80.40\ +$\thickapprox$``{=html}1x & MPANet [@nuances] & CVPR'21 & 70.58 & 96.21 & 98.8 & 68.24 & 76.64 & 98.21 & 99.57 & 80.95\ +$\thickapprox$``{=html}1x & MMN [@mmn] & ACMMM'21 & 70.60 & 96.20 & 99.00 & 66.90 & 76.20 & 97.20 & 99.30 & 79.60\ +$\thickapprox$``{=html}1.75x & MTL [@crossmodalmultitask] & PR'22 & 67.25 & 95.38 & 98.46 & 64.29 & 69.58 & 96.66 & 99.03 & 74.37\ +$\thickapprox$``{=html}1.25x & PAENet [@paenet] & ACMMM'22 & 74.22 & **99.03** & **99.97** & **73.90** & 78.04 & **99.58** & **100.00** & **83.54**\ +$\thickapprox$``{=html}2x & MSCLNet [@msclnet] & ECCV'22 & [76.99]{.underline} & [97.93]{.underline} & [99.18]{.underline} & 71.64 & [78.49]{.underline} & [99.32]{.underline} & [99.91]{.underline} & 81.17\ +$\thickapprox$``{=html}1x & Ours & - & 75.18 & 96.87 & 99.13 & 70.12 & 78.40 & 97.46 & 98.91 & 81.20\ +$\thickapprox$``{=html}2x & Ours (C) & - & **77.12** & 97.03 & 99.08 & [72.33]{.underline} & **82.07** & 97.42 & 98.87 & [82.95]{.underline}\ +::: + +[]{#tab:sysumm01 label="tab:sysumm01"} +:::: + +:::: table* +::: tabular +c\|c\|c\|cccc\|cccc \*Strategy & \*Method & \*Venue & &\ +& & & Rank-1 & Rank-5 & Rank-10 & mAP & Rank-1 & Rank-5 & Rank-10 & mAP\ +Video & MITML [@vcm] & CVPR'22 & [63.74]{.underline} & [76.88]{.underline} & [81.72]{.underline} & 45.31 & 64.54 & [78.96]{.underline} & 82.98 & 47.69\ +\*Image & LbA [@lba] & ICCV'21 & 46.38 & 65.29 & 72.23 & 30.69 & 49.30 & 69.27 & 75.90 & 32.38\ +& MPANet [@nuances] & CVPR'21 & 46.51 & 63.07 & 70.51 & 35.26 & 50.32 & 67.31 & 73.56 & 37.80\ +& DDAG [@ddag] & ECCV'20 & 54.62 & 69.79 & 76.05 & 39.26 & 59.03 & 74.64 & 79.53 & 41.50\ +& VSD [@mireid] & CVPR'21 & 54.53 & 70.01 & 76.28 & 41.18 & 57.52 & 73.66 & 79.38 & 43.45\ +& CAJL [@caj] & ICCV'21 & 56.59 & 73.49 & 79.52 & 41.49 & 60.13 & 74.62 & 79.86 & 42.81\ +& Baseline & - & 62.02 & 75.35 & 81.35 & [47.05]{.underline} & [64.90]{.underline} & 78.64 & [83.68]{.underline} & [48.21]{.underline}\ +& Ours & - & **67.65**& **80.32**& **84.73**& **52.30**& **70.23**& **82.19**& **86.11**& **52.54**\ +::: + +[]{#tab:vcm label="tab:vcm"} +:::: + +::: tabular +c\|cc\|cc \*Method & &\ +& Rank-1 & mAP & Rank-1 & mAP\ +CM-NAS [@cm-nas] & 82.57 & 78.31 & 84.54 & 80.32\ +CAJL [@caj] & 84.75 & 77.82 & 85.03 & 79.14\ +MPANet [@nuances] & 82.8 & 80.7 & 83.7 & 80.9\ +MMN [@mmn] & 87.5 & 80.5 & 91.6 & 84.1\ +MTL [@crossmodalmultitask] & 88.34 & 84.06 & 89.91 & 85.64\ +PAENet [@paenet]& **95.35** & **89.98** & **97.57** & **91.41**\ +MSCLNet[@msclnet] & 83.86 & 78.31 & 84.17 & 80.09\ +Ours &[91.07]{.underline} &[85.23]{.underline}& [92.18]{.underline} &[86.59]{.underline}\ +::: + +[]{#tab:regdb label="tab:regdb"} diff --git a/2306.01150/main_diagram/main_diagram.drawio b/2306.01150/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..affa968b7db5ffea5b52255bb7b941c6f6a88adf --- /dev/null +++ b/2306.01150/main_diagram/main_diagram.drawio @@ -0,0 +1 @@ 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 \ No newline at end of file diff --git a/2306.01150/main_diagram/main_diagram.pdf b/2306.01150/main_diagram/main_diagram.pdf new file mode 100644 index 0000000000000000000000000000000000000000..55a52132869914f426a4da862311fd6451e84aa2 Binary files /dev/null and b/2306.01150/main_diagram/main_diagram.pdf differ diff --git a/2306.01150/paper_text/intro_method.md b/2306.01150/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..fa4c13f0c3d4fae2768459d4fe527d4dceaa3d60 --- /dev/null +++ b/2306.01150/paper_text/intro_method.md @@ -0,0 +1,104 @@ +# Introduction + +:::: table* +::: center ++:-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| **RQ1:** Which parts of task definitions are important when performing zero-shot instruction learning? - For classification tasks, label-related information is crucial, as it helps the model identify the output space and identify each label's meaning when generalizing. - Additional details or constraints besides primary mentions of input and output, in general, do not improve model performance. As model size increases, additional details become important. - Task definitions can be extensively compressed with no performance degradation, particularly for generation tasks. | +| | +| **RQ2:** Is natural language the most efficient format to communicate task instructions to models? - Framing instructions as a structured input/action/output triplet is potentially a more efficient and effective way of creating task definitions. - In fact, using only basic metadata and the label space (without label definitions) in a structured format, we achieve similar, or even better performance as with full definitions. | +| | +| **RQ3:** How can we improve models' understanding of definitions as well as model performance? - Adding a meta-tuning stage for adapting models to the writing styles of definitions improves the performance. | ++--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +::: +:::: + +Large language models or LLMs [@devlin2019bert; @2020t5; @Brown2020LanguageMA] demonstrate the ability to perform zero-shot cross-task generalization through learning from instructions of tasks [@Sanh2022MultitaskPT; @Wei2022FinetunedLM; @Mishra2022CrossTaskGV; @Wang2022BenchmarkingGV; @ouyang2022training; @openai_chatgpt]. By fine-tuning an LLM with *task definitions* and a few *demonstration examples* on upstream training tasks, the model acquires the power to perform new tasks with unseen definitions and example. This is known as *instruction learning*. + +However, a natural question is: to what extent does the zero-shot generalization ability derive from the model's understanding of task definitions? Recent work in prompt-based learning suggests models might not interpret even short prompts as people expect [@webson-pavlick-2022-prompt; @shin2020autoprompt; @deng2022rlprompt; @Prasad2022GrIPS]. Task definitions are special prompts that are usually long and encode rich information. We imagine models' understanding of definitions also departs from human expectation. To investigate this question, we conduct a systematic analysis using both human annotation and computational approaches. Our study is based on the English portion of the large-scale SUPER-NATURALINSTRUCTION (NIv2) dataset [@Wang2022BenchmarkingGV], which comprises 757 training tasks and 119 unseen test tasks. + +First, we explore which type of information in task definitions is necessary for maintaining model performance. We define eight categories of content and provide a fine-grained annotation for all the sentences in task definitions. Then, we retrain the model with every occurrence of each category in NIv2 ablated out, and measure the model performance on the validation set with the same ablation. We observe variable contributions to model performance across content types. For example, input descriptions are in general not helpful to generalization performance, i.e., removing them causes little to no degradation of performance. However, larger models tend to leverage them more. On the other hand, the label information is of great importance. Providing natural-language Label Definitions helps specify the task-specific meaning of common verbalizers while providing the label verbalizer only helps in generalizing to a new label space. We also find that we can achieve similar or even better performance compared to full definitions by only providing the models with a label space along with very basic task metadata, e.g., category, domain, reasoning type, etc. This suggests that costly human generation of task definitions may not always be more helpful than available basic metadata about the task. + +Second, motivated by @feng2018pathologies, to understand what is necessary for models to perform well, we propose **S**yntax-guided **T**ask **D**efinition **C**ompression (STDC), an automatic approach to removing content in task definitions that is not helpful for model performance. STDC queries the model for predictions on inputs and only requires black-box access. We can remove around 60% of tokens while achieving \~3 points of performance improvement of T5-XL on a held-out set. This implies that instead of understanding the whole definition of the task, the models are relying on particular text while ignoring the rest. Along with similar observations as the ablation study above, STDC reveals new patterns of how models understand definitions. For example, models usually do not need to see the whole label space, but might infer the rest with a partial label space. + +Given our observations, we conclude that current instruction learning models rely on partial information in definitions. We imagine the lack of consistency in the creation process of task definitions might hinder the model from attending to all key information in definitions. Thus, we propose two complementary strategies to overcome this. The first strategy is to replace the full definition with a JSON-like formatted triplet of input, action, and output. A JSON-like triplet simplifies the creation of task definitions by asking authors of the definition to fill in blanks in templates instead of writing from scratch, and the common structure increases consistency between authors. The second strategy is to perform meta-tuning before instruction learning to adapt LLMs to any predefined styles of task definitions. We achieve 4.2, 4.0, and 2.1 Rouge-L improvements on BART-Large, T5-Large, and T5-XL, respectively, combining these two strategies. We summarize our key findings in Table [\[table:summaryoffindings\]](#table:summaryoffindings){reference-type="ref" reference="table:summaryoffindings"}. [^3] + +![Annotations of three examples that cover the eight categories of content in task definitions. ](annotationexample.pdf){#fig:annotation} + +In this section, we introduce the formulation of instruction learning, as well as the models and benchmarks used in our study. Further details are presented in Appendix [10](#appendix:details){reference-type="ref" reference="appendix:details"}. + +Instruction learning aims to train a language model so that it understands natural language task instructions and is able to generalize to a new task by solely reading new instructions. A task instruction may include several elements. In this paper, we follow @Wang2022BenchmarkingGV and adopt instructions with 1) a *task definition*: a high-level description of the input and output of the task; and 2) *demonstration examples*: some input-output examples for the task. Note that other content such as *things to avoid* and *negative examples* may also be included but have been shown to be less effective [@Mishra2022CrossTaskGV]. + +A task instruction is generally pre-pended to an input and passed to the LLM. The LLM is first fine-tuned on several upstream training tasks and then asked to conduct inference on an unseen test task, given only the task instruction. + +We adopt the English portion of NIv2 [@Wang2022BenchmarkingGV], which contains 757 training tasks and 119 unseen test tasks. The test tasks fall into 12 categories, including textual entailment, data-to-text generation, etc. However, we also consider a more coarse split of test tasks into *classification* and *generation* tasks, based on whether the output space is fixed or not. For each task, we select 100 examples for either fine-tuning or testing and report performance of Rouge-L [@lin-2004-rouge], following @Wang2022BenchmarkingGV. We use the task definition and two demonstration examples as the instruction. The original paper does not provide an official validation split, which we prepare by putting aside 76 training tasks. We fix the validation set for all experiments to ensure no data leakage. Note that for later experiments, results for Section [3](#section:ablation){reference-type="ref" reference="section:ablation"} and Section [4](#compressionsection){reference-type="ref" reference="compressionsection"} are reported on the validation split which we hold out ourselves while results for Section [5](#improvesection){reference-type="ref" reference="improvesection"} are on the official test set. + +We experiment with the T5-Large and T5-XL models [@2020t5] since the family of T5 sequence-to-sequence models has been shown by @Wang2022BenchmarkingGV to achieve superior performance after fine-tuning compared to frozen models like GPT-3 [@Brown2020LanguageMA] or InstructGPT [@ouyang2022training] on NIv2 benchmark[^4]. We also consider BART-Large [@lewis2020bart] in the experiments. **All results are reported as average performance over three random seeds**. + +# Method + +We present the pseudo-code for the compression algorithm. + +:::: algorithm +**Input:** A model $f$. a set of examples for a specific task $S$: $\mathcal{D}_{S}$. The full task definition: $X_{full} = \{w_1, w_2, ..., w_n\}$. The performance of $f$ on $\mathcal{D}_{S}$ with $x_{full}$: $f\left(\mathcal{D}_{S} | X_{full} \right)$. Constituency tree for the task definition: $\mathcal{T}$.\ +**Output:** Compressed definition $X_{compressed}$. + +::: algorithmic +Initialization: traverse the parse tree $\mathcal{T}$. Find the tree depth $Dep(\mathcal{T})$. The set of nodes $N_i$ at each layer i = 1,2, $\cdots$, $Dep(\mathcal{T})$. $X_{compressed}$ = $X_{full}$ Remove $n_i$ and compute the new performance of $f$ with $X_{full} \textbackslash n_i$: $f\left(\mathcal{D}_{S} | X_{full} \textbackslash n_i \right)$ Remove $n_i$ and its subtree. $X_{compressed}$ = $X_{compressed} \textbackslash n_i$ $X_{compressed}$ +::: + +[]{#Algo label="Algo"} +:::: + +We present examples of the input/action/output triplets as task definitions in Table [\[table:examplesoftriplet\]](#table:examplesoftriplet){reference-type="ref" reference="table:examplesoftriplet"}. + +:::: center +::: {#table:hyperil} + Hyperparameter Range + ---------------- ------------------------------------------------------------ + Learning rate $1 \times 10^{-5}$, $5 \times 10^{-5}$, $1 \times 10^{-4}$ + Batch size 16, 32, 64, 128 + \# Epochs 2, 3, 6 + + : Hyper-parameter choices +::: +:::: + +We find that we need slightly different hyperparameters for instruction learning after meta tuning. We conduct grid search of learning rate, batch size, and the number of epochs on the validation set. The list of candidates are shown in Table [4](#table:hyperil){reference-type="ref" reference="table:hyperil"}. The learning rate is scheduled with constant learning rate with warm up steps 300 for all models. + +:::: table* +::: center ++:--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| **TASK1** task1292_yelp_review_full_text_categorization You are given a review about a place. You need to provide a rating from \"1 star\" to \"5 stars\" for this place. **Input**: \[\"a review about a place\"\] **Action**: \[\"provide a rating from \" 1 star \" to \" 5 stars \" for this place\"\] **Output**: \[\"a rating from \" 1 star \" to \" 5 stars \"\"\] | +| | +| **TASK2** task321_stereoset_classification_religion Given two sentences regarding the stereotypes of religions, determine whether the second sentence is either a stereotype, anti-stereotype, or is unrelated with respect to the first sentence and the target. A stereotype is a widely held but oversimplified notion of a group of people. An anti-stereotype is an idea that goes against the widely held but oversimplified notion of a group of people. Sentences and targets are unrelated if there is no relevance between them. **Input**: \[\"two sentences regarding the stereotypes of religions\"\] **Action**: \[\"determine whether the second sentence is either a stereotype , anti-stereotype , or is unrelated with respect to the first sentence and the target\"\] **Output**: \[\"stereotype, anti-stereotype\", \"A stereotype is a widely held but oversimplified notion of a group of people\", \"An anti-stereotype is an idea that goes against the widely held but oversimplified notion of a group of people\"\] **TASK3** task628_xlwic_word_with_different_meaning_sentence_generation In this task, you are given a word, followed by a sentence. You should respond with a valid sentence which contains the word with the same meaning as in the given sentence. For example, if the given sentence refers to a 'fly' as the insect, you should not respond with a sentence which uses 'fly' as the verb. You may use the word in a different tense than is given. For example, you may use the word 'ended' in the output where the given input word is 'end'. **Input**: \[\"a word, followed by a sentence\"\] **Action**: \[\"respond with a valid sentence which contains the word with the same meaning as in the given sentence\"\] **Output**: \[\"a valid sentence\"\] | +| | +| **TASK4** task405_narrativeqa_question_generation You will be given a summary of a story. You need to create a question that can be answered from the story. You can create a question about characters, events, facts and beliefs, etc. Your question should be specific, try not to use pronouns instead of full names. As the stories are sometimes movie plots, they will contain actor names in parentheses. You should not use those names. Only use character names. Try to ask a question about all parts of the plot, not just the beginning. **Input**: \[\"a summary of a story\"\] **Action**: \[\"create a question that can be answered from the story\"\] **Output**: \[\"a question\"\] | +| | +| **TASK5** task1202_atomic_classification_xneed In this task, you are given two phrases: Head and Tail, separated with \. The Head and the Tail events are short phrases possibly involving participants. The names of specific people have been replaced by generic words (e.g., PersonX, PersonY, PersonZ). PersonX is always the subject of the event. You have to determine whether it is plausible for the Head to desire the Tail or not. In this task, desire means desires of sentient entities. For example, doctors likely desire to cure a patient. Classify your answers into \"Yes\" and \"No\". The phrase may also contain a placeholder that can be an object, a person, and/or an action. **Input**: \[\"two phrases : Head and Tail , separated with \< sep \>\"\] **Action**: \[\"determine whether it is plausible for the Head to desire the Tail or not\"\] **Output**: \[\"Yes, No\"\] | +| | +| **TASK6** task1580_eqasc-perturbed_question_generation Given a statement, generate a question such that the answer is contained in that statement. **Input**: \[\"a statement\"\] **Action**: \[\"generate a question such that the answer is contained in that statement\"\] **Output**: \[\"a question\"\] | +| | +| **TASK7** task383_matres_classification You will be given a context and a verb separated with a newline character, and you have to answer if the given verb is a negation or not. A verb is a negation if it is not going to exist, not happen, or has no effect. The output should be Ÿesïf the verb is a negation and N̈oötherwise. **Input**: \[\"a context and a verb separated with a newline character\"\] **Action**: \[\"answer if the given verb is a negation or not\"\] **Output**: \[\"Yes, No\", \"\" Yes \" if the verb is a negation and \" No \" otherwise\"\] | ++---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +::: +:::: + +[^1]: $^*$Work done when Fan Yin was an intern at Salesforce. + +[^2]: $^\diamondsuit$Jesse and Philippe contributed equally; order is random. + +[^3]: Code will be released at\ + . + +[^4]: At the time the paper is finished. See Section [6](#related work){reference-type="ref" reference="related work"} and Section [7](#discussion){reference-type="ref" reference="discussion"} for updated discussions. + +[^5]: With https://github.com/yzhangcs/parser + +[^6]: https://instructions.apps.allenai.org/ + +[^7]: https://huggingface.co/models?sort=downloads\ + &search=google%2Ft5 + +[^8]: https://github.com/yizhongw/Tk-Instruct + +[^9]: https://github.com/microsoft/DeepSpeed diff --git a/2308.02000/main_diagram/main_diagram.drawio b/2308.02000/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..d8421598ba2888464c106f8de387de79fd07505b --- /dev/null +++ b/2308.02000/main_diagram/main_diagram.drawio @@ -0,0 +1,253 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/2308.02000/main_diagram/main_diagram.pdf b/2308.02000/main_diagram/main_diagram.pdf new file mode 100644 index 0000000000000000000000000000000000000000..e238a713aed932881955164229914db401b9a784 Binary files /dev/null and b/2308.02000/main_diagram/main_diagram.pdf differ diff --git a/2308.02000/paper_text/intro_method.md b/2308.02000/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..4af29bd9af48147ea879d183b96b94b369163a2e --- /dev/null +++ b/2308.02000/paper_text/intro_method.md @@ -0,0 +1,60 @@ +# Introduction + +In recent years, there has been a desire to incorporate the interpretability, compositionality, and logistics of symbolic systems into neural networks (NNs). Existing methods combine them in an ad hoc manner, [@NLM; @SATNET] converts symbolic programs into differentiable forms, [@NS_Infer; @NS_chem] introduce symbolic modules to assist NNs, [@visprog] use neural and symbolic modules as building blocks for visual programs. These methods do not truly bring symbolic power to NNs, but simply allow them to work synergically. The essential disparity lies at a low level. NNs use distributed representations, while symbolic systems use symbolic representations. This motivates us to explore ways to bridge neural and symbolic representations, thus combining the two types of intelligence from the ground up. + +Cognitive science studies [@SE] have suggested that symbolic representation in the human brain does not appear out of nowhere; rather, there is a gradual transition from neural perception to *preliminary symbols* and eventually to symbolic languages over the course of human evolution, as people observe and interact with their environment. The transitional representation, preliminary symbols, is essential in connecting neural and symbolic representation. Taking this concept into account, we attempt to replicate the process of transitional representation arising from neural representation, through unsupervised learning on visual observations in this study, as an exploration of the potential path of unifying neural and symbolic thinking at the representation level. + +Representation explains how the input is made up of reusable components [@Bengio_reprl]. Taking visual observations as an example, distributed representation is explained by vectors, such as principal components, that illustrate the high-dimensional statistical features, and symbolic representation uses structural methods, such as logic sentences, that explain the visual parts and their connections. A transitional representation should be in between that (1) contains high-dimensional details of the input and (2) implies structural information about the semantics of the input. + +
+ +
Decomposing samples from three datasets into visual parts marked with different colors, shallowness represents confidence. Odd columns are input, even columns are decompositions.
+
+ +We propose a novel **Transitional Dictionary Learning (TDL)** framework that implicitly learns symbolic knowledge, such as visual parts and relations, by reconstructing the input as a combination of parts with implicit relations. With a simple fine-tuning that aligns the output with human preference through reinforcement learning and a heuristic reward, the model can give a human-interpretable decomposition of the input; examples are shown in Figure [1](#fig:samples){reference-type="ref" reference="fig:samples"}. TDL uses an Expectation Maximization (EM) algorithm to iteratively update dictionaries that store hidden representations of symbolic knowledge, through an online prototype clustering on the visual parts decomposed from the inputs by a novel game-theoretic diffusion model using the current dictionaries. + +We suggest two metrics to evaluate the learned representation. Clustering Information Gain, which assesses if the learned dictionary is parsimonious and representative, and a heuristic shape score that assesses if the decompositions are in line with human intuition. We conduct unsupervised learning experiments on three abstract compositional visual object datasets, which require the model to utilize the compositionality of data instead of simply visual features, and three tasks on symbol grounding to predefined classes of parts and relations, and transfer learning to unseen classes. The results show huge improvements compared to the state-of-the-art part segmentation baselines, which struggle to process abstract objects that lack distinct visual features. We also conduct human evaluations; the results demonstrate significantly improved interpretability of the proposed method and the proposed metrics are consistent with human assessments. Our contributions are concluded as follows. + +- We propose the unsupervised Transitional Dictionary Learning to learn symbolic features in representations with a novel game-theoretic diffusion model and online prototype clustering. + +- We introduce two metrics, the clustering information gain and the heuristic shape score, to evaluate the learned representation and give evaluation results agreed to human judgment. + +- We perform experiments, compare our method with state-of-the-art unsupervised part segmentation models, and conduct human evaluations for all models and proposed metrics. + +Our code and data are available at . + +# Method + +
+ +
Overview of our architecture. The decomposition process takes K steps to iteratively refine the generated visual parts. NP models generate in parallel, each generates one part and communicates through “Broadcast”. The mapped representations will be stored in a memory bank for clustering.
+
+ +To implement the TDL, we need an encoder $f: x\rightarrow R$ that decomposes the input $x$ into $R=\{r_i\}_{i=1}^{N_P}$ where each $r_i$ decoded by a decoder $g: r_i\rightarrow m_i$ into a visual part $x_i=m_i x$, $m_i\in [0,1]^{H\times W\times C}$ is a mask. Inspired by [@zeroc], we adopt a U-Net-based diffusion model [@SDE] to iteratively refine the generated mask, the encoder downsamples the $x$ to the embedding $r_i$ which upsampled by the decoder as $m_i$. The model iterates $K$ steps. $N_P$ copies of the model sharing the same parameters, generate $N_P$ masks in parallel with each produces one mask. At each step, for model $i$, the mask $m_i(t)$ generated in the previous step (a random mask for the first step) and other feature maps, such as other models' output, are inputted, to produce an updated mask $m_i(t+1)$. After $K$ steps, the model outputs $N_P$ visual parts and the corresponding representation $R$. + +To generate multiple meaningful visual parts at the same time, we propose a game-theoretic method in Section [4.1](#ssec:gt){reference-type="ref" reference="ssec:gt"} inspired by [@EigenGame] who model PCA as a competitive game of principal components. We regard the decomposition process as a cooperative game of visual parts that converge in $K$ steps to cooperatively reconstruct the input while competing with each other to avoid repetition and so on. Each part is adjusted by a "player", one of the $N_P$ copies of the model. With the visual parts generated, we use prototype clustering to implement the EM algorithm in Section [4.2](#ssec:prototype clustering){reference-type="ref" reference="ssec:prototype clustering"}. We also introduce a shape score to measure model performance and serve as a reward to tune the unsupervised learned model with reinforcement learning in Section [4.3](#ssec:rl_sp){reference-type="ref" reference="ssec:rl_sp"}. Figure [2](#fig:model overview){reference-type="ref" reference="fig:model overview"} shows an overview of our method. + +A player model adjusts the generated visual parts to maximize the utility modeled by a GT loss $$\begin{align} +\label{eq:gt} + L_{GT}=L_{Rec}+\alpha_1 L_{overlap} + \alpha_2 L_{resources} + \alpha_3 L_{norm} +\end{align}$$ evaluates the equilibrium state composed of $N_P$ generated parts $\overrightarrow{x}=(x_1,x_2,...x_{N_P})$, in detail: + +**Reconstruction Error**. $L_{Rec}$ evaluates the reconstructed input $\tilde{x}=\sum_i^{N_P} x_i$ using a combination of focal loss [@focal] and dice loss [@dice]. + +**Overlapping Penalty.** $L_{overlap}=\sum_{H,W,C} \max(0,\tilde{x}-x)$ introduces a competition between players that avoids overlap between parts by penalizing redundant parts. + +**Resources Penalty.** $L_{resources}=\sum_i^{N_P} max(0,q_R-|m_i|)$ where $q_R\in\mathbb{R}$ is the quota of a player, prevents one player from reconstructing everything while others output empty. The quota restricts one player from having enough resources to output the entire input, thus requiring cooperation. + +**L2 Norm.** $L_{norm}=\sum_i^{N_P}||\bar{m_i}||_2^2$ where $\bar{m_i}$ is the unactivated mask before input into the $Sigmoid$ function, which can simplify the search space of the model to accelerate convergence. + +See Appendix [12](#apdx:gt_loss){reference-type="ref" reference="apdx:gt_loss"} for further details. We follow SMLD [@SDE] to train a scoring network $\nabla_{m_i(t)} L_{GT}=f_S(s_i(t);\theta,\phi)$, where $\phi=\{\phi^{i}\}_{i=1}^{N_A}$ are prototype dictionaries from 1 to $N_A$ arities that will be discussed in Section [4.2](#ssec:prototype clustering){reference-type="ref" reference="ssec:prototype clustering"}, to approximate the gradient of optimal move for each player $i$ that maximizes utility $-L_{GT}$. The input state $s_i(t)=(e_i(t),\bar{x}_i(t))$ includes the feature map $\bar{x}_i(t)=concat(x;m_i(t);\sum_{k\neq i} m_k(t),...)$ that contains the input, the current mask, and the moves of other players (i.e., "broadcast"), and an embedding $e_i(t)=e_i^{time}(t)+e_i^{pid}+e_i^{pred}(t)$ covers a time step $e^{time}_i(t)\in\mathbb{R}^{d_{emb}}$ and player index $e_i^{pid}\in\mathbb{R}^{d_{emb}}$ from learned embedding tables, and a predicate embedding $e^{pred}_i(t)=\sum_{\sigma\in \phi^1} P(\sigma|x_i(t))\sigma$ computed for every step. + +The move sampled by Langevin dynamics $m_i(t+1)=m_i(t)+\epsilon\nabla_{m_i(t)} L_{GT}+\sqrt{2\epsilon}z(t), z(t)\sim N(0,I), t=0,1,...,K$ where $\epsilon$ is the step size. A loss term $L_{SMLD}$ from the SMLD paper can be used to minimize $E[||\nabla_{m_i(t)} L_{GT}-f_S(s_i(t);\theta)||^2]$ to train the scoring network to give good approximations. We apply this term as a regularization to form **Decomposition Loss** $L_{Decomposition}=L_{GT}+\beta L_{SMLD}$ in Figure [2](#fig:model overview){reference-type="ref" reference="fig:model overview"}. Further details can be found in the Appendix [8](#apdx:detail){reference-type="ref" reference="apdx:detail"}. + +We cluster $R$ to implement the EM algorithm in Equation [\[eq:em\]](#eq:em){reference-type="ref" reference="eq:em"}. As discussed in Section [\[eq:em\]](#eq:em){reference-type="ref" reference="eq:em"}, we learn multiple dictionaries for different arities. Each dictionary is composed of prototypes $\phi^i\in\mathbb{R}^{N_{\phi^i}\times d_\phi}$ where $N_{\phi^i}$ is the dictionary size and $d_\phi$ is the dimension of the prototype. We train a predicate head for each dictionary; for example, for 1-ary, $\mu_i^1=f_\mu^1(r_i)$ maps a neural logic variable $r_i$ to a representation $\mu_i^1\in\mathbb{R}^{d_\mu}$, for 2-ary, $\mu_k^2=f_\mu^2(r_i,r_j)$ maps a pair $(r_i,r_j)$. In our work, mappers are implemented as convolution layers on visual parts and their combinations, e.g., $\mu^2_k=f_\mu^2(x_i+x_j)$. + +We perform an online clustering during training by maintaining a FIFO memory bank $M=\{M^i\}_{i=1}^{N_A}$ where $M^i\in\mathbb{R}^{L^i_M\times d_\mu}$ that adds new terms $\overrightarrow{\mu^i}$ after each training step. Similar to [@deepcluster], we run K-Means for every one or a few training steps after the warm-up epochs, in a set $\overrightarrow{\mu^i}+M^i$ for each dictionary, where, taking 1-ary as an example, $\overrightarrow{\mu^1}=(\mu^{1(1)},\mu^{1(2)},...,\mu^{1(B)})$ is the set of 1-ary representations in an input batch of length $B$. $\overrightarrow{\mu^i}$ has gradients while $M^i$ does not. We drop unwanted terms, such as empty ones, and randomly sample, for example, 30% pairs to increase efficiency. + +The K-Means output assignments $C^i$ for each term in $\overrightarrow{\mu^i}$, we make pseudo-labels $Y^i$ for prototypes in the dictionary $\phi^i$ by assigning them to their nearest clustering centroids from $C^i$. Then we minimize the distance between prototypes and their assigned centroids in latent space by Cross-Entropy (CE) loss $\min_{\phi^i,\theta} L_{CE}^i= CELoss(dist(\overrightarrow{\mu^i},\phi^i),Y^i)$ where $dist$ is a distance metric (e.g. L2 distance). The **Clustering Loss** is the sum of the CE loss for all dictionaries $L_{Clustering}=\gamma\sum_{i=1}^{N_A} L_{CE}^i$. We optimize it with the decomposition loss as $L_{TDL}=L_{Clustering}+L_{Decomposition}$ to implement the TDL framework. More details can be found in the Appendix [8.2](#asec:predicate_head){reference-type="ref" reference="asec:predicate_head"}. We visualize the latent space of a model trained in LineWorld for all the decomposed parts of the test set in Appendix [11](#apdx:tsne){reference-type="ref" reference="apdx:tsne"} by tSNE, where we can see that the predicates in the dictionary are learned as different clusters, while the baseline, UPD, does not provide predicate information and shows a less organized latent space. + +We employ PPO [@PPO] to tune an unsupervised learning model by regarding the decomposition process as an episode. We use a heuristic **shape score** to evaluate the decomposed part by 3 factors. (1) Continuity $R_{cont}$: the shape is not segmented and is an integral whole. $R_{cont}=\frac{max(A_C)}{sum(A_C)}$ where $A_C$ is the list of contoured areas for *segments* in a part. $R_{cont}=1$ if the part is not segmented. We use $findContours$ in OpenCV to obtain the segments for 2D data and DBSCAN for 3D after converting to point clouds. (2) Solidity $R_{solid}$: no holes inside a part. $R_{solid}=\frac{A_P}{sum(A_C)}$ where $A_P$ is the space or area of the part. $R_{solid}=1$ if there is no hole. (3) Smoothness $R_{smooth}$: the surfaces or contours of the part are smooth. $R_{smooth}=\frac{\rho_S}{\rho_O}$, where $\rho_S$ is the perimeter of the smoothed largest contour and $\rho_O$ is for the original contour. We apply RDP to smooth 2D data and alpha shape for 3D. The shape score $R_S=R_{cont}\times R_{solid}\times R_{smooth}$ normalized between 0 and 1. It can also be used to measure model performance. See Appendix [17](#apdx:score){reference-type="ref" reference="apdx:score"} for more details. diff --git a/2308.10436/main_diagram/main_diagram.drawio b/2308.10436/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..26fc327d9aede322ad60f6b334d3b89d69e360a0 --- /dev/null +++ b/2308.10436/main_diagram/main_diagram.drawio @@ -0,0 +1,101 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/2308.10436/main_diagram/main_diagram.pdf b/2308.10436/main_diagram/main_diagram.pdf new file mode 100644 index 0000000000000000000000000000000000000000..7c4d0a1f2c4f47a9ff68f56fc4bf93639c21a7b2 Binary files /dev/null and b/2308.10436/main_diagram/main_diagram.pdf differ diff --git a/2308.10436/paper_text/intro_method.md b/2308.10436/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..c8adfcda91477c5028e78d30ab9ae2cc7a89037d --- /dev/null +++ b/2308.10436/paper_text/intro_method.md @@ -0,0 +1,132 @@ +# Introduction + +Graph Neural Networks (GNNs) are popular tools to learn functions on graphs. They are commonly designed to be permutation equivariant since the node ordering can be arbitrary (in the matrix representation of a graph). Permutation equivariance serves as a strong geometric prior and allows GNNs to generalize well [1–3]. Yet in many applications, the node ordering across different graphs is matched or fixed a priori, such as a time series of social networks where the nodes identify the same users, or a set of skeleton graphs where the nodes represent the same joints. In such settings, the natural symmetries arise from graph automorphisms, which effectively only act on the graph signals; This is inherently different from the standard equivariance in GNNs that concerns all possible permutations acting on both the signals and the graph domain. A permutation of both the graph and the graph signal can be seen as a *change of coordinates* since it does not change the object it represents, just the way to express it. This parallels the *passive symmetries* in physics, where physical observables are independent of the coordinate system one uses to express them [4, 5]. In contrast, a permutation of the graph signal on a fixed graph potentially transforms the object itself, not only its representation. This parallels the *active symmetries* in physics, where the coordinate system (in this case, the graph or domain) is fixed but a group transformation on the signal results in a predictable + +transformation of the outcome (e.g., permuting left and right joints in the human skeleton changes a left-handed person to right-handed). The active symmetries on a fixed graph are similar to the translation equivariance symmetries in Euclidean convolutional neural networks (CNNs), where the domain is a fixed-size grid and the signals are images. + +In this work, we focus on active symmetries in GNNs. Specifically, we consider a fixed graph domain G with N nodes, an adjacency matrix $A \in \mathbb{R}^{N \times N}$ , and input graph signals $X \in \mathbb{R}^{N \times d}$ . We are interested in learning equivariant functions f that satisfy (approximate) active symmetries + +$$f(\Pi X) \approx \Pi f(X)$$ + for permutations $\Pi \in \mathcal{G} \subseteq \mathcal{S}_N$ , (1) + +where $\mathcal{G}$ is a subgroup of the permutation group $\mathcal{S}_N$ that depends on G. For example, $\mathcal{G}$ can be the graph automorphism group $\mathcal{A}_G = \{\Pi : \Pi A = A\Pi\}$ . Each choice of $\mathcal{G}$ induces a hypothesis class $\mathcal{H}_{\mathcal{G}}$ for f: the smaller the group $\mathcal{G}$ , the larger the class $\mathcal{H}_{\mathcal{G}}$ . We aim to select $\mathcal{G}$ so that the learned function $f \in \mathcal{H}_{\mathcal{G}}$ generalizes well, also known as the *model selection* problem [6, Chp.4]. In contrast, standard graph learning methods (GNNs, spectral methods) are defined to satisfy passive symmetries, by treating A as input and requiring $f(\Pi A\Pi^\top, \Pi X) = \Pi f(A, X)$ for all permutations $\Pi \in \mathcal{S}_N$ and all N. But for the fixed graph setting, we argue that active symmetries are more relevant. Thus we use A to define the hypothesis class rather than treating it an input. By switching from passive symmetries to active symmetries, we will show how to design GNNs for signals supported on a fixed graph with different levels of expressivity and generalization properties. + +While enforcing symmetries has been shown to improve generalization when the symmetry group is known a priori [7–13], the problem of *symmetry model selection* is not completely solved, particularly when the data lacks exact symmetries (see for instance [14–16] and references therein). Motivated by this, we study the symmetry model selection problem for learning on a fixed graph domain. This setting is particularly interesting since (1) the graph automorphism group serves as the natural *oracle* symmetry; (2) real-world graphs tend to be asymmetric, but admit cluster structure or local symmetries. Therefore, we define *approximate symmetries of graphs* using the cut distance between graphs from graphon analysis. An approximate symmetry of a graph G is a symmetry of any other graph G' that approximates G in the cut distance. In practice, we take G' as coarse-grainings (or clusterings) of G, as these are typically guaranteed to be close in cut distance to G. We show how to induce approximate symmetries for G via the automorphisms of G'. Our main contributions include: + +- 1. We formalize the notion of active symmetries and approximate symmetries of GNNs for signals supported on a fixed graph domain, which allows us to study the symmetry group model selection problem. (See Sections 2, 4) +- 2. We theoretically characterize the statistical risk depending on the hypothesis class induced from the symmetry group, and show a bias-variance tradeoff between the reduction in expressivity and the gain in regularity of the model. (See Sections 3, 4) +- 3. We illustrate our approach empirically for image inpainting, traffic flow prediction, and human pose estimation. (See Section 5 for an overview, and Appendix D for the details on how to implement equivariant graph networks with respect to different symmetry groups). + +# Method + +**Require:** Abelian group $A_G = (S_2)^k$ + +- 1. Construct the character table of $\chi_{\text{irreps}}$ for $A_G$ , i.e. $\chi_i : A_G \to \{\pm 1\}$ $i = 1, \dots \ell$ ; +- 2. For each character $\chi_i$ in the character table, compute the projection matrix + +$$P_{\chi_i}(\mathcal{X}) = [P_{\chi_i}(e_1); \dots; P_{\chi_i}(e_N)] \in \mathbb{R}^{N \times N};$$ + (12) + +followed by computing the basis from $P_{\chi_i}(\mathcal{X})$ and call it $\mathcal{X}_{\chi_i} = [b_{\chi_i}^{(1)}, \dots, b_{\chi_i}^{(K_i)}]$ . + +3. $\mathcal{X}=\oplus_{i=1}^\ell \mathcal{X}_{\chi_i}$ where $\mathcal{X}_{\chi_i}$ are the isotypic component. Then f is any linear function satisfying that $f(\mathcal{X}_{\chi_i})\subseteq \mathcal{X}_{\chi_i}$ for all $i=1,\ldots,\ell$ . In particular, in the basis $[b_{\chi_i}^{(s)}]_{1\leq i\leq \ell,1\leq s\leq K_i}$ f can be written as a block diagonal matrix $\mathbb{R}^{n\times n}$ with each block $M_{\chi_i}$ being the linear map from $\mathcal{X}_{\chi_i}\to\mathcal{X}_{\chi_i}$ , + +$$f = \begin{bmatrix} M_{\chi_1} & & & \\ & M_{\chi_2} & & \\ & & \ddots & \\ & & & M_{\chi_{\ell}} \end{bmatrix}. \tag{13}$$ + +return f + +$$\begin{array}{c|cccc} +e & \sigma \\ +\hline +\chi_e & 1 & 1 \\ +\chi_2 & 1 & -1 +\end{array}$$ + +Table 1: Character table for $\operatorname{aut}(P_4) \cong Z_2$ + +for some scalars $\alpha_1, \alpha_2$ and $v_{\chi_1} \in \mathcal{X}_{\chi_1}, v_{\chi_2} \in \mathcal{X}_{\chi_2}$ . Then by (14), for all $g \in \mathcal{G}$ , + +$$f(g v_{\chi_1}) = f(\lambda_1(g) v_{\chi_1}) = \lambda_1(g) f(v_{\chi_1}) = \lambda_1(g) \alpha_1 v_{\chi_1} + \lambda_1(g) \alpha_2 v_{\chi_2}.$$ +(17) + +Now, since f is equivariant, for all $g \in \mathcal{G}$ , + +$$f(g v_{\chi_1}) = gf(v_{\chi_1}) = g(\alpha_1 v_{\chi_1} + \alpha_2 v_{\chi_2}) = \lambda_1(g)\alpha_1 v_{\chi_1} + \lambda_2(g)\alpha_2 v_{\chi_2}.$$ +(18) + +But there exists some $g' \in \mathcal{G}$ such that $\lambda_1(g') \neq \lambda_2(g')$ , which leads to $f(g'v_{\chi_1}) \neq f(g'v_{\chi_1})$ , a contradiction. One can easily extend the proof strategy to the general case for $f(v_{\chi_1}) = \sum_{i=1}^l v_{\chi_i}$ . + +**Example A.1.** Consider the path graph on 4 nodes (i.e., $P_4$ ). We have aut $(P_4) = \{e, (14)(23)\} \cong Z_2$ . + +Steps 1: Note that $Z_2$ is abelian and thus all irreducible characters $\chi(g) \in \{\pm 1\}$ , for all $g \in Z_2$ . The character table is shown in Table 1. + +Step 2: using (11) we have + +$$P_{\chi_e}[e_1;e_2;e_3;e_4] = \frac{1}{2} \begin{bmatrix} 1 & 0 & 0 & 1 \\ 0 & 1 & 1 & 0 \\ 0 & 1 & 1 & 0 \\ 1 & 0 & 0 & 1 \end{bmatrix} \text{ which yields basis } \mathcal{B}(P_{\chi_e}) = [e_1 + e_4;e_2 + e_3].$$ + +$$P_{\chi_2}[e_1;e_2;e_3;e_4] = \frac{1}{2} \begin{bmatrix} 1 & 0 & 0 & -1 \\ 0 & 1 & -1 & 0 \\ 0 & -1 & 1 & 0 \\ -1 & 0 & 0 & 1 \end{bmatrix} \text{ which yields basis } \mathcal{B}(P_{\chi_2}) = [e_1 - e_4;e_2 - e_3].$$ + +Step 3: Parameterize $f: \mathbb{R}^4 \to \mathbb{R}^4$ by $f: \mathcal{B}(P_{\chi_e}) \to \mathcal{B}(P_{\chi_e})$ and $f: \mathcal{B}(P_{\chi_2}) \to \mathcal{B}(P_{\chi_2})$ . For all $v \in \mathbb{R}^4$ , write $v = c_1(e_1 + e_4) + c_2(e_2 + e_3) + c_3(e_1 - e_4) + c_4(e_2 - e_3)$ , then + +$$f(v) = \begin{bmatrix} \alpha_1 & \alpha_2 \\ \alpha_3 & \alpha_4 \end{bmatrix} \begin{bmatrix} c_1 \\ c_2 \end{bmatrix} + \begin{bmatrix} \alpha_5 & \alpha_6 \\ \alpha_7 & \alpha_8 \end{bmatrix} \begin{bmatrix} c_3 \\ c_4 \end{bmatrix}, \tag{19}$$ + +where $\alpha_1, \ldots, \alpha_8$ are (learnable) real scalars. Now f is linear, equivariant by construction. + +In this section, we provide additional details of constructing equivariant linear maps for using the symmetry group induced by graph coarsening (Definition 3). Recall the symmetry group with M clusters of G (with the associated coarsened graph G') is given by + +$$\mathcal{G}_{G \to G'} := \left( \mathcal{S}_1 \times \mathcal{S}_2 \ldots \times \mathcal{S}_M \right) \rtimes \overline{\mathcal{A}}_{G'} \subset \mathcal{S}_N.$$ + +We first assume that $\overline{\mathcal{A}}_{G'}$ is trivial and show how to parameterize equivariant functions with respect to products of permutations. Then we discuss more general cases, for instance if $\overline{\mathcal{A}}_{G'}$ is abelian. For the ease of exposition, we consider $X \in \mathbb{R}^N$ , $Y \in \mathbb{R}^N$ . + +Suppose $\overline{\mathcal{A}}_{G'}$ is trivial, we claim that a linear function $f: \mathbb{R}^N \to \mathbb{R}^N$ is equivariant with respect to $\mathcal{G}_{G \to G'}$ if and only if it admits the following block-matrix form: + +$$f = \begin{bmatrix} f_{11} & f_{12} & \cdots & f_{1M} \\ f_{21} & f_{22} & \cdots & f_{2M} \\ & & \ddots & \\ f_{M1} & f_{M2} & \cdots & f_{MM} \end{bmatrix}, f_{kk} = a_k I_{c_k} + b_k \mathbb{1}_{c_k} \mathbb{1}_{c_k}^\top, f_{kl} = e_{kl} \mathbb{1}_{c_k} \mathbb{1}_{c_l}^\top \text{ for } k \neq l, \quad (20)$$ + +where $f_{kl}$ are block matrices, $I_k$ is a size-k identity matrix, $\mathbb{1}_k$ is a length-k vector of all ones, and $a_k, b_k, e_{kl}$ are scalars. The subscript $c_k$ denotes the size of the k-th cluster. Figure 5 illustrates the block structure of f. + +We provide a proof sketch of our claim. Since $\overline{\mathcal{A}}_{G'}$ is trivial by assumption, it remains to show any function that is equivariant to the product of permutations $(S_1 \times S_2 \dots \times S_M)$ admits the form in eqn. 20. We justify as follows: + +- 1. (within-cluster) $f_{kk}$ is a linear permutation-equivariant function if and only if its diagonal elements are the same and its off-diagonal elements are the same ([34, Lemma 3.]); +- 2. (across-cluster) $f_{kl}$ for $k \neq l$ is a constant matrix since nodes within a cluster are indistinguishable; moreover, $e_{kl}$ are unconstrained since the coarsened symmetry $\overline{\mathcal{A}}_{G'}$ is trivial. + +![](_page_18_Picture_10.jpeg) + +Figure 5: The block structure of equivariant linear function $f: \mathbb{R}^n \to \mathbb{R}^n$ with respect to $\mathcal{G}_{G \to G'}$ (where G, G' are asymmetric): Each diagonal block $f_{kk}$ is diagonally constant and off-diagonally constant; Each off-diagonal block $f_{kl}$ is a constant matrix. + +As an example, we illustrate the equivariant linear layer for two-cluster graph coarsening. Without loss of generality, assume that the node signals X are ordered according to the cluster assignment (e.g., $X_{[1:|V_1|]}$ are node features for the first cluster, etc). Let $X_{(1)}, X_{(2)}$ denote the node features for the first and second cluster, $W_{(1)}^s, W_{(2)}^s$ denote the weights on the block diagonal for self-feature transformation, $W_{(1)}^n, W_{(2)}^n$ denote the weights on the block diagonal for within-cluster neighbors, and $W_{(12)}^n, W_{(21)}^n$ denote the weights off the block diagonal for across-cluster neighbors. Let I denote the identity matrix, $\mathbf{1}_{(1)}, \mathbf{1}_{(2)}$ denote the all-ones matrices with the same size as the corresponding + +cluster, and $\mathbf{1}_{(12)}, \mathbf{1}_{(21)}$ denote the all-ones matrices mapping across clusters. Recall $\odot$ denotes the element-wise multiplication of two matrices. Then the equivariant linear layer is parameterized as + +$$I \odot \begin{bmatrix} X_{(1)}W_{(1)}^s \\ X_{(2)}W_{(2)}^s \end{bmatrix} + \begin{pmatrix} \begin{bmatrix} \mathbf{1}_{(1)} & 0 \\ 0 & \mathbf{1}_{(2)} \end{bmatrix} - I \end{pmatrix} \odot \begin{bmatrix} X_{(1)}W_{(1)}^n \\ X_{(2)}W_{(2)}^n \end{bmatrix} + \begin{bmatrix} 0 & \mathbf{1}_{(12)} \\ \mathbf{1}_{(21)} & 0 \end{bmatrix} \odot \begin{bmatrix} X_{(1)}W_{(12)}^n \\ X_{(2)}W_{(21)}^n \end{bmatrix}. (21)$$ + +For cases where $\overline{\mathcal{A}}_{G'}$ is nontrivial, if $\overline{\mathcal{A}}_{G'}$ is abelian, we can use a construction by Serre ([83] Section 8.2). Observe that the symmetry of $\overline{\mathcal{A}}_{G'}$ effectively constrains the patterns of $e_{kl}$ in equation 20. We provide an example where $\overline{\mathcal{A}}_{G'} = \mathcal{S}_2$ in our image inpainting experiment (see Section 5.1), and defer the investigation of general cases for future work. + +**Lemma 1** (Risk Gap). Let $\mathcal{X} = \mathbb{R}^{N \times d}$ , $\mathcal{Y} = \mathbb{R}^{N \times k}$ be the input and output graph signal spaces on a fixed graph G. Let $X \sim \mu$ where $\mu$ is a $\mathcal{S}_N$ -invariant distribution on $\mathcal{X}$ . Let $Y = f^*(X) + \xi$ , where $\xi \in \mathbb{R}^{N \times k}$ is random, independent of X with zero mean and finite variance and $f^* : \mathcal{X} \to \mathcal{Y}$ is $\mathcal{A}_G$ -equivariant. Then, for any $f \in V$ and for any compact group $\mathcal{G} \subseteq \mathcal{S}_N$ , we can decompose f as + +$$f = \bar{f}_{\mathcal{G}} + f_{\mathcal{G}}^{\perp},$$ + +where $\bar{f}_{\mathcal{G}} = \mathcal{Q}_{\mathcal{G}} f, f_{\mathcal{G}}^{\perp} = f - \bar{f}_{\mathcal{G}}$ . Moreover, the risk gap satisfies + +$$\Delta(f,\bar{f}_{\mathcal{G}}) = \mathbb{E}\left[\|Y - f(X)\|_F^2\right] - \mathbb{E}\left[\|Y - \bar{f}_{\mathcal{G}}(X)\|_F^2\right] = \underbrace{-2\langle f^*,f_{\mathcal{G}}^{\perp}\rangle_{\mu}}_{\textit{mismatch}} + \underbrace{\left\|f_{\mathcal{G}}^{\perp}\right\|_{\mu}^2}_{\textit{constraint}}.$$ + +Lemma 1 is a straightforward extension of Lemma 6 in [8], which makes use of Lemma 1 in [8]. + +**Lemma 1 in [8].** Let U be any subspace of V that is closed under $\mathcal{Q}$ . Define the subspaces S and A of, respectively, the $\mathcal{G}$ -symmetric and $\mathcal{G}$ -anti-symmetric functions in $U:S=\{f\in U:f$ is $\mathcal{G}$ -equivariant A and $A=\{f\in U:\mathcal{Q}f=0\}$ . Then A admits admits an orthogonal decomposition into symmetric and anti-symmetric parts + +$$U = S \oplus A$$ + +*Proof of Lemma 1.* The first part of Lemma 1 $f = \bar{f}_{\mathcal{G}} + f_{\mathcal{G}}^{\perp}$ follows from Lemma 1 in [8]. For the second part, by the assumption that the noise $\xi$ is independent of X with zero mean and finite variance, we can simplify the risk gap as + +$$\Delta(f, \bar{f}_{\mathcal{G}}) := \mathbb{E}\left[\|Y - f(X)\|_F^2\right] - \mathbb{E}\left[\|Y - \bar{f}_{\mathcal{G}}(X)\|_F^2\right] = \mathbb{E}\left[\|f^*(X) - f(X)\|_F^2\right] - \mathbb{E}\left[\|f^*(X) - \bar{f}_{\mathcal{G}}(X)\|_F^2\right].$$ +(22) + +Substituting $f = \bar{f}_{\mathcal{G}} + f_{\mathcal{G}}^{\perp}$ yields + +$$\mathbb{E}\left[\|f^{*}(X) - \bar{f}_{\mathcal{G}}(X) - f_{\mathcal{G}}^{\perp}(X)\|_{F}^{2}\right] - \mathbb{E}\left[\|f^{*}(X) - \bar{f}_{\mathcal{G}}(X)\|_{F}^{2}\right] += -2\langle f^{*}(X) - \bar{f}_{\mathcal{G}}(X), f_{\mathcal{G}}^{\perp}(X)\rangle_{\mu} + \mathbb{E}\left[\|f_{\mathcal{G}}^{\perp}(X)\|_{F}^{2}\right] += -2\langle f^{*}, f_{\mathcal{G}}^{\perp}\rangle_{\mu} + \|f_{\mathcal{G}}^{\perp}\|_{\mu}^{2}.$$ +(23) + +We remark that Lemma 6 in [8] assumes that $f^*$ is $\mathcal{G}$ -equivariant, so the first term in (23) vanishes. We are motivated from the symmetry model selection problem, and thereby relax the assumption of the chosen symmetry group $\mathcal{G}$ can differ from the target symmetry group $\mathcal{A}_{\mathcal{G}}$ . + +**Theorem 2** (Bias-Variance-Tradeoff). Let $\mathcal{X} = \mathbb{R}^{N \times d}$ , $\mathcal{Y} = \mathbb{R}^{N \times k}$ be the graph signals spaces on a fixed graph G. Let $S_N$ act on $\mathcal{X}$ and $\mathcal{Y}$ by permuting the rows with representations $\phi$ and $\psi$ . Let $\mathcal{G}$ be a subgroup of $S_N$ acting with restricted representations $\phi|_{\mathcal{G}}$ on $\mathcal{X}$ and $\psi|_{\mathcal{G}}$ on $\mathcal{Y}$ . diff --git a/2310.04928/main_diagram/main_diagram.drawio b/2310.04928/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..9159894df89311ebc60cda902d3b83df6b297979 --- /dev/null +++ b/2310.04928/main_diagram/main_diagram.drawio @@ -0,0 +1,88 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/2310.04928/main_diagram/main_diagram.pdf b/2310.04928/main_diagram/main_diagram.pdf new file mode 100644 index 0000000000000000000000000000000000000000..06ddb276c1c92db43750c4ea330d0e656661a03b Binary files /dev/null and b/2310.04928/main_diagram/main_diagram.pdf differ diff --git a/2310.04928/paper_text/intro_method.md b/2310.04928/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..ddab5d590eadebf78e5dfc8dc431ca573084c5e7 --- /dev/null +++ b/2310.04928/paper_text/intro_method.md @@ -0,0 +1,40 @@ +# Introduction + +The evaluation of large language models (LLMs) has predominantly relied on English datasets to assess language proficiency [@wang-etal-2018-glue; @baradaran2022survey], reasoning abilities [@zellers-etal-2019-hellaswag; @huang-etal-2019-cosmos; @bisk2020piqa; @talmor-etal-2019-commonsenseqa], and real-world knowledge [@hendrycksmeasuring]. LLMs such as GPT-3.5 [@ouyang2022training], Falcon [@penedo2023refinedweb], and BLOOMZ [@muennighoff2022crosslingual], however, are pre-trained on large-scale multilingual data, and thus it is critical to evaluate what knowledge they capture and their reasoning abilities in languages beyond English. + +School exams serve as a powerful means to assess the reasoning abilities and real-world knowledge of LLMs, given that these tests are meticulously designed by expert educators, drawing upon the principles of learning science. At various educational levels, school exams function as assessment tools, evaluating not only language proficiency but also higher-order cognitive skills such as comprehension, analytic abilities, and the application of real-world knowledge across diverse scenarios [@novak1988learning]. + +@hendrycksmeasuring proposed `MMLU`, a massive multitask language understanding benchmark in English that is compiled from different exams, covering topics including US history, computer science, and high school subjects. Recent progresses on LLMs such as LLaMA [@touvron2023llama] and GPT--4 [@OpenAI2023GPT4TR] use `MMLU` as one of the evaluation datasets. In the GPT-4 technical report, automatic evaluation is further extended to encompass various standardized exams, including SAT, GRE, and bar exams. + +
+ +
Distribution of subject areas and education levels in IndoMMLU. “Hum”, “Social”, “Indo”, and “Local” refer to Humanities, Social Science, Indonesian Language, and Local Languages and Cultures, respectively.
+
+ +While there has been a plethora of work on LLM evaluation for English [@OpenAI2023GPT4TR; @katz2023gpt; @choi2023chatgpt; @ryznar2023exams; @chalkidis2023chatgpt], there has been comparatively little work in other languages [@li2023cmmlu; @sengupta2023jais]. Recent work by @OpenAI2023GPT4TR evaluated GPT-4 using a translated version of `MMLU`, and reported strong performance. While encouraging, using translations of English evaluation datasets has serious shortcomings, including translation noise, a complete lack of content that is sensitized to the local language/culture (esp. as most English evaluation datasets are highly US centric), and conversely, the existence of content that is irrelevant to the local language/culture (e.g. questions relating to US law or customs) and incongruent with the language-specific evaluation [@liu2023multilingual]. + +In this paper, we ask professional teachers (of Indonesian nationality) to collect exam questions from various educational levels in Indonesian schools (i.e. primary school, junior high school, senior high school, and university). We categorize the collected questions into different subject areas, including: (1) STEM (Science, Technology, Engineering, and Mathematics); (2) Social Science; (3) Humanities; (4) Indonesian Language; and (5) Local Languages and Cultures. Figure [1](#fig:overview){reference-type="ref" reference="fig:overview"} presents an overview of the distribution of the resulting dataset, `IndoMMLU`, across different subject areas and education levels. It is worth mentioning that 21% of the questions specifically focus on the Indonesian language, and 25% encompass nine distinct local languages and cultures that are specific to Indonesia. + +Our contributions can be summarized as follows: + +- We introduce the first Indonesian `MMLU` dataset, namely `IndoMMLU`, which comprises 64 tasks across different subject areas and education levels in Indonesia. + +- Our dataset includes exam questions from school grades 1 to 12, as well as university entrance exams. This comprehensive coverage allows us to perform fine-grained assessment of the Indonesian language proficiency of existing LLMs. + +- Approximately 25% of our data encompasses nine distinct local languages and cultures in Indonesia, namely Lampungic (`ljp`), Balinese (`ban`), Makassarese (`mak`), Banjarese (`bjn`), Madurese (`mad`), Sundanese (`sun`), Javanese (`jav`), Dayak Ngaju (`nij`), and Minangkabau.[^2] These questions are not only in under-represented languages but also incorporate specific cultural content, such as art, poetry, and daily life. For Lampungic (`ljp`) and Makassarese (`mak`) in particular, this is the very first NLP resource to be released. + +- We evaluate various multilingual LLMs, including GPT-3.5 [@ouyang2022training], XGLM [@lin2021few], Falcon [@penedo2023refinedweb], BLOOMZ [@muennighoff2022crosslingual], mT0 [@muennighoff2022crosslingual], LLaMA [@touvron2023llama], and Bactrian-X [@li2023bactrian], across different model sizes. We find that only GPT-3.5 passes the highest primary school level exam, and no models demonstrate familiarity with local Indonesian languages and culture. + +# Method + +`IndoMMLU` is a multiple-choice problem set in 64 subjects from different education levels, following the format of English `MMLU` (see Figure [2](#fig:ex1){reference-type="ref" reference="fig:ex1"} and Figure [3](#fig:ex2){reference-type="ref" reference="fig:ex2"}). `IndoMMLU`, however, is based on the Indonesian education curriculum, and has more fine-grained education levels than `MMLU`. + +In Indonesia's curriculum, schools are categorized into three levels: (1) six years of primary school (*Sekolah Dasar* = "SD"), (2) three years of junior high school (*Sekolah Menengah Pertama* = "SMP"), and (3) three years of senior high school (*Sekolah Menengah Atas* = "SMA"). At primary school, pupils in all grades are taught the Indonesian language, civics, mathematics, art, sports, and religion. From grade 4 to 6 and in junior high school, pupils additionally learn a foreign language, a local language/culture, science, and social science.[^3] In senior high school, pupils study more specialized natural science and social science subjects, including physics, chemistry, biology, geography, sociology, economy, and history. In `IndoMMLU`, we explicitly exclude mathematics because the questions typically consist primarily of symbols with little language content, and there are existing datasets for mathematical reasoning such as `GSM-8K` [@cobbe2021training] and `NumGLUE` [@mishra-etal-2022-numglue]. + +The local language/culture subjects vary across provinces in Indonesia and depend on the local government policy. For example, in West Sumatra, Minangkabau culture is taught using the Indonesian language, while in West Java, pupils are exposed to the Sundanese language and culture. Figure [2](#fig:ex1){reference-type="ref" reference="fig:ex1"} illustrates two exam questions for Minangkabau culture, and one exam question for Sundanese. + +We asked seven professional teachers with at least a bachelor's degree in education to gather publicly-available school exam questions in Indonesia from web sources.[^4] They were tasked with gathering problems for specific subject areas and educational levels, as well as metadata such as the source (i.e. URL of the source document), school level, class level, question, multiple-choice options, and the correct answer key. We instructed the teachers to only include exams that had accompanying answer keys, and to exclude problems that contained images. Additionally, we organized an 1-hour workshop to discuss the data collection procedure with all the teachers, addressing any questions or concerns they had. All teachers are paid competitively, higher than the Indonesian average monthly wage.[^5] + +To ensure the accuracy of the data entry process, we randomly checked questions collected by each teacher. We manually verified the questions, multiple-choice options, and the corresponding answer keys based on the given URL, and found that each teacher conducted the work accurately. We also additionally performed automatic filtering to remove repetitive questions, and remove questions that have no answer key. + +After data cleansing, we obtained a total of 14,981 questions, distributed over school levels and subjects as detailed in Figure [1](#fig:overview){reference-type="ref" reference="fig:overview"}; the details of each subject area are in Table [\[tab:group\]](#tab:group){reference-type="ref" reference="tab:group"} and the Appendix. `IndoMMLU` consists of 30% primary school, 24% junior high school, 32% senior high school, and 14% university entrance exam questions. Table [\[tab:stat\]](#tab:stat){reference-type="ref" reference="tab:stat"} shows the average question length for each education level and subject area. We can observe that primary school questions tend to be shorter and university entrance exam questions are longer. Indonesian language questions have the highest average length, while local languages and culture questions are around 88 characters on average. diff --git a/2310.05719/main_diagram/main_diagram.drawio b/2310.05719/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..140c53034e9265dce80c62a8f12167561d52b76b --- /dev/null +++ b/2310.05719/main_diagram/main_diagram.drawio @@ -0,0 +1 @@ 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 \ No newline at end of file diff --git a/2310.05719/main_diagram/main_diagram.pdf b/2310.05719/main_diagram/main_diagram.pdf new file mode 100644 index 0000000000000000000000000000000000000000..91ec5ebed52e77ee9df2e13eeca08cf13c66b06a Binary files /dev/null and b/2310.05719/main_diagram/main_diagram.pdf differ diff --git a/2310.05719/paper_text/intro_method.md b/2310.05719/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..a286e1bdd72ea6d04adc249e30caa51165c28ad7 --- /dev/null +++ b/2310.05719/paper_text/intro_method.md @@ -0,0 +1,89 @@ +# Introduction + +Transformers, as introduced by Vaswani et al. (2017), have profoundly impacted machine learning, establishing a prevailing neural network architecture across various domains. Transformers consistently excel in different fields, including natural language processing (Lin et al., 2022), time series forecasting (Wen et al., 2022), and computer vision (Dosovitskiy et al., 2020). Their success can be attributed to their scaling properties (Kaplan et al., 2020) and efficient utilization of contemporary hardware architectures designed for extensive parallel computing. The unification of a single architecture across tasks facilitates immediate, far-reaching applicability of any analysis that handles general properties of the Transformer architecture. + +As large Transformer foundation models (Bommasani et al., 2021) continue to grow in size and complexity, the challenges associated with training, i.e., exponential increase in parameters and compute for a fixed incremental improvement in performance (Hoffmann et al., 2022; Zhai et al., 2022; Bachmann et al., 2023), become increasingly more perilous. Consequently, achieving state-ofthe-art results is often confined to researchers with access to ample GPU resources. To address these issues and strive for more efficient and sustainable performance improvements, we embark on the following more compelling and alternative inquiry: + +*Can we combine the capabilities of pre-trained Transformer models?* + +Merging multiple Transformer models into a single entity while preserving their unique capabilities can yield several advantages; (a) *Enhanced performance* by harnessing the collective capabilities of individual models. (b) *Reduced inference complexity*, as querying a single model replaces the need to query n models in an ensemble, reducing computational (FLOPs) and storage requirements by + +\*These authors contributed equally to this work + +a factor of n. (c) *The necessity to train from scratch can be readily eliminated*, leveraging existing public models, already available, and numerous in quantity 1 . + +A straightforward way of fusing, i.e., merging, models of the same architecture, is to average their weight matrices one-to-one, referred to as 'Vanilla Fusion' (VF). However, this method overlooks potential misalignments between the parameter matrices, arising due to neurons at the same positions, in different models, encoding different information (Godfrey et al., 2022). Instead, we propose to use Optimal Transport fusion (OTFusion) (Singh & Jaggi, 2020), which at its core, aligns the weight or parameter matrices before fusing them. + +Thus, by virtue of such an alignment, OTFusion ensures that the fused model effectively integrates the knowledge and capabilities of the individual models to be merged, rather than simply averaging the weight matrices without guaranteeing meaningful information preservation. Additionally, OTFusion accommodates the fusion of models with different widths, and in turn, different sizes, which is fundamentally not possible with VF. This is a crucial feature, as such heterogeneous models are available in plenty, to better unleash the potential of existing pre-trained models. Consequently, OTFusion has been shown to be an effective method for fusing fully connected (Singh & Jaggi, 2020), convolutional (Nguyen et al., 2021) and recurrent neural networks (Akash et al., 2022) on a variety of tasks, heavily outperforming VF. + +Yet, despite its wide adoption (Nguyen et al., 2021; Liu et al., 2022; Ainsworth et al., 2022), the layerwise procedure proposed by OTFusion does not fit well with contemporary architectural design, that comprises of constant residual streams, normalization layers, and attention operations. It is not equipped in any way to align and fuse models with complex information streams and to fuse transformer-specific components. Hence, the primary aim of our work is to develop techniques that help bridge these gaps and successfully generalize fusion to Transformer-based architectures. + +Our contributions are: (a) We analyze each of the idiosyncratic architectural components in Transformers in thorough detail, with an ultimate aim to best fuse them across different models. Throughout our discussion, we exposit our approach based on the perspective of *flow of the transportation maps*2 , that makes for intuitive visualizations and interpretation. (b) We uncover that, surprisingly, OTFusion based on a *hard-alignment underperforms* in this context, contrary to the case of fully-connected or convolutional architectures; and that, *soft-alignment plays a key role* in successful one-shot fusion. (c) We showcase the efficacy of our approach by extensive experimentation involving the fusion and finetuning of Vision Transformers (ViTs) across multiple datasets, including CIFAR10, CIFAR100, TINY IMAGENET and IMAGENET-1K, as well as BERT (Devlin et al., 2018) models for natural language tasks. We *consistently outperform* the original *converged* models across tasks and datasets, by about ∼ *1.0%*, *while significantly reducing computational and storage costs by a factor of* n*.* + +Overall, our research marks an important stride in advancing model fusion techniques, that help deliver enhanced performance and efficiency for modern Transformer based architectures. + +# Method + +With a modular architecture like the transformer, it is intuitive to use a divide-and-conquer approach to develop a fusion algorithm. Therefore, we first divide the architecture into its simplest building block — fully connected layers — that can be fused by the prevalent OTFusion strategy. The question remains; how to effectively connect these building blocks, especially if heterogeneous? How to hierarchically reconstruct a fully fused transformer ensuring consistency of the single fused blocks? + +As we provide solutions to such open questions, we will guide our discussion in this section with a transport flow perspective, which allows for an intuitive and effective concatenation of blocks of any sort, and that, therefore, in principle can be applied to every architecture. Henceforth, we will use the notation from Vaswani et al. (2017) for Transformers. We display our methods in the non-masked self-attention case, but our method can generalize to the cross-attention or causal masked attention. + +In the typical OTFusion application, the TM of the previous layer is simply passed to the next layer. However, in more complex architectures, the incoming TM of a layer can depend on multiple TMs. To formalize and visualize this flow of TMs, we present the *Transportation Map Flow Graph*. + +To introduce the concept, we use the flow graph of a residual connection (Fig. 1). Rectangles represent the neural network layers; red nodes represent any non-learnable computations or permutations inside the network; edges represent the propagation of the TMs. Layers have exactly one incoming and one outgoing edge. Computation nodes always have multiple incoming edges and one outgoing edge, where the outgoing TM must depend on the incoming TMs. A major contribution of this work is to handle the various complex transportation map flows throughout the transformer architecture. + +In residual connections, the outputs of a current layer and a residual layer are summed up. The TMs coming from these two layers will be different, therefore the ideal TM flow strategy has to be determined. We explored three heuristics to calculate a weighting vector $\gamma^{(\ell)}$ , where each entry $\gamma_i^{(\ell)}$ scales the corresponding rows of the TMs. After obtaining $\gamma^{(\ell)}$ we compute the weighted average as shown in Eq. 1. Find the results in Sec. 5.1. + +![](_page_3_Figure_12.jpeg) + +Figure 1: TM flow graph for a residual connection. + +$$\mathbf{T}_{\text{out}}^{(\ell)} = \mathbf{T}_{\text{current}}^{(\ell)} \operatorname{diag}(\mathbf{1} - \boldsymbol{\gamma}^{(\ell)}) + \mathbf{T}_{\text{residual}}^{(\ell)} \operatorname{diag}(\boldsymbol{\gamma}^{(\ell)}) \tag{1}$$ + +**Averaging.** For plain averaging, as proposed by Singh & Jaggi (2020), we set $\forall i, \gamma_i = 0.5$ . This heuristic does not depend on activations and can therefore be used even in the case of weight-based alignment. However, it introduces the strict assumption that the residual and the current layer TM are + +of equal importance when aligning the subsequent layer. We therefore extend Singh & Jaggi (2020) with two novel residual policies. + +Weighted Scalar. To alleviate the equal contribution constraint from the averaging method, we compute a weighting factor $\forall i, \gamma_i^{(\ell)} = \gamma_{\text{scalar}}^{(\ell)}$ (Eq. 2). We use the activations of the anchor model, over a batch of samples S, because only those carry information about the importance of the current and the residual branch in the anchor model to which we try to align the other models. $\mathbf{f}_{\text{residual}}^{(\ell)}(\mathbf{x})$ are the activations from the residual branch while $\mathbf{f}_{\text{current}}^{(\ell)}(\mathbf{x})$ are the activations from the current layer $\ell$ . + +$$\gamma_{\text{scalar}}^{(\ell)} = \frac{\sum_{\mathbf{x} \in S} ||\mathbf{f}_{\text{residual}}^{(\ell)}(\mathbf{x})||_1}{\sum_{\mathbf{x} \in S} ||\mathbf{f}_{\text{current}}^{(\ell)}(\mathbf{x})||_1 + \sum_{\mathbf{x} \in S} ||\mathbf{f}_{\text{residual}}^{(\ell)}(\mathbf{x})||_1}$$ +(2) + +**Weighted Matrix.** As opposed to the Weighted Scalar method, here, we calculate a weight vector $\gamma^{(\ell)}$ where each entry $\gamma_i^{(\ell)}$ weighs one strand of a residual connection. The computation of each $\gamma_i^{(l)}$ is similar to Eq. 2 but here we do not compute the $\ell^1$ -Norm over the whole activation vectors, instead, we take the absolute value of the corresponding i-th values of the activation vectors. + +We note that Ainsworth et al. (2022) propose to propagate either the identity ( $\mathbf{T}_{\text{out}} = \mathbf{I}$ ) or the residual transportation map itself ( $\forall i, \gamma_i^{(l)} = 1$ ). In the case of hard alignment, these methods perform worse than averaging. + +The attention mechanism (Eq. 3) poses multiple challenges when it comes to TM flow (Fig. 2): what are the incoming TMs for $\mathbf{W}^Q$ , $\mathbf{W}^K$ and $\mathbf{W}^V$ ? Which TM is propagated to $\mathbf{W}^O$ ? How to handle attention with multiple heads? + +Self-Attention( +$$\mathbf{x}$$ +) = softmax( $\frac{\mathbf{Q}\mathbf{K}^{\mathbf{T}}}{\sqrt{d_k}}$ ) $\mathbf{V}$ , with { $\mathbf{Q}$ , $\mathbf{K}$ , $\mathbf{V}$ } = $\mathbf{W}^{\{\mathbf{Q}, \mathbf{K}, \mathbf{V}\}}\mathbf{x}$ (3) + +The first challenge is conveniently solved by the TM flow graph. We can simply use the TM from the previous layer for each $\mathbf{W}^Q$ , $\mathbf{W}^K$ and $\mathbf{W}^V$ . This even holds true for multiple heads. The incoming TM of $\mathbf{W}^Q$ is more complex to obtain because it depends on the outgoing TMs of $\mathbf{W}^Q$ , $\mathbf{W}^K$ , and $\mathbf{W}^V$ . However, if we constrain both TMs of $\mathbf{W}^K$ and $\mathbf{W}^Q$ to be equal permutation matrices (i.e., hard alignment with $\mathbf{T}_Q = \mathbf{T}_K = \mathbf{T}_{QK}$ ), we show that the permutation matrices cancel (see Eq. 4) leaving the softmax undisturbed. Therefore, we only propagate the outgoing TM of $\mathbf{W}^V$ to $\mathbf{W}^Q$ . + +For soft-alignment Eq. 4 no longer holds, in that case we investigated alleviating the constraint of equal TMs for $\mathbf{W}^K$ and $\mathbf{W}^Q$ . Removing this constraint slightly increased one-shot accuracy. + +$$\widetilde{\mathbf{Q}} = \mathbf{Q} \mathbf{T}_{QK}$$ + and $\widetilde{\mathbf{K}} = \mathbf{K} \mathbf{T}_{QK}$ and $\widetilde{\mathbf{Q}} \widetilde{\mathbf{K}}^{\top} = \mathbf{Q} \mathbf{T}_{QK} \mathbf{T}_{QK}^{\top} \mathbf{K}^{\top} = \mathbf{Q} \mathbf{K}^{\top}$ (4) + +For multi-head attention fusion, there is an additional layer of complexity because one must align the weights and the heads. On top of that, there is no guarantee that a hard one-to-one alignment between heads exists. For that reason, we propose cross-head alignment. During cross-head alignment, $\mathbf{W}_i^V$ , $\mathbf{W}_i^K$ and $\mathbf{W}_i^V$ (where i is the + +![](_page_4_Figure_13.jpeg) + +Figure 2: Self-Attention flow graph. + +head index) are concatenated across the output dimension to form three combined weight matrices $(\mathbf{W}^Q, \mathbf{W}^K \text{ and } \mathbf{W}^V)$ . OTFusion is then applied to each of the concatenated weight matrices. Finally, $\mathbf{T}_V$ is propagated to $\mathbf{W}^O$ . Find a visualization of our cross-head alignment method in App. B. + +The layer normalization is a learnable neural network parameter and consequently must be fused. It contains only two parameters ( $\alpha$ and $\beta$ ) per input and there are no interconnections between different inputs and outputs. Therefore, no TM has to be computed for this layer. The parameters are only aligned w.r.t. to the incoming TM. The incoming TM is then propagated to the subsequent layer. + +The ViT embeddings fusion approach is most effectively conveyed by its TM flow graph, as depicted in Fig. 3. For the concatenation, we notice that the class token is only a small fraction of the full sequence, in other words, for the integrity of the sequence, it is far more important to propagate the TM of the patch embeddings than the one for the class token. After concatenation, the positional embeddings are added. We notice that the addition is the same operation as for residual connections, so we can use one of the three TM flow strategies from Sec. 4.2.1. + +The bias is only connected to the output of a neural network layer, so we align it using the outgoing TM of the corresponding layer. + +Figure 3: ViT embeddings flow graph. + +Soft vs Hard Alignment. OTFusion technically allows soft alignment for MLPs, CNNs and ResNets, but Singh & Jaggi (2020) discovered that for these simpler architectures, hard alignment outperforms soft alignment. However, we do not want to limit the search space for optimal alignment to only permutation matrices (possibly too constraining for a complex architecture such Transformers). We, therefore, broaden the perspective on alignment introduced by OTFusion using the Sinkhorn algorithm and tuning the softness of the TM by optimizing over the Sinkhorn regularizer, discovering that soft alignment outperforms hard alignment for Transformers. + +Weights vs. activations alignment. The combined methodology introduced so far, and the novel perspective on the TM flow, allow us to apply OTFusion to the single fully connected layers without further adaptations in the case of weight-based approach, while the activation-based strategy needs a bit more thought. Transformers operate on sequences of tokens as opposed to simpler architectures that only operate one token at a time. In our activations-based algorithm, we treat every token of the sequence as a possible activation. + +Sequence Filtering. For ViTs, it is obvious that not every token contributes equally to the final image classification. We hypothesize that activations-based alignment performs best if only the most important tokens of a sequence are considered. Therefore, we explored filtering out unimportant tokens. For datasets where images are centered, we propose window filtering, where only the *n* by *n* center patches are considered as activations for activations-based alignment (window *n*). Additionally, we explored using only the class token for activation-based alignment (only cls). diff --git a/2311.07884/main_diagram/main_diagram.drawio b/2311.07884/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..b42b5b13b7e2baefe1a1c1f9063923b3b08478db --- /dev/null +++ b/2311.07884/main_diagram/main_diagram.drawio @@ -0,0 +1,160 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/2311.07884/main_diagram/main_diagram.pdf b/2311.07884/main_diagram/main_diagram.pdf new file mode 100644 index 0000000000000000000000000000000000000000..2c4eab33357a6f5cf09a52cda8a87acd01e3ba54 Binary files /dev/null and b/2311.07884/main_diagram/main_diagram.pdf differ diff --git a/2311.07884/paper_text/intro_method.md b/2311.07884/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..62392f89301fcc068285d4ac60f58f6f8a919d6f --- /dev/null +++ b/2311.07884/paper_text/intro_method.md @@ -0,0 +1,100 @@ +# Introduction + +Different social and demographic groups of people hold diverse and potentially even conflicting perspectives and opinions, which can be expressed in user-generated text data in domains and topics such as product reviews, social debates, healthcare, law, and politics [@shandilya2018fairness; @brazinskas2020few; @zhang-etal-2016-conversational; @huang2023embrace]. We formalize fair summarization of diverse perspectives as generating a fair summary for user-generated data by providing an accurate and comprehensive view of various perspectives from these groups. For example, Figure [1](#fig:example){reference-type="ref" reference="fig:example"} shows the reviews about an Amazon product. The summary generated by GPT-3.5 [@Ouyang2022instructgpt] unfairly represents sentiments as it summarizes more content from positive reviews than negative, ignoring negative groups and thus misleading the reader. + +However, existing summarization metrics cannot adequately measure fairness. On the one hand, the summarization metrics such as ROUGE [@lin-2004-rouge] and BERTScore [@bert-score] fail to measure fairness for two reasons. First, they are not inherently designed for quantifying and measuring fairness directly, but rather measuring similarity between system outputs and references based on n-gram overlapping or embeddings. Second, the reference summaries can also be biased. On the other hand, current work on summarization [@shandilya2018fairness] and LLM fairness evaluation [@chang2023survey] has not explored fair abstractive summarization which is more dominating than extractive summarization with the development of LLMs [@openai2023gpt4]. + +In this work, we present the first study that systematically defines, quantifies, and benchmarks fairness in abstractive summarization using large language models in a wide range of domains and topics. First, we define fair abstractive summarization that takes into consideration a social attribute, such as gender, and aim to generate summaries that are "fair" with respect to this attribute. We curate a benchmark [PerspectiveSumm]{.smallcaps}(Table [\[tab:fairsumm_data\]](#tab:fairsumm_data){reference-type="ref" reference="tab:fairsumm_data"}) by unifying and cleaning six existing datasets in domains where unfair summarization is a significant issue including healthcare, politics, restaurant, product, law, and policy, which covers social attributes of gender, party, sentiment, rating, and speaker. The fairness of the summary on these attributes greatly influences the information and even the opinions of the readers. Next, we quantify fairness in abstractive summarization as a distribution alignment of the generated summary and the source text. The distribution is calculated by ratios of tokens written by different groups of people (e.g. tweets written by male/female). According to the definition, we propose several metrics to evaluate fairness, including Binary Unfair Rate (BUR) which measures the overall fairness in abstractive summarization by counting the number of fair summaries in the whole dataset, Unfair Error Rate (UER) which measures the distance between generated summaries and source text by computing how target distribution differs from source's. Third, we benchmark fairness in abstractive summarization. We conduct a comprehensive evaluation of five main-stream LLMs, including GPTs (text-davinci-003, gpt-turbo-3.5, gpt-4), Alpaca, and Claude on [PerspectiveSumm]{.smallcaps}. + +Our results show that GPT-4 shows the least number of unfair summary based on our definitions. However, even GPT-4 [@openai2023gpt4] suffers from fairness problems as the BUR score for GPT-4 on Amazon dataset reaches 74.27%. We also analyze the common factors that can influence fairness, including summary length, perspective ratio [@shandilya2018fairness; @shandilya2020fairness], and temperature [@guo2017calibration]. Then we discovered three simple but effective ways to increase the fairness of the summaries including changing temperature, changing generated length, and appending the definition of fairness to the prompt. The result shows the BUR decreases from 51.11% to 47.48% on GPT-turbo-3.5 [@Ouyang2022instructgpt] by providing instructions of fairness definition. + +Our contributions are as follows: (1) We propose new metrics for quantifying and measuring the fairness of abstractive summarization over diverse perspectives. (2) We collect and unify a comprehensive benchmark for analyzing fairness in abstractive summarization in various domains and topics. (3) We conduct comprehensive experiments to investigate the fairness of reference summaries and summaries generated by popular LLMs finding that both reference/generated summaries fail to achieve the fairness. (4) We analyze the factors that influence fairness. We also propose three simple but effective methods to improve fairness. + +::: table* +::: + +In this section, we first formulate our task of summarization from diverse perspectives with social attributes (Section [2.1](#sec:task){reference-type="ref" reference="sec:task"}), then present data collection of [PerspectiveSumm]{.smallcaps} for this task (Section [2.2](#sec:data){reference-type="ref" reference="sec:data"}). + +# Method + +A fair abstractive summary should include comprehensive perspectives without underrepresenting certain groups of people. We use social attributes to indicate the properties that form groups of people. As shown in Table [\[tab:fairsumm_data\]](#tab:fairsumm_data){reference-type="ref" reference="tab:fairsumm_data"}, social attributes can be gender, party, and sentiment. + +After defining social attributes, we give the formal definition of summarization with social attributes. + +[]{#def:2.1 label="def:2.1"} + +A summarization dataset $\mathcal{D}$ consists of pairs of source text and target summary $(\bm{x}, \bm{y})$. Here, $\bm{x}=\{x_1,x_2,\cdots,x_n\}$ is the source text consisting of $n$ token $x_i$ and $\bm{y}=\{y_1,y_2,\cdots,y_m\}$ is the target summary consisting of $m$ tokens $y_i$. We consider a social *attribute* $a$ having $r$ different values $\mathcal{V} = \{v_1, v_2, \cdots, v_r\}$, $r>1$. Furthermore, we use $v(\cdot)$ to denote the value of a token. We assume each source unit $x_i$ belongs to one value $v(x_i) \subset \mathcal{V}$, $|v(x_i)|=1$, while each token in target is associated with one or multiple values $v(y_i) \subseteq \mathcal{V}$, $|v(y_i)|\geq 1$. For instance, when considering gender as an attribute $a$ in Twitter review summarization, the $r=2$ possible values are $\mathcal{V}=\{\texttt{male}, \texttt{female}\}$. For each tweet token $x_i$ in the source, if it is written by a male, $v(x_i) = \texttt{male}$, otherwise $v(x_i) = \texttt{female}$[^1]. For each token $y_i$ in the summary of tweets, it could come from $\texttt{male}$, $\texttt{female}$, or both. For instance, if both party says a product is "good". Then "good" can be both from $\texttt{male}$, and $\texttt{female}$. + +To conduct a comprehensive analysis of fairness, we select and then process existing datasets to form our fair summarization benchmark [PerspectiveSumm]{.smallcaps}. We follow two main principles to choose six datasets from a large variety of existing datasets. First is the *quality*, the source texts need to be human-written and marked with clear, and precise social attributes. These attributes are existing meta-data in the dataset except that we use a classifier for sentiment classification. Second is *diversity*, the collected datasets need to cover various domains and perspectives. + +Table [\[tab:fairsumm_data\]](#tab:fairsumm_data){reference-type="ref" reference="tab:fairsumm_data"} shows the statistics of [PerspectiveSumm]{.smallcaps}. It covers six different domains with six attributes. It is worth noting that, two dialogue datasets contain more than one attribute, which emphasizes that people can have diverse perspectives of fairness on the same input source text. Details of these six datasets are in Appendix [9](#sec:data_construct){reference-type="ref" reference="sec:data_construct"}. + +
+ +
Overview of our proposed metrics. Source/Target Dist. indicates source/target value distribution.
+
+ +In this section, we provide definitions for fairness and propose metrics to evaluate it. Figure [2](#fig:metrics){reference-type="ref" reference="fig:metrics"} shows the logic of computation of fairness in abstractive summarization and corresponding evaluation metrics. The first stage is to produce the value distribution. Given a sample pair of source and target text, we first split the source according to the annotated values of social attributes. Then, we apply Rule/BERT/BART scores to compare target tokens with source tokens to obtain the value of each token in the summary. Next, we count the number of tokens for each value in the source and target and obtain the value distribution of them. + +Computation of source and target value distribution (source/target distribution for short) is the first step of fairness computation. The formal definition is listed as follows. + +We define the value distribution $\bm{p}$ in three scenarios: (1) the value distribution in source: $\bm{p}_{\bm{x}} = [\bm{p}_{\bm{x}}(v_1), \bm{p}_{\bm{x}}(v_2),\cdots,\bm{p}_{\bm{x}}(v_r)]$, where $\bm{p}_{\bm{x}}(v_k)$ is the proportion of source tokens associated with value $v_k$, $0\leq \bm{p}_{\bm{x}}(v_k) \leq 1$, $\sum_{k=0}^r \bm{p}_{\bm{x}}(v_k) = 1$. (2) Similarly, the value distribution in target: $\bm{p}_{\bm{y}} = [\bm{p}_{\bm{y}}(v_1), \bm{p}_{\bm{y}}(v_2),\cdots,\bm{p}_{\bm{y}}(v_r)]$. (3) a gold value distribution $\bm{p}_{\bm{g}} = [\bm{p}_{\bm{g}}(v_1),\bm{p}_{\bm{g}}(v_2), \cdots, \bm{p}_{\bm{g}}(v_r)]$ which means the distribution that aligned with user's requirement (Section [3.3](#sec:define_fair){reference-type="ref" reference="sec:define_fair"}). + +To compute $\bm{p}_{\bm{x}}$, since the meta-data of the dataset shows the values of each token in the source, it can be acquired by counting the number of tokens of each value. However, the calculation of $\bm{p}_{\bm{y}}$ cannot be obtained easily due to the abstract nature of summaries. As abstractive summarization does not directly copy from the source text, it contains words that are not in the source but still semantically come from the source. This makes it difficult to attribute a sentence in the summary to a certain part of the source. To tackle this, we propose two algorithms for calculating the target distribution $\bm{p}_{\bm{y}}$: + +For a given k, we scan each k-gram in target text, if this k-gram is also shown in the source text, we assign this k-gram the value of the source text it appears. It is worth noting that the k-gram in the target can be assigned multiple values according to the definition. These metrics can automatically compute $\bm{p}_{\bm{y}}$ given $\bm{p}_{\bm{x}}$, with no additional annotation on $\bm{p}_{\bm{y}}$. We use $k=1$ for all experiments. In this paper, we use RuleScore to represent this matching approach. + +We use BARTScore [@yuan2021bartscore], and BERTScore [@bert-score] to measure the distance between the target and source. These metrics capture the similarity in semantics, overcoming the new word matching issues in the n-gram algorithm [@banerjee-lavie-2005-meteor]. Specifically, for a given source $\bm{x}$ and target $\bm{y}$, we first split $\bm{x}$ by value. For each $v_i \in \mathcal{V}$, $\bm{x_i} = \{x \in \bm{x} | v(x) = v_i\}$. Let $r = |\mathcal{V}|$. Then, we compute the correlation between $\bm{x_i}$ and $\bm{y}$ to obtain $\bm{p}_{\bm{y}}$: $$\begin{equation} +\bm{p}_{\bm{y}} = \text{Softmax}(\text{Score}(\bm{x_1},\bm{y}),\cdots, \text{Score}(\bm{x_{r}}, \bm{y})) +\end{equation}$$ Where $\text{Score}$ is BARTScore and BERTScore, and the temperature of $\text{Softmax}$ is a hyperparameter. We pick 0.1 for experiments because it better aligns with human evaluation (Appendix [11.2](#sec:meta-eval){reference-type="ref" reference="sec:meta-eval"}). + +The design of the metrics is different in terms of the following two aspects. First, they have different granularity. The N-gram score is based on the token which controls the fairness of the token level. BERT score computes the similarity of the sentences which is in sentence level. BART score computes via the entailment of the entire summary. This is on the summary level. Second, these scores have different advantages. Rule score can be applied to diverse length source text while the other two scores have length limitations. BERT score can capture semantic similarity while BART score captures entailment. These two models rely on the accuracy of the backbone models as well. + +After defining the source/target distribution, the goal of fair summarization is equivalent to producing a $\bm{p}_{\bm{y}}$ not underrepresenting any value in $\bm{p}_{\bm{x}}$. + +We define unfair summary as exhibiting underrepresentation of user groups in summaries. Given value distributions, we can define two types of fairness: (1) **Ratio Fairness**: the target value distribution $\bm{p}_{\bm{y}}$ shall follow the same value distribution as source $\bm{p}_{\bm{x}}$. In this case, $\bm{p}_{\bm{g}} = \bm{p}_{\bm{x}}$. (2) **Equal Fairness**: the target value distribution $\bm{p}_{\bm{y}}$ shall follow the uniform value distribution $\bm{p}_{\bm{g}} = [1/r,1/r,\cdots,1/r]$, regardless of source. This distribution keeps the balance of each value. However, users can define any $\bm{p}_{\bm{g}}$ to indicate their own idea distribution, not restricted to Ratio Fairness or Equal Fairness. We mainly discuss ratio fairness in the experiment section because summarization aims to compress but still follow the value distribution in the source text [@shandilya2018fairness]. + +As shown in Figure [2](#fig:metrics){reference-type="ref" reference="fig:metrics"}, we define four types of metrics to evaluate whether the generated summaries are fair. First, given the source and target distribution and definition of fairness summary. Binary Unfair Rate gives direct fair-or-not binary unfair classification results. + +We define Binary Unfair Rate (BUR), a binary function that outputs 1 if the sample is unfair; and 0 otherwise. To be specific, we define the summary as fair if and only if $\bm{p}_{\bm{y}}(v_k) \geq +\tau \cdot \bm{p}_{\bm{x}}(v_k)$ holds for each value $k = 1, 2, \cdots, r$, where $0 \leq \tau \leq 1$ is a tolerance hyperparameter that is chosen based on application scenarios [@shandilya2018fairness]. Otherwise, if $\bm{p}_{\bm{y}}(v_k) < +\tau \cdot \bm{p}_{\bm{x}}(v_k)$, we say the value $v_k$ is underrepresented. Thus $$\begin{equation} +\small +f_{\text{BUR}}(\bm{x},\bm{y};\tau) = \mathbbm{1}\left(\bigvee_{i=1}^r \bm{p}_{\bm{y}}(v_k) < \tau \cdot \bm{p}_{\bm{\bm{x}}}(v_k)\right) +\end{equation}$$ + +This metric is either zero to one and it can be averaged over the dataset, demonstrating the proportion of summaries that are *unfair* in the dataset. However, this hard metric may not be able to describe which one is more unfair/fair if BUR of two samples is the same. Thus, we propose Unfair Error Rate as a soft metric to compute the sum of the distance to source distribution of each value. + +This metric defines a function $f_{\text{UER}}(\bm{x},\bm{y})$ measuring the distance between $\bm{p}_{\bm{x}}$ and $\bm{p}_{\bm{y}}$. UER serves as the supplementary metric for BUR as it shows more fined-grained details compared with BUR. $$\begin{equation} +f_{\text{UER}}(\bm{x},\bm{y}) = \frac{1}{r}\sum_{k=1}^r \max(0, \bm{p}_{\bm{y}}(v_k) -\bm{p}_{\bm{x}}(v_k) ) +\end{equation}$$ + +UER shows the average percentage of values that are underrepresented. Furthermore, since people have different tolerance and requirements for fairness, we propose Area Under the Curve to compute the expected result of BUR when the tolerance is randomly picked from zero to one. + +As mentioned, $\tau$ is a hyperparameter chosen according to the scenarios. However, people may have different assignments of $\tau$ according to their own experiences. To incorporate different tolerance hyper-parameters, we propose AUC of fairness threshold (AUC for short) for BUR metric $f_{\text{AUC}}(\bm{x},\bm{y})$. $$\begin{equation} +f_{\text{AUC}}(\bm{x},\bm{y}) = \int_{0}^{1} f_{\text{BUR}}(\bm{x},\bm{y};\tau) \; d\tau +\end{equation}$$ + +It is a number between zero and one showing the expected BUR before assigning a fairness threshold $\tau$ to it because we assume that the users evenly pick the threshold between zero and one. It is worth noting that, practically, we sample ten points evenly to approximate the integration. + +These three metrics cannot show which value is more unfair in one sample. Thus, we further propose Second-Order Fairness to compute the variance of UER among all values. + +First, we define UER set $S_\text{UER}$ as $$\begin{equation} +S_\text{UER} =\{\max(0, \bm{p}_{\bm{g}}(v_k) -\bm{p}_{\bm{x}}(v_k) )\ | k \in [r]\} +\end{equation}$$ Second-Order Fairness (SOF) measures the unfairness difference among all values, defined as: $$\begin{equation} +\label{eq:SOF} +f_{\text{SOF}}(\bm{x},\bm{y}) = \frac{1}{r}\sum_{s \in S_\text{UER}} \left|s- \frac{\sum_{s \in S_\text{UER}}s}{r}\right| +\end{equation}$$ + +Inspired by @macqueen1967some, SOF measures the coherence of set $S_{\text{UER}}$ via computing the average distance to the center $\sum_{s \in S_\text{UER}}s /r$. + +If all values' UER are close to their averaged center, it indicates the generated summary leans towards certain values. In this case, we say that each value has a similar possibility to be unfair, so-called second-order fairness. + +**Tolerance Hyperparameter.** Regarding $\tau$, the threshold that decides whether a value is under-represented is based on the application scenarios [@celis2018fair]. When strict fairness is required, such as the political opinions of two parties, this threshold should be set to a high value, for instance, $\tau = 0.8$ or higher [@shandilya2018fairness]. However, in some other text summarization tasks in our daily lives, we do not require each value to have strict constraints, such as product review summarization. In the extreme case, we only want each value to appear in the summary rather than constraining the time of appearance, so we can set $\tau = 0$. In this paper, we set $\tau = 0.8$ for all experiments [@biddle2017adverse], if not specified. + +**Overall Score.** The score of a metric $\text{M} \in \{\text{BUR}, \text{UER}, \text{AUC}, \text{SOF}\footnote{SOF computes $S_\text{UER}$ on $\mathcal{D}$ first. Then apply Equation~\ref{eq:SOF}. }\}$ on a dataset $\mathcal{D}$ is the average corresponding metrics $f_\text{M}$ on each sample: $$\begin{equation} +\text{M} = \frac{1}{|\mathcal{D}|}\sum_{(\bm{x},\bm{y}) \in \mathcal{D}} f_\text{M}(\bm{x},\bm{y}) +\end{equation}$$ + +::: table* +::: diff --git a/2311.16937/main_diagram/main_diagram.drawio b/2311.16937/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..9d0ee1ec632c3b5a74e4b8ceb1ceada85cfe7c4a --- /dev/null +++ b/2311.16937/main_diagram/main_diagram.drawio @@ -0,0 +1,52 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/2311.16937/main_diagram/main_diagram.pdf b/2311.16937/main_diagram/main_diagram.pdf new file mode 100644 index 0000000000000000000000000000000000000000..12bbe0812ed86a1eef2113564efaf33721b0cf3e Binary files /dev/null and b/2311.16937/main_diagram/main_diagram.pdf differ diff --git a/2311.16937/paper_text/intro_method.md b/2311.16937/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..203c1ce284ba15e6049efc8fc3a9f71e4d2b1c9b --- /dev/null +++ b/2311.16937/paper_text/intro_method.md @@ -0,0 +1,75 @@ +# Introduction + +Inverse rendering of outdoor scenes has diverse downstream applications such as scene relighting, augmented reality, game asset generation, and environment capture for films and virtual production. However, accurately estimating the underlying scene model that produced an image is an inherently ambiguous task due to its ill-posed nature [2]. To address this, many works use some combination of handcrafted [7, 17] or learned priors [4, 6, 19, 47], inductive biases in model architectures [13], or multi-stage training pipelines [35, 36, 50]. This process is made even more difficult when considering in-the-wild image collections from the internet that contain transient objects, image filters, unknown camera parameters and changes in illumination. + +Outdoor scenes present particular challenges. Natural illumination from the sky is complex and exhibits an enormous dynamic range. This causes strong cast shadows when the brightest parts of the sky are occluded. These occlusions are non-local and discontinuous making them hard to incorporate within a differentiable renderer. Outdoor scene geometry can also exhibit arbitrary ranges of scale. On the other hand, sky illumination dominates secondary bounce lighting, meaning it is reasonable to assume a spatially non-varying, distant illumination environment. In addition, natural illumination contains statistical regularities [14] that make it easier to model. For example, luminance generally increases with elevation (the 'lighting-from-above' prior), the sun can only be in one position and the range of possible colours from sun and sky light is limited. + +In this paper, we tackle the outdoor scene inverse rendering problem by fitting a neural scene representation to a multi-view, varying-illumination photo collection. We name our method NeuSky, and make four key contributions relative to prior work. First, we make a key observation: Any pixel in an image that observes the sky provides a direct constraint on the illumination environment in that direction. Second, we combine this insight with an HDR neural field natural illumination model [16] learnt from natural environments, constraining this model to outpaint plausible illuminations given the direct observations of illumination seen from the camera. Thirdly, we propose outside-in visibility, a novel, differentiable, neural approximation to sky visibility, computed with a single forward pass through a directional distance function network. Finally, we deploy this visibility representation to enable end-to-end training, removing the need for phased training. Crucially, this means that shadows can influence illumination and geometry estimation by appearance losses backpropagating through the visibility network, enabling geometry estimation for non-observed scene regions and also avoiding shadow baking into albedo. + +# Method + +Our method takes as input a dataset of N images. From these images we compute poses with COLMAP [31, 32] and semantic segmentation maps with ViT-Adapter [10] according to the Cityscapes [11] convention. The preprocessed dataset comprises $\mathcal{D} = \{(\mathcal{I}_i, \mathbf{E}_i, \mathbf{K}_i, \mathcal{S}_i)\}_{i=1}^N$ , where $\mathcal{I} \in \mathbb{R}^{H \times W \times 3}$ is an image, $\mathcal{S} \in \mathbb{Z}^{H \times W}$ is the segmentation map and $\mathbf{E} = [\mathbf{R}|\mathbf{t}] \in \mathbb{R}^{3 \times 4}$ and $\mathbf{K} \in \mathbb{R}^{3 \times 3}$ are the camera extrinsics and intrinsics respectively. To align the vertical axis of our scene with gravity, we robustly fit a plane to the camera positions and rotate to align with the x-y plane. + +Scene Representation We model scene geometry as a neural SDF, such that at any point $\mathbf{x} \in \mathbb{R}^3$ , the signed distance is given by $f_{\text{SDF}}(\mathbf{x}) \in \mathbb{R}$ . We assume that the scene is Lambertian, with diffuse albedo modelled by the neural field $\mathbf{a}(\mathbf{x}) \in [0,1]^3$ . We also assume that illumination is a distant environment that depends only on direction $\mathbf{d} \in S^2$ , with HDR RGB incident radiance given by $L_i(\mathbf{d}) \in \mathbb{R}^3_{>0}$ . + +**Rendering** We follow NeuS [40] and derive a volume density, $\sigma(\mathbf{x})$ , from the SDF value. This allows volume rendering of the SDF in the same fashion as in NeRF. For a ray $\mathbf{r}$ with origin $\mathbf{o}$ and direction $\mathbf{v}$ , the time-discrete volume rendered RGB colour is given by: + +$$\mathbf{c}(\mathbf{r}) = \sum_{j=1}^{S} w_j \mathbf{a}(\mathbf{x}_j) \sum_{k=1}^{D} L_i(\mathbf{d}_k) V(\mathbf{x}_E, \mathbf{d}_k) \max(0, \mathbf{n}(\mathbf{x}_j) \cdot \mathbf{d}_k), \tag{1}$$ + +where the first summation is over the S samples along the ray, while the second is over the D lighting direction samples. The lighting direction samples are distributed approximately uniformly over the sphere by using an 8-subdivided icosahedron giving D = 642. $w_j$ is the volume rendering blending weight for the jth sample point which depends on $t_{1...j}$ and $\sigma(\mathbf{x}_{1...j})$ , with $\mathbf{x}_j = \mathbf{o} + t_j \mathbf{v}$ . $V(\mathbf{x}, \mathbf{d}) \in \{0, 1\}$ is the sky visibility in direction $\mathbf{d}$ at position $\mathbf{x}$ with $\mathbf{x}_E$ being the position at the expected termination depth of the ray $\mathbf{x}_j$ . $\mathbf{n}(\mathbf{x}) = \nabla f_{\mathrm{SDF}}(\mathbf{x}) / \|\nabla f_{\mathrm{SDF}}(\mathbf{x})\|$ is the surface normal at $\mathbf{x}$ , derived from the gradient of the SDF. + +We define our appearance loss for a batch of rays $\mathcal{R}$ as: + +$$\mathcal{L}_{app} = \sum_{\mathbf{r} \in \mathcal{R}} \ell(\mathbf{c}_{gt}(\mathbf{r}), sRGB(\mathbf{c}(\mathbf{r}))), \tag{2}$$ + +where $\mathbf{c}_{\mathrm{gt}}(\mathbf{r})$ is the ground truth colour for ray $\mathbf{r}$ , $\mathrm{sRGB}(\cdot)$ tonemaps the linear image provided by our model and $\ell$ computes the sum of L1 and cosine errors (to match both absolute RGB values and hue). To avoid overfitting we apply a random rotation $R \sim \mathcal{U}(SO(3))$ to jitter the direction vectors $\mathbf{d}_k$ in every batch. + +**Neural Illumination Model** To restrict $L_i$ to the space of plausible natural illumination environments, we use a neural illumination prior, RENI++ [16]. This is a conditional neural field, $f_{L_i}: S^2 \times \mathbb{R}^{3 \times K} \to \mathbb{R}^3$ that outputs log + +HDR RGB colours in the given input direction, conditioned on a normally distributed 3D latent code $\mathbf{Z} \in \mathbb{R}^{3 \times K}$ , $\text{vec}(\mathbf{Z}) \sim \mathcal{N}(\mathbf{0}_{3K}, \mathbf{I}_{3K})$ . The latent space of RENI++ provides a low dimensional characterisation of natural, outdoor illumination environments since it was trained on several thousand real-world outdoor environment maps. This provides useful global constraints on the estimated illumination, which is only partially observable in any one image. In addition, the normally-distributed latent space provides a prior while the latent code is vertical-axis rotation-equivariant (rotating **Z** about the vertical axis corresponds to similarly rotating the environment). This vertical axis corresponds to gravity and we therefore align the vertical axis of our scene with gravity as described above. We optimise a RENI++ latent code, $\mathbf{Z}_i$ , and absolute scale, $\gamma_i$ , for each image i in the training set and replace $L_i(\mathbf{d}_k)$ with $\gamma_i \exp(f_{L_i}(\mathbf{d}_k, \mathbf{Z}_i))$ in (1). To ensure the estimated illumination is plausible, we include a prior loss: $\mathcal{L}_{\text{prior}} = \|\mathbf{Z}\|_2^2$ for all latent codes. In Section 3.1 we describe how the illumination environment in an image can be additionally constrained via sky pixel observations. + +Reducing Visibility Tests Visibility of the illumination environment from a scene point is required in our rendering equation (1) and is essential for recreating cast shadows and ambient occlusion effects. However, computing sky visibility from a neural SDF is computationally expensive. It requires sphere tracing from the query point in the light direction until the ray hits another part of the surface or leaves the scene bounds. To render a single pixel, this must be performed D times for each of the S sample points. We therefore propose two methods to drastically reduce the number of visibility samples. First, since we are only concerned with visibility on the surface of the scene, we define $\mathbf{x}_E = \mathbf{o} + t_E \mathbf{v}$ , where $t_E$ is the current expected termination depth of the ray, and evaluate visibility only at $\mathbf{x}_E$ . This means we only need D visibility tests per pixel since we reuse the computed visibilities for all sample points along the ray. Second, any light direction in the lower hemisphere, i.e. where $(\mathbf{d}_i)_z < 0$ , will strike either the scene or the ground. For these directions we set $V(\cdot) = 1$ , i.e. visible. The rationale for this is that the RENI++ illumination environment will learn to capture the colour of the ground or lower hemisphere of the scene, averaged over all spatial positions. This provides an approximation to secondary illumination from the ground. We found this to perform considerably better than setting these directions as non-visible. In spite of these two speedups, the remaining D/2 visibility tests still prove too expensive if performed via sphere tracing of the SDF. For this reason, in Section 3.2 we propose a fast, softened approximation for visibility. + +Pixels labelled in the semantic segmentation maps with the 'sky' class (hereafter referred to as *sky pixels*) provide a direct observation of the distant illumination environment in the direction given by the ray for that pixel. To the best of our knowledge, this constraint has never been used to aid illumination estimation + +in inverse rendering methods. Since our illumination model, RENI++ [16], captures the space of plausible natural illuminations, even observing only a portion of the sky provides a strong statistical cue. For example, if a bright region corresponding to the sun is observed, then RENI++ cannot create another sun in an unobserved part of the environment. Alternatively, if all observed sky is white, it is likely to be an overcast day and RENI++ will predict an ambient environment without a discernible sun. Using sky pixel constraints alone can be viewed as statistical outpainting of the whole environment from the portion observed in an image. In practice, we incorporate this within our inverse rendering framework such that the appearance loss of non-sky pixels also provides a rich, indirect constraint on the illumination. + +The sky segmentation also provides an additional constraint that is similar to the widely used mask loss. Since we know that sky ray pixels miss the scene, we penalise our neural scene representation from placing any density along the ray, providing geometric supervision. Together, these form our sky loss: + +$$\mathcal{L}_{\text{sky}} = \sum_{\mathbf{r} \in \mathcal{R} \cap \mathcal{S}_{\text{sky}}} \varepsilon(\mathbf{c}_{\text{gt}}(\mathbf{r}), \mathbf{c}_{\text{sky}}(\mathbf{r})) - \log(1 - \sum_{j} w_{j}), \tag{3}$$ + +where $S_{\text{sky}}$ is the set of sky pixels. The first term is the error between the observed sky pixel colour and predicted, $\mathbf{c}_{\text{sky}}(\mathbf{r}) = \text{sRGB}(\gamma \exp(f_{L_i}(\mathbf{r}, \mathbf{Z})))$ , and the second term is the binary cross entropy loss on the accumulated density in sky pixels. + +Shadows offer a wealth of information about geometry, both within and beyond the view frustum. For instance, if the sun is predicted to be behind the camera and a prominent cast shadow appears on the floor, we can infer there is geometry behind the camera and the likely sun direction. However, to fully leverage this information it is necessary to have a differentiable model of visibility. + +To address this, we draw inspiration from works, NeRFactor [50] and NeRV [35] and learn a neural model of visibility. However, to make training tractable, [50] learn their visibility representation in a second training phase with geometry pretrained and frozen and [35] require known illumination. Initial attempts to model visibility using the same parameterisation as [35] were unable to fit in our less constrained and end-to-end task. We desire a model of visibility that is consistent with the geometry of our scene, fast to sample from and differentiable, enabling gradients from visibility to inform illumination, albedo and geometry estimation. However, this model must be constrained enough that training end-to-end with our scene representation is tractable. To achieve this, we propose outside-in visibility in which visibility is represented implicitly via a Spherical Directional Distance Field (SDDF) defined on the radius 1 sphere that bounds our scene and is tied to our SDF scene representation via consistency losses. Our geometric volume is represented with the Mip-NeRF 360 [3] scene contraction. This means that parallel rays converge to a point on the radius 2 sphere + +(representing infinity). Hence, our visibility model resides on the radius 1 sphere where position-dependent visibility can be reasoned about, while our distant illumination model is defined on the radius 2 sphere (see Figure 3). + +Spherical Directional Distance Function Consider a point $\mathbf{s} \in S^2$ lying on a bounding sphere of radius 1. The Spherical Directional Distance Function (DDF), $f_{\rm DDF}: S^2 \times$ $S^2 \to \mathbb{R}$ , returns the (positive) distance from s for any inward-pointing direction $\mathbf{d}$ to the first intersection with the surface. In other words, the spherical DDF stores an inward looking depth map of the scene from any viewpoint on the radius r sphere. The DDF is related to the SDF: $f_{SDF}(s +$ $f_{\rm DDF}(\mathbf{s}, \mathbf{d})\mathbf{d}) = 0$ , such that moving the distance given by the DDF must arrive at the surface where the SDF is zero. However, there may be multiple such points and the DDF must return the minimum, giving us another constraint: $f_{\text{DDF}}(\mathbf{s}, \mathbf{d}) = \min\{t | f_{\text{SDF}}(\mathbf{s} +$ $t\mathbf{d}) = 0$ . + +The DDF is required to learn a very complex function: essentially an + +![](_page_7_Picture_5.jpeg) + +Fig. 3: We model our illumination and illumination visibility via two spherical neural fields at radius $\infty$ and 1 respectively. However our world space is contracted as per Mip-NeRF-360 [3], such that any point at infinity is placed on the sphere of radius 2. Since we model distant illumination, the sampled colour only depends on direction, and two samples at different locations but in the same direction will sample RENI++ [16] at the same point. However, visibility of distant illumination is dependent on location and the intersection of the ray on the sphere of radius 1 is used to sample our visibility network. + +inward-facing depth map of the scene from any position on the sphere. We found that this function is easier to learn if we define a consistent coordinate frame to parameterise directions for any given point on the sphere. We normalise the inward-facing directions from world coordinates to a local coordinate system such that the y-axis aligns with s (the sample position on the DDF), the x-axis is orthogonal to y and to our world-up, and the z-axis is orthogonal to y and x. See Figure 7 for a visualisation. + +Sky Visibility via Directional Distance Fields Our key insight is to show how to use the inward looking DDF as a representation for computing outward sky visibility (see Figure 4). Consider a point $\mathbf{x} \in \mathbb{R}^3$ lying on the surface (and inside the bounding sphere, such that $\|\mathbf{x}\| \leq 1$ ). We can use the DDF to check whether $\mathbf{x}$ can see the sky or is occluded in a direction $\mathbf{d}$ . First we compute the point $\mathbf{s}$ as the solution to $\mathbf{s} = \mathbf{x} + t\mathbf{d}$ , s.t. $\|\mathbf{s}\| = 1$ and $t \geq 0$ , i.e. the point on the radius r sphere that is intersected by the ray in direction $\mathbf{d}$ from $\mathbf{x}$ . Next, we evaluate the DDF at $\mathbf{s}$ in direction $-\mathbf{d}$ (i.e. outside-in): $f_{\text{DDF}}(\mathbf{s}, -\mathbf{d})$ . If $\mathbf{x}$ is not occluded then the DDF value should be similar to the actual distance between $\mathbf{s}$ and $\mathbf{x}$ : $f_{\text{DDF}}(\mathbf{s}, -\mathbf{d}) \approx \|\mathbf{s} - \mathbf{x}\|$ . However, if $\mathbf{s}$ is occluded then the DDF will return a distance significantly less than the actual distance: $f_{\text{DDF}}(\mathbf{s}, -\mathbf{d}) < \|\mathbf{s} - \mathbf{x}\|$ . Binary visibility can be computed by testing whether this difference is below + +a threshold ϵ: V = (∥s − x∥ − fDDF(s, −d) < ϵ). Note that this is equivalent to classical shadow mapping [42] with the exception that we rely on a DDF forward pass as opposed to (non-differentiable) rasterisation of a mesh from the light source perspective. + +However, binary visibility is discontinuous and so not suitable for propagating loss gradients through visibility and back into geometry. For this reason, we replace the discrete threshold with a softened approximation (see Figure 5): + +$$V(\mathbf{x}, \mathbf{d}) = 1 - \kappa \left( \eta(\|\mathbf{s} - \mathbf{x}\| - f_{\text{DDF}}(\mathbf{s}, -\mathbf{d}) - \epsilon) \right), \tag{4}$$ + +where κ is the sigmoid function. The threshold ϵ controls the tolerance on what is considered a shadow. We make this learnable and initialise it with a large value (equal to the scene radius). When ϵ is large, no parts of the scene will be considered occluded. As training converges, ϵ can be reduced to gradually introduce more illumination occlusions. The parameter s controls the sharpness of the transition between occluded and unoccluded. + +Supervising the DDF The DDF indirectly determines visibility which in turn determines appearance via the rendering equation in (1). This means that the DDF is partially supervised by the appearance loss. However, we also require that the DDF's representation of scene geometry is consistent with the SDF geometry. We enforce this consistency through four losses. First, Lddf\_depth, enforces that the depth predicted by the DDF should match that of the scene parameterised by the SDF. Second, Lddf\_levelset, ensures that travelling the distance predicted by the DDF should arrive at the SDF zero level set. Third, we encourage multiview consistency in the DDF via a multiview consistency loss Lddf\_multiview. Finally, with, Lddf\_sky, we further take advantage of our sky segmentation maps as an additional constraint on our DDF. Rays that intersect the sky have no occlusions between the camera origin and our DDF sphere. Our DDF should therefore predict at least the distance to + +![](_page_8_Picture_7.jpeg) + +Fig. 4: Visibility of our neural illumination from a point in the scene is implicitly represented via our Directional Distance Field (DDF) which represents the depth to the surface of our scene from any point on the unit sphere. The DDF is a spherical neural field that surrounds our scene at radius 1. The DDF is fully differentiable allowing gradients obtained from shadowing to inform illumination and geometry. + +the camera origin for those intersecting rays. Detailed descriptions of these losses can be found in the supplementary. diff --git a/2312.06093/main_diagram/main_diagram.drawio b/2312.06093/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..54f1e4b31f5bcbb9899ecbe658d5a7dc6f83d008 --- /dev/null +++ b/2312.06093/main_diagram/main_diagram.drawio @@ -0,0 +1,152 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/2312.06093/paper_text/intro_method.md b/2312.06093/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..b60a14571782bd9248cf7681b6febf2330291a42 --- /dev/null +++ b/2312.06093/paper_text/intro_method.md @@ -0,0 +1,138 @@ +# Introduction + +Motivation. The proliferation of social media has given rise to a new form of communication: memes. Memes are multimodal and often contain a combination of text and visual elements that convey meaning, humor, and cultural significance. Understanding the topics and themes of memes is important for a variety of applications, including content moderation, social media analysis, and cultural studies. Although meme analysis has been an active area of research for some time, with studies focusing on various aspects such as sentiment analysis [17], semantics [13, 35], abusive content [23, 26], and cultural significance [5, 11], there has been little work done on unsupervised multimodal topic modeling of memes. This represents a significant research gap to address, as memes have become an essential form of communication, and understanding their topics and themes is crucial for various applications. + +A significant challenge in performing unsupervised multimodal topic modeling of memes lies in comprehending the semantic interaction between the text and visual modalities. For instance, in Figure 1, if we consider the superimposed text and image separately, they could be associated with topics such as "Marriage" and "Animals" respectively. However, when we consider both modalities together, the appropriate topic should be "Humor" or "Relationship Humor". Hence, it is essential to consider both text and visual elements together to extract meaningful topics from memes. Despite this, most existing topic modeling techniques are predominantly text-based [37] and are not designed to handle multimodal data. + +**Research Objectives.** To address these research gaps, we propose PromptMTopic1, a novel multimodal prompt-based model designed to learn topics from both text and visual modalities by leveraging the powerful language modeling capabilities of large + + $^{1}https://github.com/Social-AI-Studio/PromptMTopic \\$ + +![](_page_1_Picture_2.jpeg) + +Figure 1: Example of a multimodal meme from FHM dataset. + +language models (LLMs). Specifically, PromptMTopic first utilizes a visual language model to extract descriptions of the visual aspect of the meme. These visual captions are then combined with the superimposed text extracted from the memes and used as input to a series of carefully constructed prompts. These prompts guide the LLM in identifying and grouping relevant topics in memes. + +Contributions. We summarize our contributions as follows: + +- We propose a novel multimodal prompt-based model to perform topic modeling on memes. To the best of our knowledge, this is the first multimodal topic modeling model that effectively extracts and clusters topics learned from memes. +- We evaluate our proposed model through extensive experiments on three real-world meme datasets. Our automatic evaluations demonstrate that our proposed model outperforms the state-of-the-art topic modeling baselines in learning descriptive topics in memes. +- We also perform a qualitative analysis of the learned topics and show that our model can identify meaningful and culturally relevant topics from memes. + +# Method + +The problem of multimodal meme topic modeling involves identifying common topics within a collection of memes, using both visual + +![](_page_2_Figure_2.jpeg) + +Figure 2: PromptMTopic Model. + +and textual data. The problem can be defined as follows: Given a collection of n memes $D=m_1,m_2,\ldots,m_n$ , with each meme m=i,l comprising an underlying image i and superimposed text l, the goal is to generate the latent topics and their associated concept probabilities. Traditionally, topic modeling techniques compute a score for each meme-topic pair and topic-word pair through matrix decomposition or the computation of statistical distributions. + +Existing statistical and embedding-based topic modeling techniques require textual data for matrix decomposition or for the computation of statistical distributions. Therefore, to apply these techniques to memes, it is necessary to (1) extract the superimposed text l and (2) convert the image i into an acceptable textual representation. Superimposed text can be extracted automatically using existing libraries such as EasyOCR or keras-ocr, or by human annotators. Regarding the visual element, a common approach for image-totext conversion involves representing the image's semantics with a textual description, or image captioning. Hence, we cleaned the superimposed text from the images and input the resulting images into a pre-trained image captioning model. For this task, we used the BLIP-2 model, which has demonstrated its ability to generate high-quality captions for crowdsourced images by leveraging the implicit knowledge in frozen large language models such as Open Pre-trained Transformers [40]. Using BLIP-2, we were able to generate captions that effectively describe the dominant objects or events depicted in the meme's image. + +Our PromptMTopic model is an unsupervised approach that leverages the strong language understanding capabilities of ChatGPT for topic generation. Recent studies have demonstrated that ChatGPT achieves state-of-the-art performance on many natural language + +tasks [20, 32, 33]. Notably, we observed that ChatGPT2 excels at understanding informal language, including internet jargon and slang. To perform the topic modeling task with ChatGPT, we designed two distinct prompts that extract relevant topics and consolidate a multitude of generated topics into a unique set of themes. Each of these prompts is structured as a multi-turn conversation, where prior turns serve as task demonstrations that guide the model's generation. This process aligns with the concept of *in-context learning*, where large language models are prompted with instructions and/or demonstrations to solve new tasks without additional training [8]. This approach enables the model to learn and perform the task efficiently without the need for fine-tuning. Figure 2 provides an overview of our PromptMTopic model, which consists of two stages: *Topic Generation* and *Topic Collapse*. + +Table 1 shows an example of a prompt for topic generation. To guide the behavior of ChatGPT ("assistant"), we meticulously crafted a system message that instructs the assistant to identify and extract high-level topics. Moreover, we instructed the assistant to provide only unique topics and to avoid overly verbose details. We then introduced N conversation turns, containing prompt inputs and expected answers, for the model to comprehend and perform the task. For our experiments, we chose N=8 conversation turns as there were noticeably fewer hallucinations with this setting. Finally, we fed the prompt input for the new sample and obtained the topics for individual memes. + +However, we observed two scenarios where ChatGPT sometimes failed to generate topics. First, ChatGPT may not comprehend the sample meme due to unclear or incoherent textual inputs. These occurrences are typically because the meme has too few superimposed words, which does not provide enough context for ChatGPT to infer from. Second, the content moderation layer in the ChatGPT API detects inappropriate inputs, such as profanity, and prevents + +&lt;sup>2https://api.openai.com/v1/chat/completions + +Table 1: Example ChatGPT's prompt for topic generation (reformatted for visualisation purposes). The prompt input consists of the extracted textual features (in yellow), the question (in green), the demonstration answers provided (in purple) and the ChatGPT's generated answers (in pink) + +**System**: You are designated as an assistant that identify and extract high-level topics from memes. You should avoid giving specific details and provide unique topics solely. + +**User:** Please list the high-level topics in the following meme using its image caption and superimposed text. + +Meme's Image Caption: president obama in front of the white house Meme's Superimposed Text: i did not divide the country the republican decision to obstruct every single thing i proposed to help us dig out of the financial crisis they caused divided the country Topics: + +Assistant: ['Politics', 'Financial Crisis'] + +**User**: Please list the high-level topics in the following meme using its image caption and superimposed text. + +Meme's Image Caption: a man in a car with the caption, when you're in the car and you see in the car + +Meme's Superimposed Text: recruit: why tf gotta wear my id tag and bring my FAD everywhere? BMT spec: IT'S the LAW. + +Topics: + +Assistant: ['Military'] + +ChatGPT from generating topics. To address these situations, we classified these memes into "Miscellaneous" and "Inappropriate" topic clusters, respectively. + +While ChatGPT generates multiple high-level topics for individual memes, many of these topics appear similar and overlap. For instance, topics such as "occupation", "job search", "labour" and "employment" can be consolidated under the umbrella term "employment". To address this issue, we propose two approaches for grouping similar topics: *Prompt-Based Matching (PBM)* and *Word Similarity Matching (WSM)*. + +**Prompt-Based Matching.** To group similar topics from a collection of memes, we prompt ChatGPT to combine unique topics until a specified number of topics, denoted by K, is reached. Initially, we sort the unique topics in descending order based on their respective frequency counts, creating the set $T_n = t_1, t_2, \ldots, t_n$ . The rationale behind this is that topics with higher occurrence rates are likely to be more distinctive. Subsequently, we select the first n-1 topics from $T_n$ to form the set $T_{n-1} = t_1, t_2, \ldots, t_{n-1}$ . We then prompt ChatGPT to merge the current topic $t_n$ into a topic within the set $T_{n-1}$ . If ChatGPT fails to merge $t_n$ with any of the topics in $t_n$ , we incorporate $t_n$ into the "Miscellaneous" topic. We repeat this process until we reduce $t_n$ to the set $t_1, t_2, \ldots, t_K$ . See Table 2 for an example of a topic grouping prompt. + +Table 2: Example ChatGPT's prompt for topic collapse (reformatted for visualisation purposes). The prompt input consists of the extracted textual features (in yellow), the question (in green), the demonstration answers provided (in purple) and the ChatGPT's generated answers (in pink) + +**System:** You are designated as an AI assistant that determine the most appropriate overarching topic from a given list of topics that best corresponds to a current topic. Specifically, you are required to identify the most fitting topic from the provided list that is associated with the given topic, without providing any topics that are not included in the given list of topics. + +User: Topics: + +1. Military , ... , 6. Healthcare + +Current topic : War + +What is the most suitable topic from the given topics that corresponds to the current topic? + +Assistant: ['Military'] + +User: Topics: $[t_1, t_2, ..., t_K]$ + +Current topic : $t_{K+1}$ + +What is the most suitable topic from the given topics that corresponds to the current topic? + +**Assistant**: $t \in \{t_1, t_2, ..., t_K\}$ + +In our experiment, we encountered an issue with some datasets having a large number of unique topics, which exceeded the maximum token length allowed by the ChatGPT API. To overcome this, we used a sliding window of M topics from the sorted unique topic set and performed topic grouping iteratively. If ChatGPT successfully merged a topic into one of the M topics, we merged the topic and restarted the iteration cycle. However, if ChatGPT failed to merge the current topic into any topic in the unique topic list, we assigned the topic to a category named "Miscellaneous". + +**Word Similarity Matching**. In order to consolidate similar topics into a specified number of unique topics, denoted by K, we calculate the similarity between pairs of topics. Specifically, we aggregate the documents belonging to each topic and compute the Class-based Term Frequency - Inverse Document Frequency (c-TF-IDF) word representation: + +$$W_{x,c} = t f_{x,c} \log(1 + \frac{1+C}{1+df_{x,c}})$$ + (1) + +where $tf_{x,c}$ represents the frequency of word x in class c, $df_{x,c}$ represents the number of classes in the document set that contain the word x, and C represents the total number of classes in the document set. The resulting representation captures the frequency + +Table 3: Example ChatGPT's prompt for topic representation words (reformatted for visualisation purposes). The prompt input consists of topic and the corresponding top 100 words (in yellow), the demonstration answers provided (in purple) and the ChatGPT's generated answers (in pink) + +**System**: You are designated as an AI assistant that given a list of words and a topic, determines the top 10 words related to the topic, without providing any words that are not included in the above list of words. + +Words: 1. singapore, 2. tracetogether, ..., 16. message ... + +Given the above topic, list the top 10 words related to the topic. + +**Assistant:** ['computer', 'robot', 'google', 'iphone', 'airpods', 'internet', 'tracetogether', 'video', 'tiktok', 'message'] + +User: Topic : tWords : $[w_1, w_2, ..., w_k]$ + +Given the above topic, list the top 10 words related to the topic. + +**Assistant**: $[s_1, s_2, ..., s_{10}]$ + +of words in each topic, weighted by their importance across all topics. We then select the top 20 words from the c-TF-IDF representation for each topic. To measure the similarity between topics, we count the number of common words between the top 20 words of each topic pair. If a topic has fewer than 20 words, we normalize the number of common words by the smaller word count between the two topics. This way, we identify topics that share common words and phrases, and group them together to form a smaller set of unique topics. We can express the topic similarity computation using the following formula: + +$$TopicSimilarity(t_i, t_j) = \frac{t_i \cap t_j}{\min(t_i, t_j)}$$ + (2) + +where t represents a topic, which consists of a bag of topic-related words. Once we have calculated the topic similarity scores for all possible pairs of topics, we merge the topic pair that has the highest word similarity to form a new topic. We repeat this process iteratively until only K unique topics remain, where K represents the desired number of final topics. + +To evaluate the efficacy of our proposed PromptMTopic model, we assess the generated topics using standard topic model metrics, as detailed in Section 5. Evaluation necessitates that topics be represented as a word mixture. Given our model did not supply the required topic-word distributions for selecting the top W representative words, we computed the c-TF-IDF for each cluster and employed two methods to retrieve representative words. In the + +Table 4: Dataset statistics: Size represents the number of memes in each dataset, Caption represents average length of BLIP-2 caption per meme, Text represents average length of superimposed text per meme. + +| | | Avg. Length | | | +|--------------|--------|-------------|-------|--| +| Dataset | Size | Caption | Text | | +| TotalDefMeme | 2,513 | 10.75 | 17.98 | | +| FHM | 10,000 | 8.34 | 11.54 | | +| Memotion | 6,992 | 8.44 | 14.84 | | + +PromptMTopic-WSM approach, we select the top 10 words with the highest c-TF-IDF scores for each topic, under the assumption that a word's importance is closely tied to the topic's meaning. In the PromptMTopic-PBM approach, we prompt ChatGPT to return the top 10 words related to each topic, feeding it the top 100 words arranged in descending order of their c-TF-IDF scores. Table 3 provides an example of the word ranking prompt using ChatGPT. + +Using these two approaches, we can evaluate the effectiveness of our methodology and the coherence of the predicted topics. diff --git a/2312.09788/paper_text/intro_method.md b/2312.09788/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..3507b928e568c52ce598b563bbc097947266d9b7 --- /dev/null +++ b/2312.09788/paper_text/intro_method.md @@ -0,0 +1,23 @@ +# Introduction + +Deep Neural Networks have showcased remarkable ability in scene understanding tasks like Semantic Segmentation (SS) [\[7,](#page-8-0) [71,](#page-10-0) [73\]](#page-10-1), when the training and test distribution are the same. This dependency reveals a significant vulnerability: their performance substantially diminishes when encountering domain shifts [\[67\]](#page-10-2), highlighting a fundamental challenge in generalizing these networks to unseen domains [\[43,](#page-9-0) [46,](#page-9-1) [62\]](#page-10-3). To address this, Domain Generalized Semantic Segmentation (DGSS) has gained prominence [\[28,](#page-9-2) [31,](#page-9-3) [33,](#page-9-4) [83\]](#page-11-0). DGSS aims to develop models that leverage a source-annotated dataset while remaining effec- + +![](_page_0_Figure_9.jpeg) + +Figure 1. Performance over time by various methods on the GTA → {Cityscapes, BDD, Mapillary} benchmark. Recent openvocabulary approaches, like FC-CLIP, are shown to excel in zeroshot learning and surpass traditional domain generalization methods trained in closed-set scenarios, challenging the relevance of the DGSS setting. CLOUDS, by harnessing multiple foundation models, demonstrates its ability to effectively utilize the source dataset, thereby outperforming both conventional DGSS and openvocabulary methods. + +tive across unseen domains, thus overcoming the limitations of traditional DNNs in handling unseen environments. + +Recently, the advent of large-scale pretrained models, often referred to as Foundation Models (FMs) [\[2,](#page-8-1) [35,](#page-9-5) [56,](#page-10-4) [60,](#page-10-5) [68\]](#page-10-6), have brought a paradigm shift in computer vision tasks. The FMs comprise of contrastively trained image classification models (e.g., CLIP [\[56\]](#page-10-4), ALIGN [\[30\]](#page-9-6)), text-conditioned generative models (e.g., Stable Diffusion [\[60\]](#page-10-5), DALL-E [\[57\]](#page-10-7)), vision transformer-based segmentation model trained on mammoth dataset (e.g., Segment Anything Model (SAM) [\[35\]](#page-9-5)), to name a few. Of particular interest to SS, SAM is a promptable segmentation model that accepts as input an image and geometric prompts (points, scribbles, boxes, masks) and outputs classagnostic masks. Owing to its pre-training on billion-scale dataset [\[35\]](#page-9-5), SAM has demonstrated excellent performance on varied out-of-distribution tasks namely, medical imaging + +![](_page_1_Figure_0.jpeg) + +Figure 2. Qualitative comparison at inference: (a) Input image, (b) SHADE, a traditional style diversification DGSS method, (c) SAM, a foundation model that predicts precise class-agnostic maps, (d) GroundingSAM that leverages SAM and text-prompts to output semantic maps, and (d) Our proposed CLOUDS that leverages an assembly of foundation models to predict high-quality semantic maps. + +[\[22,](#page-8-2) [44,](#page-9-8) [72\]](#page-10-8), crater detection [\[19\]](#page-8-3), and so on. In summary, large-scale pre-training holds promise for DGSS as robustness is greatly improved when a model is trained on large datasets that cover various possible distributions [\[80\]](#page-11-1). + +While the FMs present an excellent solution for addressing DGSS, they have not been adopted so far in the literature. A great majority of DGSS methods rely on the technique of Domain Randomization (DR) [\[55,](#page-10-9) [69,](#page-10-10) [75,](#page-10-11) [83\]](#page-11-0), where the goal is to diversify the labeled source domain images by photometric and geometric image transformations. Examples of such transformations include style diversification [\[83,](#page-11-0) [84\]](#page-11-2), adversarial style augmentation [\[84\]](#page-11-2), and difficulty-aware photometric augmentations [\[31\]](#page-9-3), among others [\[33,](#page-9-4) [79\]](#page-11-3). Albeit effective to some extent, these techniques can not bring content augmentation, resulting in sub-optimal performance (see Fig. [2b\)](#page-1-0). Moreover, adopting SAM for the task of DGSS is not straightforward as it outputs class-agnostic masks, making it unsuitable for DGSS (see Fig. [2c\)](#page-1-0). While there are recent works, such as Grounding-SAM [\[35,](#page-9-5) [41\]](#page-9-7), that equip SAM with semantic predictions, it misses out on several important objects (*e.g.*, "poles") and *stuff* classes (*e.g.*, "tree", "sidewalk") (Fig. [2d\)](#page-1-0). The lack of semantic awareness can be attributed to ambiguous text prompts or poor vision and text alignment. These results indicate that DGSS is far from solved and accommodating FMs in DGSS is an open question. + +Driven by the motivation of narrowing down the research gap, in this work we propose to amalgamate an *assembly* of FMs into a cohesive system to address DGSS. In detail, we propose to exploit: (i) strong representations of contrastive FMs, (ii) diverse content augmentation using generative FMs, and (iii) near accurate class-agnostic mask prediction of SAM and turn them into semantic predictions. Concretely, for the segmentation network we use CLIP [\[56\]](#page-10-4) as a backbone, serving as a robust feature extractor. To introduce content diversification, we generate synthetic images using a combination of Large Language Models (LLMs) [\[68\]](#page-10-6) and text-to-image Diffusion Model (DM) [\[60\]](#page-10-5). The LLM generates task-aware diverse textual prompts which then guide the DM in producing photorealistic synthetic images. Since the generated images are unlabeled, we employ self-training strategy [\[24,](#page-8-4) [81,](#page-11-4) [85,](#page-11-5) [86\]](#page-11-6) where we use the pseudo labels from the teacher network to train the student model. However, as the pseudo labels from the teacher can be noisy, we leverage SAM's excellent class-agnostic mask predictions to refine the pseudo labels. As shown in Fig. [2e,](#page-1-0) the pseudo-label refinement greatly assists in improving the reliability of self-training. + +In summary, in this work we make the following contributions: (i) We highlight for the first time in DGSS the importance of content diversification, which is more effective than traditional DGSS methods relying on style diversification; (ii) We propose CLOUDS, a system of collaborative Foundation models for DGSS. We run extensive experiments on several DGSS benchmarks and demonstrate its effectiveness. As shown in Fig. [1,](#page-0-0) our FM-based DGSS method outperforms the traditional DGSS and openvocabulary segmentation models by a non-trivial margin. diff --git a/2312.16023/main_diagram/main_diagram.drawio b/2312.16023/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..4f00119cee8fd8d62f57f7d5ab5691ae17f2188e --- /dev/null +++ b/2312.16023/main_diagram/main_diagram.drawio @@ -0,0 +1 @@ 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 \ No newline at end of file diff --git a/2312.16023/paper_text/intro_method.md b/2312.16023/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..d7275fe455a6812892dac0c70e8a40b91235b580 --- /dev/null +++ b/2312.16023/paper_text/intro_method.md @@ -0,0 +1,41 @@ +# Introduction + +Sarcasm is a form of verbal irony that often uses positive words to convey a negative message, such as frustration, anger, contempt and even ridicule [@WILSON20061722]. In real-world cases, a piece of sarcastic news often lacks explicit linguistic markers, and thus requires additional cues to reveal the true intentions. The accompanying visual information provides helpful cues to better perceive ironic discrepancies. Multimodal sarcasm [@wang-etal-2022-multimodal; @KaiShu2017FakeND] is omnipresent in social media posts, forum discussions, and product reviews, and hence the multimodal sarcasm understanding is of great significance for a wide range of applications in the news field such as sentiment analysis[@Mao_Shen_Yu_Cai_2021], fake news detection[@Ying2022BootstrappingMR; @Qi_Bu_Cao_Ji_Shui_Xiao_Wang_Chua_2023], and public opinion analysis. + +![Example of a piece of sarcastic news. It shows that a fire extinguisher, a tool used to extinguish fires, has caught fire. This sarcasm can remind us of the quality of the fire extinguisher.](imgs/example1.pdf){#fig:introexample width="76mm"} + +Figure [1](#fig:introexample){reference-type="ref" reference="fig:introexample"} illustrates a piece of multimedia sarcastic news. To understand this sarcastic news, a model must capture textual cues from multiple sentences, including *fire fighting* and *cause danger*. The accompanying visual cue, *fire on a fire extinguisher* in the figure below, plays an important role in this sarcasm. The figurative and creative nature of such multimodal sarcasm poses a great challenge to the effective perception of the true intention under the guise of overt positive surface involved in a whole document and an image. This requires the design and development of document-level multimodal sarcasm understanding (MSU) methods that specifically take the characteristics of such an ironic expression into consideration. + +Prior research has underscored the critical significance of utilizing extensive, high-quality, and challenging benchmarks for the development and evaluation of state-of-the-art deep learning methods across various natural language processing (NLP) tasks[@8269806; @10.1007/978-3-030-58577-8_8]. In this context, existing sarcasm benchmarks [@castro-etal-2019-towards; @wang-etal-2022-multimodal] have demonstrated considerable promise. However, when addressing document-level multimodal sarcasm understanding in the real-world news domain, they exhibit certain limitations, including (1) Limited length of text. In real-world scenarios, a piece of news may include more than $70$ words across multiple sentences [@KaiShu2017FakeND], concealing ironic discrepancies beyond sentence boundaries. However, samples in existing multimodal sarcasm datasets [@cai-etal-2019-multi; @castro-etal-2019-towards; @wang-etal-2022-multimodal] only include about $20$ words within a single utterance on average, which greatly simplifies the challenges of MSU in real-world cases. (2) Limited quality of annotations. Existing large satirical datasets [@riloff-etal-2013-sarcasm; @ptacek-etal-2014-sarcasm; @barbieri-etal-2014-modelling] are mostly generated by bootstrapping algorithm or remote supervision with noisy labels. These annotations can be disruptive to systems that harness such data for downstream applications due to the subtle nature of sarcasm. Furthermore, these datasets only contain text modality. (3) Very limited number of samples. As sarcasm lacks explicit linguistic or visual markers, a model requires a large volume of samples to learn the rules or ways that reveal the true underlying intentions. A large-scale dataset benefits the generalization capability of an MSU model that alleviates the over-fitting issue during the training procedure. + +The aforementioned limitations in existing datasets highlight the need for a comprehensive, challenging, and higher-quality document-level multimodal sarcasm dataset to enhance irony understanding in the domain of news. Towards that, we developed DocMSU, a comprehensive benchmark that contains high-quality annotations of $102,588$ pieces of news with text-image pairs, covering $9$ hot topics such as science, business, and sports. We collect these samples from social websites, including "New York Times" and "UN News", each involving $63$ tokens across $5$ sentences on average. To alleviate the ambiguity of sarcasm, we manually annotated these documents and images in $3$ rounds with $15$ workers, ensuring the annotation quality with confidence scores. Each pair of text-image involves a binary label for sarcasm detection, $2.7$ textual spans and visual bounding boxes on average for sarcasm localization. + +The proposed DocMSU facilitates the research of multimodal sarcasm perception for real-world applications. It also introduces two new challenges: (1) capturing the nuanced sarcastic clues in two modalities, where the clues are concealed within very few words in a document or a tiny area in an image; (2) aligning the visual and linguistic features for irony understanding, where the incongruity nature of sarcasm requires cross-modal interactions. To fill this gap, we propose a novel sarcasm comprehension method that aims to fuse the pixel-level image features with the word-level textual features of a whole document in a fine-grained manner. Experimental results show the effectiveness of our method. We will release our dataset and the code. The main contributions of our work can be summarised as follows: + +- We curate DocMSU, a new benchmark for document-level multimodal sarcasm understanding in the real-world news field. Compared with existing ones, our dataset is more comprehensive and more challenging with much higher quality annotations. + +- We come up with a novel document-level MSU method for sarcasm detection and localization, mitigating the issues in sarcastic cues detection across sentences and across modalities under inconsistent context. + +- We conduct extensive experiments on our DocMSU. Results show that the created benchmark enables us to develop and evaluate various deep learning methods for the task of MSU closer to the real-world application. + +# Method + +Figure [3](#fig:example){reference-type="ref" reference="fig:example"} shows two examples selected from our DocMSU. The first example highlights the irony in National Geographic's practices, as they extensively employ plastic packaging despite advocating against the overuse of plastic. The subsequent example exposes the contradictions within Netflix's service delivery. Despite boasting about their discs being "unbreakable\" and implying exceptional durability, customers received damaged discs, contradicting the advertised durability. + +A model may face two new challenges for sarcasm understanding the above two examples. (1) Capturing the nuanced sarcastic clues that are concealed within very few words (e.g., "*a broken disc*") in a document or in a very tiny area (e.g., the tag "*unbreakable*") of the image. (2) Aligning the visual and text features for the accurate irony understanding (e.g., "*unbreakable*" and the broken disc). Existing approaches such as recent MSTI [@wang-etal-2022-multimodal], CLIP, and VILT have limitations in tackling these two challenges as they focus more on learning the overall information of the whole text and image representations. This motivates us to develop a new method to capture the fine-grained linguistic and visual sarcastic clues and align the two different types of clues for a better MSU. + +Figure [4](#fig:model){reference-type="ref" reference="fig:model"} illustrates the architecture of our model, which mainly consists of three components, including a document encoder, an image encoder, and a fusion module. To capture the underlying subtle clues concealed within very few words in a document and a very tiny area in an image, our model generates two matrices for pixels-level image representations and token-level document representations. For the cross-modal interactions, we fuse the representations in two matrices with a sliding window for multimodal alignment. We will explore the specifics of this design in the following sections. + +We denote a document as $s = \{w_i\}_{i=1}^{n}$, where $w_i$ indicates the $i$-th token and $n$ is the total number. We use BERT [@devlin-etal-2019-bert] to output contextualized token-level representations $\mu \in R^{n \times d}$, where $$\begin{gather} +\mu = [\nu_1,\nu_2,...,\nu_n] = \mathrm{BERT}(s) +\end{gather}$$ We use a fully connected layer $f_c$ to transform word representations. Then, we convert the document representation into a square shape $\varpi \in R^{L \times L \times d},$ $$\begin{align} +\varpi(i,j,:) = \big(f_c(\theta)\big)_{L \times (i-1)+j} +\end{align}$$ where $1 \leq i,j \leq L$. We add paddings when $n < L \times L$. This square document representation is used for the fine-grained alignment of the pixel-level visual representations. + +To keep the high spatial resolution of the feature maps and retain the information of the image details, we only use the early three convolution layers of ResNet [@2016arXiv160305027H]. In this way, we can keep the original resolution of images. Then, we use a projection layer $f_p$ to generate the visual representations for each pixel. Third, the image feature map is spatially divided into $m$ sliding windows, with $L\times L$ pixels for each window. In this way, the representation of the entire image is as follows, $$\begin{align} + \omega = [\omega_{1},\omega_{2},...,\omega_{m}] = f_p\big((\mathrm{ResNet}(\sigma)\big), +\end{align}$$ where $\sigma$ is the input image and $\omega_{k} \in R^{L \times L \times d}$ denotes the $k$-th window. + +During the process of multimodal fusion, we add the document representation $\varpi$ to each window $\omega_{k}$. The result of addition is denoted as $\hat{\omega}_{k}$, $$\begin{align} +\hat{\omega}_{k} = \varpi + \omega_{k}\quad k = 0,1,...,m +\end{align}$$ We apply four stages in Swin-Transformer [@liu2021Swin] to deeply fuse the two modalities. Specifically, each stage contains one patch merging layer and several blocks containing mechanisms of shifted window attention, which calculates the attention between each element in each shifted window with little computational complexity. By doing so, interactions are built between each word of the document and each image pixel without adding additional calculations. The output of this method delivers multimodal fine-grained features that could be applied to sarcasm understanding. diff --git a/2401.13837/main_diagram/main_diagram.drawio b/2401.13837/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..2d791dc65f6a2bf0b37b4ac7d984d93e7029c498 --- /dev/null +++ b/2401.13837/main_diagram/main_diagram.drawio @@ -0,0 +1,408 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/2401.13837/paper_text/intro_method.md b/2401.13837/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..e48c332f74cdf34bc417ea6b778e2d73ca856de5 --- /dev/null +++ b/2401.13837/paper_text/intro_method.md @@ -0,0 +1,128 @@ +# Introduction + +\vspace{-.1in} +\lblsec{introduction} + +Fine-grained visual recognition (FGVR) is an important task in computer vision that deals with identifying subordinate-level categories, such as species of plants or animals . It is challenging due to the fact that different species of birds can differ in subtle attributes, such as the Lincoln's Sparrow mainly differs from Baird's Sparrow in the coloration of the breast pattern (see Fig. ). Due to small inter-class and large inner-class variations, the FGVR methods typically require auxiliary information namely part annotations , attributes , natural language descriptions collected with the help of experts in the respective fields. Hence, this expensive expert annotation presents as a bottleneck and prevents FGVR in new domains from being rolled out as software or services to be used by common users. + +On the contrary encyclopedia of textual information about plants and animals can be found on the internet that document at great lengths about the characteristics and appearance of each species. In other words, modern day Large Language Models (LLMs) (e.g., ChatGPT ), trained on enormous internet-scale corpuses, already encode in its weights the expert knowledge that is quintessential in FGVR. In this work we ask ourselves, can we build FGVR systems by exploiting the unpaired auxiliary information already contained in LLMs? The implications are tremendous as it endows an average layperson with the expert knowledge of a mycologist, ornithologist and so on. + +{R}{0.54\textwidth} +\centering +\vspace{-.1in} +\includegraphics[width=\linewidth]{figures/motivation.pdf} + \caption{Comparing our proposed \ours with the state-of-the-art visual question answering models: BLIP-2 , LLaVA , LENS , and MiniGPT-4 .} + + \vspace{-5mm} + +Leveraging unlabelled FGVR dataset and the LLM to build a FGVR model is not straightforward as there exists a modality gap between the two. In detail, a LLM cannot see, as it can only ingest text (or prompts) at input and give text at output. To bridge this modality gap, we first use a general purpose visual question answering (VQA) model (e.g., BLIP-2 ) to extract a set of visual attributes from the images. Specifically, for every image in a bird dataset we prompt BLIP-2 to extract part-level descriptions of wing color, head pattern, tail pattern and so on. We then feed these attributes and their descriptions to a LLM, and prompt it to reason on a set of candidate class names using its internal knowledge of the world. As a final step, we use the candidate class names to construct a semantic classifier, as commonly used in the vision-language models (VLMs) (e.g., CLIP ), to classify the test images. We call our method Fine-grained Semantic Category Reasoning (FineR) as it discovers concepts in unlabelled fine-grained datasets via reasoning with LLMs (see Fig. ). + +Through our proposed \ours system we bring together the advances in VQA, LLMs and VLMs to address FGVR that was previously not possible without accessing high-quality paired expert annotations. Since our proposed \ours does not require an expert's intervention, it can be viewed as a step towards democratizing FGVR systems by making them available to the masses. In addition, unlike many end-to-end systems our \ours is interpretable due to its modularity. We compare our method with FGVR, vocabulary-free classification and multi-modal language and vision assistant models (e.g., BLIP-2) and show its effectiveness in the challenging FGVR task. As shown in Fig. our \ours is more effective than the state-of-the-art methods, which are accurate at predicting the super-category but fail at identifying the subordinate categories. Unique to our \ours, it works in a few-shot setting by accessing only a few unlabelled images per category. This is a very pragmatic setting as fine-grained classes in real-world typically follow an imbalanced class distribution. Finally, our method is completely training-free, which is the first of its kind in FGVR. Extensive experiments on multiple FGVR benchmarks show that our \ours is far more effective than the existing methods. + +In summary our contributions are: (i) We introduce the challenging task of recognizing fine-grained classes using only a few samples per category, where neither labels nor paired expert supervision is available. (ii) We propose the \ours system for FGVR that exploits a cascade of foundation models to reason about possible subordinate category names in an image. (iii) With extensive experiments on multiple FGVR benchmarks we demonstrate that our \ours is far more effective than the existing approaches. We additionally introduce a new fine-grained benchmark of Pokemon characters and show that \ours exhibits the potential to work truly in the-the-wild, even for fictional categories. + +\vspace{-3mm} + +# Method + +\lblsec{overview} +\vspace{-3mm} +\noindent Problem Formulation. In this work, we study the problem of fine-grained visual recognition (FGVR) given an unlabeled dataset $\datatrain$ of fine-grained objects. Different from the traditional FGVR task we do not use any label or paired auxiliary information annotated by an expert , and $\datatrain$ contains a few unlabelled samples per category . Our goal is to learn a model that can assign correct labels to the images in the test set $\datatest$. In detail, when presented with a test image we output a semantic class name $c \in \classes$, where $\classes$ is the set of class names that are not known a priori, making it vocabulary-free . Our proposed \ours first aims to generate the set $\classes$ and then assigns the test image to one of the classes. As an example, opposed to the FGVR methods that output a class-index at inference, our \ours directly outputs a semantic class name (e.g., Blue-winged Warbler). Next we discuss some preliminaries about the pre-trained models that constitute our proposed \ours. + +\vspace{-.1in} + +\vspace{-.1in} + +\noindent \textbf{\llmtitle{s} (\llm{s})}, such as ChatGPT and LLaMA , are auto-regressive models that are trained on large internet-scale text corpuses possess a wealth of world knowledge encoded in their weights. These models can be steered towards a particular task by conditioning on demonstrations of input-output example pairs, relevant to the task at hand, without the need of fine-tuning the weights. This emergent behaviour is referred to as in-context learning (ICL) . The LLMs have also demonstrated adequate reasoning abilities in answering questions when prompted with well-structured prompts , another emergent behaviour that we exploit in this work. Formally, a \llm $\ctrlllm: \inllm \mapsto \outllm$ takes a text sequence (or \llm-prompt) $\inllm$ at input and maps it to a text output sequence $\outllm$. + +\noindent \textbf{Vision-Language Models (\vlm{s})}, such as CLIP and ALIGN , consist of image and text encoders that are jointly trained to predict the correct pairings of noisy image and text pairs. Due to well-aligned image-text representation space, they demonstrate excellent zero-shot transfer performance on unseen datasets, provided the class names of the test set is known a priori . A \vlm $\ctrlvlm$ maps an image $\mbf{x}$ to a category $\class$ in a given vocabulary set $\classes$ based on the maximum cosine similarity: + + \hat{\class} = \arg\max_{\class \in \classes} \langle \encimg(\mbf{x}), \enctxt(c) \rangle, + +where $\hat{c}$ is the predicted class, and $\encimg$ and $\enctxt$ denote the vision and text encoder of the VLM, respectively. $\langle \cdot, \cdot \rangle$ represents the cosine similarity function. + +\noindent \vqatitle (\vqa) models, such as BLIP-2 and LLaVA , that combine LLM and VLM in a single framework, learn to not only align the visual and text modalities, but also to generate the caption of the image-text pair. This visual-language pre-training (VLP) is then used to perform \vqa as one of the downstream tasks. In detail, at inference a +\vqa model $\ctrlvqa: (\invqaimg, \invqatxt) \mapsto \outvqa$ receives an image $\invqaimg$ and a textual question (or \vqa-prompt) $\invqatxt$ as inputs, and outputs a textual answer $\outvqa$ that is conditioned on both the visual and textual inputs. + +While these \vqa models excel at identifying super-categories and general visual attributes (e.g., color, shape, pattern) of objects when prompted with a question, they fail to recognize many subordinate-level categories (see Fig. ) due to lack of well-curated training data. We show how the aforementioned pre-trained models can be synergistically combined to yield a powerful FGVR system. + +\lblsec{method} + +To recall, our goal is to recognize the fine-grained categories using unlabelled images in $\datatrain$, without having access to expert annotations and the ground-truth categories in $\classes$. Thus, the challenges in such a FGVR task are to first identify (or discover) the classes in $\classes$ and then assign a class name $c \in \classes$ to each instance in $\datatest$. The idea underpinning our proposed \ourslong (\ours) is that the LLMs, which already encode the world knowledge about different species of animals and plants, can be leveraged to reason about candidate class names $\classespred^{*}$ when prompted with visual attributes extracted from the \vqa models. Concretely, +\ours takes a few unlabeled images in $\datatrain$ as input, and outputs the discovered class names $\classespred^{*}$. Subsequently, the discovered candidate class names and images in $\datatrain$ are utilized to yield a multi-modal classifier $\MW_\text{mm}$ to classify a test instance using a VLM. Therefore, our \ours aims to learn a mapping: + + \ctrlours: \datatrain \mapsto (\classespred^{*}, \MW_\text{mm}), + +where $\ctrlours$ is a meta-controller that encapsulates our system. \ours operates in three phases: (i) Translating Useful Visual Information from Visual to Textual Modality; (ii) Fine-grained Semantic Category Reasoning in Language; and (iii) Multi-modal Classifier Construction. An overview of our system is depicted in Fig. and the details follow next. + +[!t] + \centering + \includegraphics[width=\linewidth]{figures/framework-v3.pdf} + \caption{The pipeline of the proposed \ourslong (\ours) system.} + + \vspace{-5mm} + +Before we can tap into the expert world knowledge inherent to the LLMs for facilitating FGVR, we need a mechanism to translate the visual information from the images into an input format which the LLM understands. Furthermore, we require an understanding about which visual cues are informative at distinguishing two fine-grained categories. For instance, what separates two species of birds (Lincoln's Sparrow vs Baird's Sparrow) is drastically different than what separates two types of mushrooms (Chanterelle vs Porcini). Motivated by this we design the following steps (see Fig. top). + +\noindent Identify super-category. +To make our framework general purpose and free of any human intervention we first identify the super-category of the dataset $\datatrain$. In detail, we use a VQA model as a \vietitle (\vie) $\ctrlvqa_{\text{\vie}}$ for this purpose. The \vie is capable of identifying the super-category (e.g., ``Bird") of the incoming dataset as it has been pre-trained on large amounts of data from super-categories. Formally, to get super-category $g_n$ we query the \vie for an image $\mbf{x}_n \in \datatrain$: + +\vspace{-5mm} + + g_n = \ctrlvqa_{\text{\vie}}(\mbf{x}_n, \pptidentify), + +where $\pptidentify$ is a simple Identify \vqa-prompt of the format shown in Fig. , and $\supercat = \{g_1, g_2, \dots, g_N\}$ is a set containing the super-category names corresponding to $N$ images in $\datatrain$. To guide the \vie to output only super-category names we provide simple in-context examples as a list of names (``Bird'', ``Pet'', ``Flower'', ...). Note that the ground-truth super-category name need not be present in the list (see \refapp{prompt_details} for details about in-context examples). + +\noindent Acquire expert knowledge. After having a general understanding of the dataset this step consists in identifying the useful set of attributes that set apart the subordinate-level categories. For instance, the ``eye-color" attribute can be used to easily distinguish the bird species Dark-eyed Junco from the species Yellow-eyed Junco. To help the \vie in focusing on such targetted visual cues we tap into the expert knowledge of the LLMs, which is otherwise only restricted to experts. In other words, we essentially ask the LLM: given a super-category what to look for while distinguishing subordinate-level categories. In detail, we base our Acquire Expert Knowledge (\aek) module $\ctrlllm_{\aek}$ on a LLM that takes an input a super-category $g_n$ and outputs a list of $M$ useful attributes: + +\vspace{-5mm} + + \va_n = \ctrlllm_{\text{\aek}}(g_n, \ppthowto), + +where $\va_n = \{a_{n,1}, a_{n,2}, \dots, a_{n,M}\}$ denote the $M$ attributes corresponding to the sample with index $n$, and How-to \llm-prompt $\ppthowto$ is the prompt, as shown in Fig. . The intuition behind the How-to \llm-prompt is to emulate how to be an expert and forms the crux of our \ours. All the attributes in the dataset are denoted as $\attrset = \{\va_{1}, \dots, \va_{N}\}$. + +\noindent Perceive visual attributes. With the attributes $\mathcal{A}$ available we leverage the \vie to extract the description for each attribute $a_{n,m}$ conditioned on the visual feature. For instance, if an attribute $a_{n,m}$ = ``eye-color", then given the image of an unknown bird the \vie is asked to describe it's eye color, which is a much easier task compared to directly predicting the subordinate-level category. In addition, we hard-code a general attribute $a_{0} = "General description of the image"$ and its prompt "Questions: Describe this image in details. Answer:" to extract the global visual information. Therefore, $\va_{n} \leftarrow \va_{n} \cup a_{0}$ and $M \leftarrow M + 1$. We found this is helpful to capture the useful visual information from the background like the habitat of a bird. Concretely, given an image $\mbf{x}_n$, its super-category $g_n$ and the attribute $a_{n,m}$, the attribute description is given as: + + s_{n,m} = \ctrlvqa_{\text{\vie}} (\mbf{x}_n, g_n, a_{n,m}, \pptdescr), + +where $\pptdescr$ denotes the Describe \vqa-Prompt, as illustrated in Fig. , and $\descrset_n = \{s_{n,1}, s_{n,2}, \dots, s_{n,M}\}$ represents a set of $M$ attribute descriptions extracted from the image $\mbf{x}_n$. + +\vspace{-3mm} + +At the end of the first phase (Sec. ) we have successfully translated useful visual information into text, a modality that is understood by the LLM. In the second phase our goal is leverage a LLM and utilize all the aggregated textual information to reason about class names in the dataset and predict a candidate set $\classespred^{*}$, as illustrated in the middle part of Fig. . + +\noindent Reasoning class names. To discover the candidate class names present in the dataset we build a \scrtitle (\scr) module that contains a LLM under the hood. The goal of \scr to evoke reasoning ability of the LLM by presenting a well-structured prompt constructed with the attributes and descriptions obtained in Sec. . Formally, the \scr is defined by a function $\ctrlllm_{\text{\scr}}$ that outputs a preliminary set of candidate classes $\classespred$ as: + +\vspace{-5mm} + + \classespred = \ctrlllm_{\text{\scr}}(\supercat, \attrset, \descrset, \pptinfer), + +where $\pptinfer$ is the Reason LLM-prompt as shown in Fig. . The $\pptinfer$ can be decomposed into two parts: (i) Structured Task Instruction is designed keeping in mind the FGVR task and is mainly responsible for steering the LLM to output multiple possible class names given an image, thereby increasing diversity in names. We find that when only one name is asked, \llm{s} tend to output general categories names (e.g., ``Gull") from a coarser but more common granularity. (ii) Output Instruction embeds the actual attribute description pairs extracted from the image, and task execution template to invoke the output (see \refapp{prompt_details} for details). + +\noindent Noisy name refinement. In the pursuit of increasing the diversity of candidate class names we accumulate in the set $\classespred$ a lot of noisy class names. It is mainly caused by the unconstrained knowledge in LLMs that is not relevant to the finite concepts in the $\datatrain$ at hand. To remedy this, in addition to name deduplication, we propose a \nndtitle (\nnd) module that uses a VLM (e.g., CLIP). +Specifically, \nnd uses the text encoder $\enctxt$ and the vision encoder $\encimg$ of the \vlm to encode $\classespred$ and $\datatrain$, respectively. Each image in $\datatrain$ is assigned to a category $\classpred \in \classespred$ based on the maximum cosine similarity as in Eq. . The class names that are not selected by Eq. constitute the noisy class names set $\mathcal{V} \subset \classespred$ and are eliminated. The refined candidate set is given as $\classespred^{*} = \classespred \setminus \mathcal{V}$. + +Given the refined class names in $\classespred^{*}$ we construct a text-based classifier for each class $c \in \classespred^{*}$ using the text-encoder $\enctxt$ of the \vlm as: +\vspace{-2mm} + + \vw^c_\text{txt} = \frac{\enctxt(c)}{||\enctxt(c)||_2}. + +Since there can be ambiguity in some of the class names we also consider a vision-based classifier by using the images in $\datatrain$. In detail, we first pseudo-label the images in $\datatrain$ with $\classespred^{*}$ using the maximum cosine similarity metric, as outlined in Sec. . This results in $U_c$ pseudo-labelled images per class $c \in \classespred^{*}$. However, as $\datatrain$ contains only a few images per class it might make the visual features strongly biased towards those samples. For example, only a few colors or perspectives of a car of make and model Toyota Supra Roadster is captured in those images. Thus, we apply a random data augmentation pipeline $K$ times on each sample $\mbf{x}$ (see \refapp{data_augmentation} for details) and then compute the vision-based classifier for each class $c \in \classespred^{*}$ as: +\vspace{-2mm} + + \mbf{w}^c_\text{img} = \frac{1}{U_c(K+1)} \Biggl( \sum^{U_c}_{i=1} \frac{\encimg(\mbf{x}^c_i)}{||\encimg(\mbf{x}^c_i)||_2} + \sum^{K \times U_c}_{j=1} \frac{\encimg(\tilde{\mbf{x}}^c_j)}{||\encimg(\tilde{\mbf{x}}^c_j)||_2} \Biggl), + +where $\tilde{\mbf{x}}$ is the augmented version of $\mbf{x}$. We set $K=10$ in all our experiments (see \refapp{study_k} for an analysis on the sensitivity of $K$). + +Being equipped with classifiers of two modalities, we construct a multi-modal classifier (MMC) for class $c$ that potentially can capture complementary information from the two modalities: +\vspace{-0mm} + + \mbf{w}^c_\text{mm} = \alpha \mbf{w}^c_\text{txt} + (1 - \alpha) \mbf{w}^c_\text{img}, + +where $\alpha$ is the hyperparameter that controls the degree of multi-modal fusion. We empirically set $\alpha = 0.7$ in all of our experiments (see \refapp{study_alpha} for an analysis on the sensitivity of $\alpha$). Finally, the MMC for all the classes in $\classespred^{*}$ is given as $\MW_\text{mm}$. + +As shown in Fig. right, we replace the text-based classifier in \vlm with the MMC $\MW_\text{mm}$, and classify the test images $\mbf{x} \in \datatest$ by + + $\hat{c} = \arg\max_{c \in \classespred^{*}} \langle \encimg(\mbf{x}), \mbf{w}_\text{mm}^{c} \rangle$, + +where $\hat{c}$ is the predicted class name having the highest cosine similarity score. diff --git a/2402.10093/main_diagram/main_diagram.drawio b/2402.10093/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..bfc5f99467ade5bb7e3915c5e87e00baffd942c2 --- /dev/null +++ b/2402.10093/main_diagram/main_diagram.drawio @@ -0,0 +1 @@ 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 \ No newline at end of file diff --git a/2402.10093/main_diagram/main_diagram.pdf b/2402.10093/main_diagram/main_diagram.pdf new file mode 100644 index 0000000000000000000000000000000000000000..4fa478abb0baeff70c6fc0163eb9355601d373c9 Binary files /dev/null and b/2402.10093/main_diagram/main_diagram.pdf differ diff --git a/2402.10093/paper_text/intro_method.md b/2402.10093/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..d23cdf36aa0f8520ef340fb5ae5c2710b41a9393 --- /dev/null +++ b/2402.10093/paper_text/intro_method.md @@ -0,0 +1,86 @@ +# Introduction + +::: wrapfigure +r6.0cm ![image](res/timeline.pdf){width="6.0cm"} +::: + +Self-supervised learning (SSL) has attracted considerable attention, owing to its data efficiency and generalization ability [@liu2021self]. SSL leverages pre-training tasks and creates intricate input representations without the need for explicit supervision via expensive annotations. In computer vision, Masked Image Modeling (MIM) [@chen2020mim; @vincent2010mim; @pathak2016mim] has emerged as one of the prevalent SSL pre-training paradigms, enabling an efficient pretraining of large models on unlabeled data by reconstructing masked parts of the input images. + +The success of MIM is driven by methods like Masked Autoencoder (MAE) [@he2022mae], data2vec [@baevski2022data2vec; @baevski2023data2vec2], and others [@bao2021beit; @xie2022simmim]. For example, MAE opens the door for sparse pre-training of Vision Transformers (ViTs) [@dosovitskiy2021vit] by masking large parts of the image and not processing the masked areas. The computational efficiency, coupled with the data efficiency of a generative reconstruction task [@xie2023data; @el2021large; @singh2023maewsp] fosters the scaling to larger architectures on datasets of limited size. However, MIM models tend to spread their attention across the whole image [@walmer2023teachingmatters]. When adapting to downstream tasks, a sufficient amount of labels is required to rewire the attention to focus on important regions in the image. In the absence thereof, MIM models perform poorly. In contrast, for example, Instance Discrimination (ID) [@he2020moco; @chen2020simclr; @caron2020swav; @grill2020byol] methods implicitly focus on objects and form semantic clusters in the latent space [@dino2021caron], which eases adaption to downstream tasks in the absence of vast amounts of labels. In summary, the most important desiderata for efficient SSL pre-training methods in computer vision are rich representations of the input -- ideally in the form of semantic clusters in the latent space -- alongside efficiency in both compute and data, and, most notably, favorable scaling to larger architectures. + +
+
+ +
+
+ +
+
Left: Downstream evaluations of pre-trained SSL models. MIM-Refiner effectively combines their respective advantages of MIM and ID without suffering from their respective disadvantages. Right: Representation quality of SSL methods evaluated via linear probing. MIM-Refiner advances the state-of-the-art on ImageNet-1K pre-training in low- and high-compute regimes.
+
+ +In this work, we first analyze pre-trained MIM models, where we find that MIM models have different types of blocks: those that mainly improve the pre-training objective and others that are responsible for abstraction within the MIM encoder. The origin of this behavior can be traced back to the fact that MIM architectures usually comprise a large ViT encoder together with a *very* light-weight decoder. For larger models, the light-weight decoder reaches a point, where it cannot further improve the pre-training objective on its own and passes part of the reconstruction task back to the last encoder blocks. Consequently, the feature quality for downstream tasks of the later blocks degrades, and, somewhat unusual, the representation quality peaks in the middle blocks of the encoder. + +Based on these insights, we introduce MIM-Refiner, a simple -- yet highly effective -- sequential refinement approach tailored to MIM models. MIM-Refiner applies an ensemble of ID heads that enforce semantic clusters via an ID objective. Most importantly, those ID heads are attached to intermediate blocks of the encoder including those that exhibit peak representation quality, instead of only a single ID head attached to the last block. + +We introduce Nearest Neighbor Alignment (NNA) as a new ID variant. Given a batch of samples, NNA aligns each sample with its NN and repels each sample from all other samples in the batch. NNA improves performance slightly and, most importantly, stabilizes training on larger models. + +Experimentally, we show that within few epochs, MIM-Refiner refines the features of a MIM model to (i) incorporate the beneficial properties of ID objectives (ii) preserves the advantages of the MIM model (iii) exploits the synergies of both methods to improve upon each individual pre-training objective, advancing the state-of-the-art across various benchmarks, see Figure [1](#fig:gpuhours_and_1shot){reference-type="ref" reference="fig:gpuhours_and_1shot"}. + +Our contributions can be summarized as follows: + +1. We analyze MIM models, and find that MIM models have different types of blocks: those that mainly improve the pre-training objective and others that are responsible for abstraction. + +2. We introduce MIM-Refiner, a sequential approach to refine the representation of a pre-trained MIM model to form semantic clusters via an ID objective. Motivated by the findings in (1), MIM-Refiner is designed to exploit the intermediate representations via an ensemble of ID heads attached to multiple encoder blocks. + +3. We introduce Nearest Neighbor Alignment (NNA), an ID objective that aligns each sample with its nearest neighbor while repelling other samples in a batch. + +
+
+ +
Encoder-decoder Parameter Count
+
+
+ +
ImageNet-1K k-NN Accuracy per Block
+
+
+ +
ImageNet-1K Low-shot Accuracy.
+
+
+ +
Relative Improvements per Block.
+
+
(a) MIM models are asymmetrically designed where the encoder has most of the parameters, as indicated by the percentages in the bars. (b) The representation quality of MAE encoders (measured by k-NN accuracy) peaks in the middle blocks before degrading when the later blocks take over parts of the decoder’s task. Various MIM models and also a semantic segmentation task follow this pattern (see Appendix 9.4) (c) The decline in representation quality in later blocks primarily contributes to the degradation in downstream performance. ID methods (represented by iBOT ) and our MAE-Refined do not suffer from this issue. (d) Correlation of the relative improvement of the reconstruction loss and the relative improvement in the k-NN accuracy per block. The relative k-NN accuracy improvement is obtained from the k-NN accuracy shown in (b). Similarly, the relative improvement of the reconstruction loss is obtained from Figure 13 in the appendix. Middle blocks form abstract representations (large improvements in the k-NN accuracy, almost no improvement in reconstruction loss), later blocks take over parts of the reconstruction task (decrease in the k-NN accuracy, large improvement in the reconstruction loss). Additional details can be found in Appendix 11.2.
+
+ +MIM models, such as MAE [@he2022mae] and data2vec 2.0 [@baevski2023data2vec2], enable an efficient pre-training of large models. In terms of architecture, the encoder and decoder are intentionally designed asymmetrically. The encoder, on the one hand, is a large ViT, where discarding 75% of the input patches drastically reduces the cost of a forward pass. The decoder, on the other hand, operates on the full sequence length -- by concatenating mask tokens to the encoded visible patches -- and, thus, is typically *very* lightweight to compensate for the increased number of tokens (Figure [2](#fig:asymmetric_encoder_decoder_design){reference-type="ref" reference="fig:asymmetric_encoder_decoder_design"}). As models increase in size, the decoder eventually reaches a point where it cannot further improve the pre-training objective on its own. Consequently, it begins to delegate a portion of the reconstruction task back to the last encoder blocks. This transfer adversely affects the feature quality for downstream tasks associated with those blocks (Figure [3](#fig:mae_knn_per_block){reference-type="ref" reference="fig:mae_knn_per_block"}), especially when only few labels are available (Figure [4](#fig:mae_vs_ibot_vs_mimrefiner_lowshot){reference-type="ref" reference="fig:mae_vs_ibot_vs_mimrefiner_lowshot"}). We observe this phenomenon by correlating the relative improvement in the reconstruction loss vs the $k$-NN accuracy (Figure [5](#fig:mae_loss_vs_knn_improvement_h14){reference-type="ref" reference="fig:mae_loss_vs_knn_improvement_h14"}). Roughly speaking, the blocks of the encoder operate in three different regimes: + +1. In early ViT blocks, general purpose features are learned, which improve the reconstruction loss and the $k$-NN accuracy simultaneously. + +2. In middle ViT blocks, abstractions are formed. The reconstruction loss improves only slightly, while the $k$-NN accuracy improves drastically. + +3. In late ViT blocks, features are prepared for the reconstruction task. The reconstruction loss improves at a faster rate, while the $k$-NN accuracy decreases. + +When naïvely using the features of the last encoder block, those features are suited for reconstruction, but not particularly for downstream tasks. If lots of labels are available, this can be compensated by fine-tuning the last encoder blocks on the downstream task. However, if not enough labels are available, or the last encoder blocks are not fine-tuned, downstream performance suffers. + +One would think that simply using a larger decoder solves these problems. However, there are multiple problems with this solution. (i) The decoder commonly operates on the full sequence length, making it costly to increase its size. (ii) Scaling the decoder can decrease performance as shown, for example, in MAE [@he2022mae] (Table 1). (iii) Models that can use a deeper decoder (such as CrossMAE [@fu2024crossmae]) also show degrading representation quality in later blocks, as shown in Appendix Figure [10](#fig:crossmae_knn_per_layer){reference-type="ref" reference="fig:crossmae_knn_per_layer"}. + +# Method + +We propose MIM-Refiner, a novel approach aimed at improving downstream performance by refining the later blocks of a pre-trained MIM model. MIM-Refiner leverages the abstract intermediate representations with an ensemble of Instance Discrimination (ID) heads, which are attached to multiple blocks towards the end of the encoder, as visualized on the left side of Figure [7](#fig:schematic){reference-type="ref" reference="fig:schematic"}. The resulting experimental improvements in various downstream tasks are discussed in Section [5](#sec:experiments){reference-type="ref" reference="sec:experiments"}. + +![ **Left:** Comparison of different pre-training schemes. ID uses a single ID head, whereas MIM models use a light-weight decoder to train an encoder. MIM-Refiner attaches multiple ID heads to the later third of the blocks of a pre-trained MIM encoder. **Right:** Nearest Neighbor Alignment (NNA). An anchor sample is attracted by its NN and simultaneously repelled from other samples in the batch. The NN is retrieved from a first-in-first-out queue of samples from previous batches.](res/schematic_combined.pdf){#fig:schematic width="\\linewidth"} + +Inspired by NN based contrastive learning [@dwibedi2021nnclr; @azabou2021mineyourownview], we propose Nearest Neighbor Alignment (NNA). NN contrastive objectives introduce an inter-sample correlation by retrieving NNs of samples in a batch and subsequently applying an objective between the samples and their NNs. In practice, the NN retrieval is typically done by tracking samples from previous iterations in a first-in-first-out queue [@dwibedi2021nnclr; @he2020moco]. Therefore, the samples in the queue are from a previous model state which creates a tradeoff between the benefit of the NN-swap and the worse signal from the out-of-sync samples. We argue that the NN-swap does not offer a benefit for negative samples, since they are already different images, and instead, degrades the signal from the contrastive objective. We therefore propose to use the NN only for the alignment of positive samples, as visualized on the right side of Figure [7](#fig:schematic){reference-type="ref" reference="fig:schematic"}. Omitting the NN-swap for the negatives stabilizes training for large models and slightly improves performance (see Table [\[tab:ablations\]](#tab:ablations){reference-type="ref" reference="tab:ablations"}). NNA is a variant of NNCLR and we visualize their difference in Figure [16](#fig:nnclr_vs_nna){reference-type="ref" reference="fig:nnclr_vs_nna"}. + +Formally, given a batch of features $\mathcal{Z} = \{z_1, \dots, z_N\}$ and a queue $\mathcal{Q}$ the NNA objective is: $$\begin{align} +\mathcal{L}^\text{NNA}_i &= - \log \frac +{\exp(\text{NN}(z_i,\mathcal{Q}) \cdot z_i / \tau)} +{ +\exp(\text{NN}(z_i,\mathcal{Q}) \cdot z_i / \tau) ++ \sum\limits^N_{j=1} \exp(\text{SG}(z_j) \cdot z_i / \tau) [i \neq j] +} \ \\ +\text{NN}(z_i, \mathcal{Q}) &= \underset{q\in\mathcal{Q}}{\text{argmax}}(z_i \cdot q) \ +\label{eq:nn_swap} +\end{align}$$ where $z_i$ is the anchor, $\text{NN}(z_i, \mathcal{Q})$ is the positive, $z_j$ are the negatives, $\tau$ is the temperature, and $\text{SG}$ is the stop-gradient operation. $[i \neq j]$ denotes the Iverson bracket that evaluates to $1$ if $i \neq j$ and to $0$ if $i = j$. All vectors are assumed to be normalized to length $1$. diff --git a/2402.10175/main_diagram/main_diagram.drawio b/2402.10175/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..ba8cdabbb3e93963a66baf4c164e8b5a06feab3e --- /dev/null +++ b/2402.10175/main_diagram/main_diagram.drawio @@ -0,0 +1,76 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/2402.10175/main_diagram/main_diagram.pdf b/2402.10175/main_diagram/main_diagram.pdf new file mode 100644 index 0000000000000000000000000000000000000000..bfd4a7b36332ffdfb0cfefbdea771776dff7b6ba Binary files /dev/null and b/2402.10175/main_diagram/main_diagram.pdf differ diff --git a/2402.10175/paper_text/intro_method.md b/2402.10175/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..c5876cbdfb9545ddd5201fa7fad898479e26afe6 --- /dev/null +++ b/2402.10175/paper_text/intro_method.md @@ -0,0 +1,91 @@ +# Introduction + +Real-life texts often exhibit underlying structures. News articles, for instance, adhere to a specific narrative order, as illustrated in Fig. [1,](#page-0-1) employed by journalists to efficiently convey messages and improve reader experience. Despite recent advances in generation of fluent text, [Deng et al.](#page-4-0) [\(2022\)](#page-4-0) demonstrate that transformer-based models struggle to effectively capture and learn the underlying latent transition structure of coherent text. Consequently, generating structurally coherent text remains an under-explored area of research. Follow the theory of functional discourse structure, elaborated in Appendix [A.1,](#page-6-0) we leverage the discourse structure to model the coherence of long-form texts. Several recent works [\(Spangher et al.,](#page-5-0) [2022;](#page-5-0) [Liu](#page-5-1) [et al.,](#page-5-1) [2022\)](#page-5-1) have addressed the problems of generating long-form text while following specific indomain discourse schema. + +**"Britain's Vision for 2100: Spaceports and Sky Farms Propel the Nation's Innovation"** + +The United Kingdom envisions a futuristic landscape in the year 2100, featuring state-of-the-art spaceports and innovative sky farms as part of its evolving infrastructure. The integration of spaceports and sky farms into Britain's infrastructure in 2100 is expected to revolutionize the country's economy and environmental sustainability, leading to increased job opportunities, advanced agricultural practices, and reduced carbon emissions. The United Kingdom has been steadily investing in research and development for space exploration and vertical farming in the years leading up to 2100, laying the foundation for the implementation of spaceports and sky farms in the future. Looking back at the history of British innovation, it is evident that the United Kingdom has a rich legacy of pioneering breakthroughs, from the Industrial Revolution to the modern computing era. [...] + +**
, , , , , , , ** + +Figure 1: A news article example with discourse role annotations. The discourse schema follows the News discourse theory by [Van Dijk](#page-5-2) [\(2013\)](#page-5-2). + +While established automatic metrics such as BLEU [\(Papineni et al.,](#page-5-3) [2002\)](#page-5-3), ROUGE-L [\(Lin and](#page-5-4) [Hovy,](#page-5-4) [2002\)](#page-5-4) and BertScore [\(Zhang et al.,](#page-5-5) [2019\)](#page-5-5) exist for Natural Language Generation evaluation, they predominantly measure lexical n-gram overlaps or semantic similarities. The evaluation of structural coherence has been a long-existing challenge [\(Guan et al.,](#page-4-1) [2021;](#page-4-1) [Cho et al.,](#page-4-2) [2019;](#page-4-2) [Zhu and](#page-6-1) [Bhat,](#page-6-1) [2020;](#page-6-1) [Deng et al.,](#page-4-0) [2022\)](#page-4-0). A common baseline metric for measuring functional discourse structure is the exact match, which compares structure elements one-to-one at each exact position. However, this metric is notably sensitive to local variations and differences in the lengths of articles. + +To address this gap, we propose a novel automatic model-free metric, Positional Discourse Divergence (PDD), specifically designed to evaluate the underlying discourse structure of articles in comparison to references. PDD partitions the sentences of an article into multiple position bins and calculates the divergence in discourse structures within each bin. This approach renders PDD re- + +1Our code is available at [https://github.com/](https://github.com/williamLyh/pos_div_metric) [williamLyh/pos\\_div\\_metric](https://github.com/williamLyh/pos_div_metric) + +silient to various challenges encountered in longform text generation, such as accommodating local variations and handling misaligned numbers of sentences. + +To validate the effectiveness and generalizability of the PDD, we evaluate the inter-agreement with human evaluations and GPT-4 coherence evaluations on three representative datasets with different discourse schema: News Discourse (Choubey et al., 2020), Long-Form Question Answering (Xu et al., 2022a) and Recipe1M+ (Liu et al., 2022). Across all three domains, PDD demonstrates the highest agreement with human judgements on coherence. + +Texts within a specific genre often exhibit similar patterns in their discourse sequences, albeit with some variations at a local level. In other words, the distribution of discourse roles is inherently tied to their approximate positions within the articles. For instance, News reports commonly present main events and their consequences at the beginning to capture the reader's interest, even though the precise order can differ. Likewise, recipes tend to follow a predictable structure, where the preparation of ingredients is generally mentioned first, followed by cooking actions towards the middle or end of the text. + +Despite the fluency achieved by (large) Language Models, they struggle to organize discourse structures like humans. In Fig. 2, we observe disparities between the discourse distributions of model predictions and human-written references when the News articles are divided into 5 positional bins. To quantitatively capture these gaps, we introduce the Positional Discourse Divergence (PDD), denoted as $D_{pos}$ , as an automatic metric. Equation 1 outlines the calculation for applying PDD to compare a *predicted* article against its corresponding *reference*: + + +$$D_{pos} = \frac{1}{N} \sum_{n=1}^{N} D_{KL}(p^{n}(r) + \epsilon || q^{n}(r) + \epsilon) \quad (1)$$ + +Firstly, both articles are segmented into N positional bins. Note the number of bin, N, should be smaller than the number of sentences in both the reference and the prediction. We denote $p^n(r)$ to represent the distribution of discourse role r for the reference in the n-th position bin, and $q^n(r)$ to represent the distribution for the generated article. These discourse distributions are calculated by the + +![](_page_1_Figure_7.jpeg) + +Figure 2: Positional discourse distribution comparisons (N=5). Top row: The discourse distribution of model predictions on News Discourse test set (Llama2-7b, finetuned on Kaggle All the News). Bottom row: Test set reference distributions. + +frequency density of the discourse roles within each bin. These discourse distributions are calculated by the frequency density of the discourse roles within each bin. Then, for each bin n, the KL divergence between the discourse distributions is calculated. 2 To address the sparsity in the discourse distribution of a single article, small-value terms, denoted as $\epsilon$ , are introduced. This addition helps in avoiding instances of zero probabilities in the distribution. + +To compute PDD or other discourse measurements like exact match, it is inevitable to employ a discourse role classifier for labeling both prediction and reference articles. An off-the-shelf discourse classifier, trained on human-annotated data with a defined schema (e.g., the News Discourse dataset for news domain), can serve this purpose. Further information regarding the discourse classifiers is provided in Appendix B. + +Much like the BLEU score, the Positional Discourse Divergence can be applied to a set of text, including both the set of model predictions and the set of reference articles. The underlying assumption is that all articles within a given set adhere to similar discourse structures, for example, being News articles of the same sports genre. Consequently, the discourse distributions of this set of articles offer a more accurate estimate of the discourse distribution specific to that genre. + +In the assessment of a single predicted article against a reference set, the focus is on how well + +&lt;sup>2Due to the asymmetry nature of KL divergence, $D_{KL}(P||Q)$ is interpreted as the information divergence of Q against P. Accordingly, we employ $q^n(r)$ to denote the discourse distributions of the predictions and $p^n(r)$ for the reference. + +![](_page_2_Figure_0.jpeg) + +Figure 3: Positional Discourse Divergence vs. Bin number (N) for predictions by two language models on the News Discourse test set. Training details in Appendix C. Curves represent best-fit quadratic curves. + +the article aligns with the target genre. In contrast, when comparing a prediction set against a reference set, the evaluation exams the model's overall ability to generate content of that specific genre. + +The bin number, N, plays a crucial role in determining the sensitivity of PDD to local variations, as illustrated in Fig. 3. Therefore, the behavior of PDD varies with the choice of N. Intuitively, a larger N implies lower tolerance for local perturbations. When N equals the number of sentences, the PDD is essentially equivalent to the exact match metric. Whereas, when N equals 1, it describes the overall discourse role distribution gaps between the prediction to the reference. + +To illustrate, we fine-tuned Llama2-7b and Llama2-13b (Touvron et al., 2023) and compared their predictions with the reference News articles. The details of the supervised fine-tuning process are explained in Appendix C. The PDD curves, illustrating the performance with different choices of bin numbers, are presented in Fig. 3. The gaps in performance are effectively captured by the disparities between the two PDD curves. + +To validate the efficacy of the Positional Discourse Divergence metric, we evaluate its agreement with human assessments, and GPT-4 on article coherence. Additionally, we compare PDD against baseline automatic metrics, such as exact match, BLEU, and BertScore. To assess generalizability, we conduct this validation across three different domains. + +each characterised by distinct human annotated, sentence-level discourse schemas: + +(I) News. We utilize the News Discourse dataset (Choubey et al., 2020), comprising 802 documents across four genres and three media sources. The average number of sentences per article is 14.6. Manual annotations for the News Discourse dataset follow the theory of functional discourse schema proposed by Van Dijk (1988, 2013). This schema defines discourse based on eight types of relations between each sentence and the main event. + +(II) Long-form QA. Long-Form Question Answering (LFQA) involves providing comprehensive answers composed of multiple sentences. Xu et al. (2022a) proposed an discourse ontology of six sentence-level functional roles also following the theory of functional discourse structure. The discourse annotations are collected on three recent LFQA datasets (ELI5 (Fan et al., 2019), WebGPT (Nakano et al., 2022), and Natural Questions (Kwiatkowski et al., 2019)). A total of 640 answer paragraphs were released, with an average of 6.1 sentences per paragraph. + +(III) Recipes. We adopt the discourse schema proposed by Liu et al. (2022) which includes seven discourse roles based on cooking actions specifically designed for recipes. They annotated the Recipe1M+ dataset (Moryossef et al., 2019; Marín et al., 2021) with a rule-based annotation system following the proposed schema. The Recipe1M+ contains over 1M textual recipes and ingredients. + +For further information regarding dataset details and schema definitions, please refer to Appendix E. + +We validate the effectiveness of our metric, PDD, by assessing its inter-agreement with human evaluations and GPT-4 coherence evaluations. The human evaluation setup details can be found in Appendix D. In our comparison, PDD is evaluated alongside several automatic metrics, including exact match, BLEU (Papineni et al., 2002), ROUGE-L (Lin and Hovy, 2002) and BertScore (Zhang et al., 2019). Notably, only PDD and exact match focus on directly measuring discourse structure, while the others are designed for assessing n-gram or semantic similarity. + +As rating long-form articles with absolute scores is a relatively complicated and subjective task, we instead ask evaluators compare two perturbed variations of the original reference article. Cohen's Kappa is computed between the metrics and eval- + + + +| Metrics | Human | | | GPT-4 | | | +|-------------|----------------|------|-----------|----------------|------|-----------| +| | News Discourse | LFQA | Recipe1M+ | News Discourse | LFQA | Recipe1M+ | +| Exact Match | 0.26 | 0.29 | 0.43 | 0.42 | 0.24 | 0.25 | +| ROUGE-L | 0.30 | 0.24 | 0.39 | 0.44 | 0.19 | 0.26 | +| BLEU | 0.24 | 0.31 | 0.46 | 0.48 | 0.28 | 0.32 | +| BertScore | 0.45 | 0.42 | 0.63 | 0.63 | 0.42 | 0.68 | +| PDD | 0.42 | 0.49 | 0.66 | 0.65 | 0.38 | 0.71 | + +Table 1: Cohen's Kappa with human and GPT-4 coherence evaluations. For News Discourse, bin number N = 8 was used, while for LFQA and Recipe1M+, N = 3. Human evaluation involves 50 randomly selected example pairs for each dataset, while GPT-4 evaluation uses 300 pairs. + +uators based on these preference annotations. We create two variations in the way that prevents resulting PDD values from exhibiting a heavy left-tail issue and ensuring a more accurate kappa estimation. In Variation 1, we randomly shuffle all the sentences, whereas in Variation 2, we initially segment the article into a randomly selected number of bins and then shuffle sentences only within their respective positional bins. + +In Tab. [1,](#page-3-0) we report the Kappa with both human and GPT-4 coherence evaluations. The details of the prompt and survey templates are shown in Appendix [G.](#page-8-0) Our PDD metric demonstrates consistent good agreement (0.4-0.6) in News Discourse and LFQA, achieving substantial agreement (>0.7) on Recipe1M+ dataset. The notable performance on the Recipe1M+ dataset can be attributed to the strong order-dependent nature of recipes: A shuffled question-answer format may be challenging to understand, but a disordered recipe is nearly incomprehensible. + +Another observation on News Discourse dataset, indicates Kappas with human evaluations are generally lower than those with GPT-4. This discrepancy is likely due to the much longer length of news articles compared to question answering and recipe datasets, posing a more challenging task for human readers. + +Our PDD metric significantly outperforms baseline metrics of Exact Match, Rouge-L and BLEU, while achieving comparable Kappa with the BertScore. We attribute the good performance of BertScore to its ability in carrying textual knowledge from the pre-trained BERT. In our experiment setup, both Variation 1 and 2 are shuffled from the same articles. Consequently, the metrics based on n-gram and semantic similarities can effectively distinguish examples closer to the original version and therefore achieve high kappa values. However, when comparing the discourse structure between two different articles of the same genre, they are likely to have very different n-gram or semantic content while maintaining a similar discourse structure. In this case, only Exact Match and PDD can capture the divergence between discourse structure. diff --git a/2402.15351/main_diagram/main_diagram.drawio b/2402.15351/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..8a8769dfdd5f71b269fc948b98036d62f5bd2a5b --- /dev/null +++ b/2402.15351/main_diagram/main_diagram.drawio @@ -0,0 +1,315 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/2402.15351/paper_text/intro_method.md b/2402.15351/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..d0df55998e57b86764dc2d8fd1b6c36c4b70fde4 --- /dev/null +++ b/2402.15351/paper_text/intro_method.md @@ -0,0 +1,89 @@ +# Introduction + +Machine learning (ML) has achieved considerable successes in recent years and an ever-growing number of disciplines rely on it. However, this success crucially relies on human ML experts to perform tasks including data preparation, model algorithm selection, model training with hyperparameter tuning and model deployment. The rapid growth of machine learning applications has created a demand for off-theshelf machine learning methods that can be used easily and without expert knowledge. To this end, automated machine learning, or AutoML, has been developed in literature. AutoML aims to minimize the workload of human experts in the ML development process, making AI algorithms available and easily accessible even for non-AI experts. + +Recent public computer vision (CV) communities (*e.g*. OpenMMLab) have provided a wide variety of toolboxes **Input:** Request (Data & Model & Deploy) + +![](_page_0_Figure_11.jpeg) + +Figure 1: AutoMMLab autmatically creates deployable models from user's language instructions. + +in various research areas such as image classification (MM-Pretrain (Contributors 2023)), object detection (MMDetection (Chen et al. 2019)), semantic segmentation (MMSegmentation (Contributors 2020a)), and pose estimation (MM-Pose (Contributors 2020b)). Considering large language models (LLMs) have exhibited exceptional ability in language understanding, generation, interaction, and reasoning, we are inspired to employ LLMs to connect AutoML with public CV communities (*e.g*. OpenMMLab) for automating the *whole* CV model production workflow via a natural language based interface. + +In this paper, we introduce a novel comprehensive task called request-to-model. Unlike traditional AutoML task that focus on one particular step in the model production workflow (*e.g*. neural architecture search or hyperparameter optimization), request-to-model extends the reach of AutoML to evolve towards fully automated model production through a user-friendly language interface. This empowers individuals to leverage the capabilities of AI without extensive expertise, driving progress in various fields. + +To facilitate the development of the request-to-model task, we propose a LLM-empowered platform called AutoMMLab, which can serve as a testbed for developing and evaluating the request-to-model algorithms. As shown in Figure 1, given language-based user requirements, our + +\*Corresponding Author Copyright © 2025, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. + +AutoMMLab platform will schedule each module to execute the entire workflow and finally output production-ready models for computer vision tasks. AutoMMLab connects diverse datasets, CV models, training pipelines and deployment tools to facilitate the solution of numerous CV tasks. The whole workflow consists of 5 major modules, including request understanding, data construction, model construction, model training with hyperparameter optimization, and model deployment. + +Thanks to the modular design of AutoMMLab, we also develop the LAMP (Language-instructed Automated Model Production) benchmark for research on the languageinstruction based model production. The LAMP is the *first* benchmark for evaluating the request-to-model AutoML algorithms in computer vision, which enables the community to explore key components in the end-to-end requestto-model pipeline. + +Hyperparameter optimization (HPO) aims to choose a set of optimal hyperparameters that yields an optimal performance based on the predefined objective function. Thanks to its wide applications, it becomes one of the most important components for AutoML. Traditional approaches (*e.g*. bayesian optimization) mostly rely on trial-and-error, leading to inefficient parameter search. To solve this problem, we propose a novel LLM-based hyperparameter optimization algorithm, called HPO-LLaMA. Equipped with rich knowledge of model training, HPO-LLaMA can skillfully search for the optimal model training hyperparameters with significantly reduced numbers of trials. + +Our main contributions can be summarized as follows: + +- We propose the novel request-to-model task, automating the whole model production workflow. We present the *first* request-to-model AutoML platform for computer vision tasks, called AutoMMLab. By integrating AutoML and language interface, it enables non-expert users to easily build and deploy CV models, unlocking the potential of AI for a wider audience. +- Based on the AutoMMLab platform, we build a benchmark termed LAMP for evaluating end-to-end promptbased model production, and also studying each component in the whole production pipeline. +- We propose a novel LLM-based hyperparameter optimization algorithm, called HPO-LLaMA. To the best of our knowledge, HPO-LLaMA is the first supervised finetuned LLM specifically designed for HPO. + +# Method + +The overview of AutoMMLab platform is illustrated in Figure 2. It directly takes language-based model request from users as the input, and automatically schedule and execute the entire workflow to produce deployable models. The whole pipeline of AutoMMLab consists of 5 main stages, including request understanding, data selection, model selection, model training with hyperparameter optimization (HPO), and model deployment. The Request Understanding stage understands the language-based model request and schedules the model production workflow, generating a structured configuration. The configuration contains detailed data information, model constraints and deployment requirements. AutoMMLab then construct the training data inside built-in dataset zoo and select the most appropriate model that meets the model constraints. When the data and model are ready, AutoMMLab execute model training with hyperparameter optimization (HPO) to obtain the final trianed model. In the model deployment stage, AutoMMLab apply MMDeploy (Contributors 2021) to deploy the model based on the user's hardware requirements. + +Request understanding aims to automatically and accurately understand the user's request, schedule the model production workflow, and generate the JSON-format configuration. Since it is non-trival for general-purpose LLMs to succeed in request understanding solely through prompting, we also train a task-specific LLM (RU-LLaMA) to achieve this goal. + +GPT-assisted Training Data Generation. Tuning LLMs typically require numerous training data (Liu et al. 2023b,a), which is exceptionally costly and time-consuming to collect with human crowd-scouring. Inspired by the success of recent LLMs in data generation (Wang et al. 2022), we leverage GPT-4 to build our data generation pipeline. For each task type, (*i.e*. image classification, object detection, semantic segmentation, and keypoint estimation), we first ask 5 professional computer vision researchers to manually design 100 diverse request-config pairs. These human annotations are used as seeds in in-context-learning to query GPT. In our pipeline, we first design prompts to ask GPT to generate diverse requests, and then use carefully designed prompts to query GPT to parse the requests and generate corresponding configurations. In our preliminary experiments, we ablated the use of GPT-3.5 and GPT-4 and empirically found that GPT-4 can consistently produce higher quality data. Therefore, we use GPT-4 in our data generation pipeline and generate 1,000 request-config pairs. Finally, the generated data samples are carefully checked and manually corrected by a group of professional annotators. + +RU-LLaMA. LLM has learned massive knowledge in the pre-training stage and have strong general question and answer capabilities. However, its usually require supervised fine-tuning on instruction-tuning dataset to better activate its domain-specific capabilities. We construct a instruction fine-tuning dataset based on the dataset generated by GPT-4. Following the mainstream methods of fine-tuning large language models in specific domains, we use LoRA (Hu et al. 2022) technology to fine-tune the LLM. Specifically, we utilize LoRA to fine-tune Llama-2 (Touvron et al. 2023b), and obtained RU-LLaMA that is skilled in parsing user's requirements into JSON-format configurations. + +Data selection stage aims to use the JSON-format configuration parsed by RU-LLaMA to automatically retrieve and build request-related training dataset from the dataset zoo. Dataset Zoo. To facilitate dataset preparation, we construct a comprehensive dataset zoo encompassing a wide variety of public datasets (Krizhevsky, Sutskever, and Hinton 2017; Lin et al. 2014; Cordts et al. 2016). Each dataset is complemented by a data card that contains the meta information appertaining to the dataset intricately. We also take into account AI safety considerations, by ensuring openness, transparency, and reliability of data. Dataset Selection Pipeline. We design an elaborate pipeline to automatically retrieve pertinent images from the dataset zoo. First, we extract the relevant data cards from the dataset zoo based on the specified task category. Next, if the user requests the specific dataset, we compute the fuzzy match scores, quantifying the similarity between the user-designated name and the name of data cards. These cards are subsequently reordered in a descending pattern, according to their scores. Then the images that meet the requirements are sequentially collected from each dataset. Meanwhile, we use WordNet (Miller 1995) to identify semantic relationships between objects. + +Model selection automates the process of choosing the most suitable model from the model zoo. Model Zoo. We first construct a comprehensive model zoo where each model is accompanied by a detailed model card and the pre-trained model weights. Each model card incorporates attributes such as the model's name, structure (*e.g*. network depth), parameters, floating point operations per second (FLOPs), inference speed, and performance. We have considered AI safety issues when constructing the model zoo. We try to keep the model source open and transparent, and all models are pretrained with reliable public data. Model Selection Pipeline. We also design a elaborate pipeline able to automatically determine the most suitable model in model zoo, tailored to the user's model-specific requirements. First, we extract the task-related model cards from the model zoo according to predicated task categorization specified in parsed config. If the user has preference to specific model, then calculate the fuzzy matching score between the card's name containing + +![](_page_3_Figure_0.jpeg) + +Figure 2: Overview of AutoMMLab. The workflow of AutoMMLab consists of five stages. **Request understanding:** Parse the language requests into formated configuration. **Data selection:** Select appropriate training data from the dataset zoo. **Model selection:** Select the optimal model from the model zoo. **Model training with HPO:** Train the model and optimize the hyperparameters. **Model deployment:** Convert the model into a package compatible with the deployment environments. + +the model structure information and the user's specified requirements. Next, we filter out model cards that did not meet the constraints imposed by users regarding the model's parameters, FLOPs, or the inference speed. Finally, we select the optimal model from the remaining model cards based on fuzzy matching scores and model performance. + +Hyperparameter optimization (HPO) or hyperparameter tuning is the problem of choosing a set of optimal hyperparameters for model training, which is crucial for the enhancement of the model performance. Hyperparameters are configuration settings that are not learned during training, but are predefined to control the learning process (*e.g.* learning + +rate and batch size). AutoMMLab supports various types of HPO baselines, including classic methods (*e.g.* random sampling) and LLM-based methods. + +Model deployment in computer vision means integrating a trained computer vision model into a real-world production environment. Model deployment mainly involves model conversion, graph structure optimization, and model quantization. Our AutoMMLab platform is equipped with MMDeploy (Contributors 2021), a comprehensive model deployment toolbox, for one-click end-to-end model deployment. It supports a growing list of models that are guaranteed to be exportable to various backends (*e.g.* ONNX, NCNN, Open- + +![](_page_4_Figure_0.jpeg) + +Figure 3: Example of LAMP dataset. + +VINO). The model converters enable converting OpenMM-Lab models into backend models that can be run on target devices. The resulting package is also with a inference SDK, supporting data preprocessing, model forward and postprocessing modules during the model inference phase. + +Large language models (LLMs) have shown amazing emergent abilities in recent studies, however, there lacks a comprehensive benchmark to evaluate the capability of recent LLMs for AutoML. To fill in this blank, we introduce the LAMP (Language-instructed Automated Model Production) benchmark3 . The benchmark provides a common ground for evaluating and developing the AutoML platform especially for language-instruction based model production. + +Testing Data Collection. The benchmark covers four basic computer vision tasks, *i.e*. image classification, object detection, semantic segmentation, and keypoint detection. To ensure the quality of the test data, we ask 5 professional computer vision researchers to manually design 20 diverse CV model requests for each task. In total, we collect 80 unique user instructions and each instruction is accompanied with a ground truth configuration and a test dataset. To enable comparisons of different LLMs for request understanding, we also provide the corresponding ground-truth configuration for each instruction. An example of our test sample is delineated in Figure 3. + +Configuration Definition. As shown in Figure 3, we define three dimensions of model requirements in the configuration file: "data", "model", and "deploy". "Data" contains the user's data-related requirements, including: application scenarios, target objects, data modalities, etc. "Model" contains the requirements related to the model, including: task type, model constraints, and the performance that the model is expected to achieve. The constraints of the model may include parameter size, floating point operations per second (FLOPs) and inference speed. "Deploy" contains the user's deployment requirements for the model, including whether to use GPU, which inference engine to use, etc. Note that all values in the configuration can be empty or "none". + +We establish three types of evaluation protocols for our LAMP benchmark: + +- Evaluation of Request Understanding. The evaluation of request understanding assesses the ability of LLMs to parse user's requests into configurations, adhering to the predefined JSON format. We measure performance based on the accuracy of parsing key-value pairs in the configuration. These pairs can be categorized into item-type pairs and listtype pairs, depending on the data format. In the case of listtype pairs, such as target objects, the pair is considered correctly parsed only if all items in the list are parsed accurately. Note that we apply post-processing and fuzzy matching before calculating the parsing accuracy. Two evaluation metrics are used to gauge the request understanding capabilities: (1) Key-level accuracy calculates the average accuracy of each key-value pair. (2) Req-Level accuracy calculates the accuracy of understanding the entire request. +- Evaluation of Hyperparameter Optimization. LAMP allows researchers to assess the effectiveness of various Hyperparameter Optimization (HPO) algorithms across a wide range of tasks. Different evaluation metrics are utilized for different tasks, with the mean values of the generated model serving as the primary evaluation metric for HPO. For image classification, the top-1 accuracy is adopted. For object detection, the standard mean Average Precision (mAP) is reported. In the case of semantic segmentation, mean Intersection over Union (mIOU) is employed. For keypoint estimation, mAP based on Object Keypoint Similarity (OKS) is used. Simultaneously calculate the mean value and standard deviation of four tasks for overall evaluation. +- End-to-end Evaluation. End-to-end evaluation focuses on the functional quality of the model generated by the request-to-model platform and is assessed using a grading scale ranging from 'F' for total failure to 'P' for perfect conformity to users' specifications: 'F' for total failure, scoring 0 points. The platform either fails to generate a functional model. 'W' for workable model, scoring 1 point. The model is runnable, but it may not fully meet all the users' requirements, such as lower accuracy. 'P' for perfect model, scoring 2 points. The model perfectly matches the users' expectations. There are four types of tasks. The full score for each task is 40, and the total full score is 160. + +We propose to tackle the challenging HPO task by tuning a LLM-based optimizer, called HPO-LLaMA. Figure 4 demonstrates an overview of our proposed HPO-LLaMA workflow. Initially, HPO-LLaMA begins by receiving a nature language description of the HPO problem in conjunction with request-specific details relating to the data and model. Following this, it generates a specific hyperparameter setting tailored to the request. Subsequently, HPO-LLaMA acquires the outcomes from the model trainings conducted using the specified hyperparameter setting on the test data and then endeavors to recommend an optimally enhanced hyperparameter setting. + +Search Space We carefully constructed a compact yet expressive search space for HPO, which includes: optimizer type, initial learning rate, learning rate decay policy, weight decay value, batch size, and the training iterations. Consid- + +3This is the v1 version of LAMP, which will be updated as LLMs evolve. + +![](_page_5_Figure_0.jpeg) + +Figure 4: Overview of HPO-LLaMA. At the initial step (t = 1), HPO-LLaMA proposes a hyperparameter configuration based on the description of model and task. Model training is then executed and the training results are passed back to HPO-LLaMA via a text prompt for further rounds (t > 1). + +ering the heterogeneity of the four tasks, a bespoke search space is delineated for each individual task. + +Training HPO-LLaMA We first adopt GPT-assisted data generation method (introduced in Request Understanding) to generate 100 diverse training requests for each of the four CV tasks. And for each training request, we stochastically select 20 distinct hyperparamter settings from the search space. Consequently, we extensively execute model training for 8000 experiments. The outcomes of these experiments are compiled into a substantial dataset, comprising triplets of "request-hyperparameter-performance". we further conduct a comprehensive evaluation to elucidate the correlation between hyperparameters and model performance. This is primarily to substantiate the appropriateness of the predetermined selection for hyperparameter space. Please refer to supplementary for details. + +The triplets of "request-hyperparameter-performance" contains rich knowledge about model training and hyperparameter tuning. We utilize the triplet data to construct both single-round and multi-round dialogues, thereby empowering HPO-LLaMA to activate the capability of predicting efficient hyperparameters from these dialogues. Specifically, when constructing the first round of dialogue, the contextual prompts includes the hyperparameter search space, along with a detailed description of the data and the model. The hyperparameter settings to be predicted are derived from the top-k optimal hyperparameter settings corresponding to the request. In multiple rounds of conversations, the contextual prompt provided to HPO-LLaMA includes historical prompt of previous conversations and the performance of the model trained on the test set for the last time the model predicted hyperparameters. The predicted hyperparameter settings are derived from top-k of all hyperparameter settings corresponding to the request and are the best performing hyperparameter setting in current multi-round dialogue. Finally, we construct the training dataset for up to three dialogue rounds and selectively sample a subset from the entire pool of potential multi-turn conversation data. We then use LoRA to fine-tune LLaMA-7B and get our HPO-LLaMA. Considering the limited volume of data in a single-round dialogue, we mix dialogue rounds ranging from one to three together to train the HPO-LLaMA. diff --git a/2402.17595/main_diagram/main_diagram.drawio b/2402.17595/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..cb44fcc7fad39e828c10a9edb7dd1685ce9a6ff5 --- /dev/null +++ b/2402.17595/main_diagram/main_diagram.drawio @@ -0,0 +1 @@ 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 \ No newline at end of file diff --git a/2402.17595/main_diagram/main_diagram.pdf b/2402.17595/main_diagram/main_diagram.pdf new file mode 100644 index 0000000000000000000000000000000000000000..5ed2a5d617e679d09a3b4a1b7dd38b9d4f361acb Binary files /dev/null and b/2402.17595/main_diagram/main_diagram.pdf differ diff --git a/2402.17595/paper_text/intro_method.md b/2402.17595/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..521f15f52b3304793e53e4d9c8911781bade5081 --- /dev/null +++ b/2402.17595/paper_text/intro_method.md @@ -0,0 +1,215 @@ +# Introduction + +A longstanding pursuit of deep learning theory is to explain the astonishing ability of neural networks to generalize despite having far more learnable parameters than training data, even in the absence of any explicit regularization. An established understanding of this phenomenon is that the gradient descent algorithm induces a so-called *implicit regularization* effect. In very general terms, implicit regularization entails that gradient flow dynamics in many neural nets, even without any explicit regularizer in the loss function, converges to the solution of a regularized learning problem. In a sense, this creates a learning paradigm that automatically favors models characterized by "low complexity". + +A standard test-bed for mathematical analysis in studying implicit regularization in deep learning is the *matrix sensing* problem. The goal is to approximate a matrix $X^\star$ from a set of measurement matrices $A_1,\dots,A_m$ and observations $y_1,\dots,y_m$ where $y_i=\langle A_i, X^\star \rangle$. A common approach, *matrix factorization*, parameterizes the solution as a product matrix, i.e., $X=UV^\top$, and optimizes the resulting non-convex objective to fit the data. This is equivalent to training a depth-$2$ neural network with a linear activation function. + +In an attempt to explain the generalizing ability of over-parameterized neural networks, [@neyshabur2014search] first suggested the idea of an implicit regularization effect of the optimizer, which entails a bias towards solutions that generalize well. [@gunasekar2017implicit] investigated the possibility of an implicit norm-regularization effect of gradient descent in the context of shallow matrix factorization. In particular, they studied the standard Burer-Monteiro approach [@burer2003nonlinear] to matrix factorization, which may be viewed as a depth-2 linear neural network. They were able to theoretically demonstrate an implicit norm-regularization phenomenon, where a suitably initialized gradient flow dynamics approaches a solution to the nuclear-norm minimization approach to low-rank matrix recovery [@recht2010guaranteed], in a setting where the involved measurement matrices commute with each other. They also conjectured that this latter restriction on the measurement matrices is unnecessary. This conjecture was later resolved by [@li2018algorithmic] in the setting where the measurement matrices satisfy a restricted isometry property. Other aspects of implicit regularization in matrix factorization problems were investigated in several follow-up papers [@neyshabur2017exploring; @arora2019implicit; @razin2020implicit; @tarmoun2021understanding; @razin2021implicit]. For instance, [@arora2019implicit] showed that the implicit norm-regularization property of gradient flow, as studied by [@gunasekar2017implicit], does not hold in the context of deep matrix factorization. [@razin2020implicit] constructed a simple $2 \times 2$ example, where the gradient flow dynamics lead to an eventual blow-up of any matrix norm, while a certain relaxation of rank---the so-called *e-rank*---is minimized in the limit. These works suggest that implicit regularization in deep networks should be interpreted through the lens of rank minimization, not norm minimization. Incidentally, [@razin2021implicit] have recently demonstrated similar phenomena in the context of tensor factorization. + +Researchers have also studied implicit regularization in several other learning problems, including linear models [@soudry2018implicit; @zhao2019implicit; @du2019width], neural networks with one or two hidden layers [@li2018algorithmic; @blanc2020implicit; @gidel2019implicit; @kubo2019implicit; @saxe2019mathematical]. Besides norm-regularization, several of these works demonstrate the implicit regularization effect of gradient descent in terms of other relevant quantities such as margin [@soudry2018implicit], the number of times the model changes its convexity [@blanc2020implicit], linear interpolation [@kubo2019implicit], or structural bias [@gidel2019implicit]. + +A natural use case scenario for investigating the implicit regularization phenomenon is the problem of matrix sensing. Classical works in matrix sensing and matrix factorization utilize convex relaxation approaches, i.e., minimizing the nuclear norm subject to agreement with the observations, and deriving tight sample complexity bound [@srebro2005rank; @candes2009exact; @recht2010guaranteed; @candes2010power; @keshavan2010matrix; @recht2011simpler]. Recently, many works analyze gradient descent on the matrix sensing problem. [@ge2016matrix] and [@bhojanapalli2016global] show that the non-convex objectives on matrix sensing and matrix completion with low-rank parameterization do not have any spurious local minima. Consequently, the gradient descent algorithm converges to the global minimum. + +Despite the large body of works studying implicit regularization, most of them consider the linear setting. It remains an open question to understand the behavior of gradient descent in the presence of non-linearities, which are more realistic representations of neural nets employed in practice. + +In this paper, we make an initial foray into this problem, and investigate the implicit regularization phenomenon in more realistic neural networks with a general class of non-linear activation functions. We rigorously demonstrate the occurrence of an implicit regularization phenomenon in this setting for matrix sensing problems, reinforced with quantitative rate guarantees ensuring exponentially fast convergence of gradient descent to the best approximation in a suitable class of matrices. Our convergence upper bounds are complemented by matching lower bounds which demonstrate the optimality of the exponential rate of convergence. + +In the bigger picture, we contribute a network architecture that we refer to as the Spectral Neural Network architecture (abbrv. SNN), which is particularly suitable for matrix learning scenarios. Conceptually, this entails coordinatizing the space of matrices by their singular values and singular vectors, as opposed to by their entries. We believe that this point of view can be beneficial for tackling matrix learning problems in a neural network setup. SNNs are particularly well-suited for theoretical analysis due to the spectral nature of their non-linearities, as opposed to vanilla neural nets, while at the same time provably guaranteeing effectiveness in matrix learning problems. We also introduce a much more general counterpart of the near-zero initialization that is popular in related literature, and our methods are able to handle a much more robust class of initializing setups that are constrained only via certain inequalities. Our theoretical contributions include a compact analytical representation of the gradient flow dynamics, accorded by the spectral nature of our network architecture. We demonstrate the efficacy of the SNN architecture through its application to the matrix sensing problem, bolstered via both theoretical guarantees and extensive empirical studies. We believe that the SNN architecture has the potential to be of wide applicability in a broad class of matrix learning problems. In particular, we believe that the SNN architecture would be natural for the study of rank (or e-rank) minimization effect of implicit regularization in deep matrix/tensor factorization problems, especially given the fact that e-rank is essentially a spectrally defined concept. + +Let $X^{\star} \in \RR^{d_1 \times d_2}$ be an unknown rectangular matrix that we aim to recover. Let $A_1, \dots, A_m \in \RR^{d_1 \times d_2}$ be $m$ measurement matrices, and the label vector $y \in \RR^m$ is generated by $$\begin{equation} + y_i = \langle A_i, X^{\star} \rangle, +\end{equation}$$ where $\langle A, B \rangle = \tr(A^\top B)$ denotes the Frobenius inner product. We consider the following squared loss objective $$\begin{equation} + \label{eq:loss} + \ell(X) := \dfrac{1}{2} \sum_{i=1}^m (y_i - \langle A_i, X \rangle)^2. +\end{equation}$$ This setting covers problems including matrix completion (where the $A_i$'s are indicator matrices), matrix sensing from linear measurements, and multi-task learning (in which the columns of $X$ are predictors for the tasks, and $A_i$ has only one non-zero column). We are interested in the regime where $m \ll d_1 \times d_2$, i.e., the number of measurements is much less than the number of entries in $X^{\star}$, in which case [\[eq:loss\]](#eq:loss){reference-type="ref" reference="eq:loss"} is under-determined with many global minima. Therefore, merely minimizing [\[eq:loss\]](#eq:loss){reference-type="ref" reference="eq:loss"} does not guarantee correct recovery or good generalization. + +Following previous works, instead of working with $X$ directly, we consider a non-linear factorization of $X$ as follows $$\begin{equation} + \label{eq:X} + X = \sum_{k=1}^{K} \alpha_k \Gamma(U_k V_k^\top), +\end{equation}$$ where $\alpha \in \RR$, $U_k \in \RR^{d_1 \times d}, V_k \in \RR^{d_2 \times d}$, and the matrix-valued function $\Gamma:\RR^{d_1 \times d_2} \rightarrow \RR^{d_1 \times d_2}$ transforms a matrix by applying a nonlinear real-valued function $\gamma(\cdot)$ on its singular values. We focus on the over-parameterized setting $d \geq d_2 \geq d_1$, i.e., the factorization does not impose any rank constraints on $X$. Let $\alpha = \{ \alpha_1, \dots, \alpha_K \}$ be the collection of the $\alpha_k$'s. Similarly, we define $U$ and $V$ to be the collections of $U_k$'s and $V_k$'s. + +For each $k \in [K]$, let $\alpha_k(t), U_k(t), V_k(t)$ denote the trajectories of gradient flow, where $\alpha_k(0), U_k(0), V_k(0)$ are the initial conditions. Consequently, $X(t)=\sum_{k=1}^{K} \alpha_k(t) \Gamma(U_k(t) V_k(t)^\top)$. The dynamics of gradient flow is given by the following differential equations, for $k \in [K]$ $$\begin{align} + \nonumber + \alphadot_k = -\nabla_{\alpha_k} \ell(X(t)), \\ \nonumber + \Udot_k = -\nabla_{U_k} \ell(X(t)), \\ \label{eq:gradient-flow} + \Vdot_k = -\nabla_{V_k} \ell(X(t)). +\end{align}$$ + +# Method + +In this work, we contribute a novel neural network architecture, called the Spectral Neural Network (*abbrv.* SNN), that is particularly suitable for matrix learning problems. At the fundamental level, the SNN architecture entails an application of a non-linear activation function on a matrix-valued input in the spectral domain. This may be followed by a linear combination of several such spectrally transformed matrix-structured data. + +To be precise, let us focus on an elemental neuron, which manipulates a single matrix-valued input $X$. If we have a singular value decomposition $X = \Phi \bar{X} \Psi^\top$, where $\Phi,\Psi$ are orthogonal matrices and $\bar{X}$ is the diagonal matrix of singular values of $X$. Let $\gamma$ be any activation function of choice. Then the elemental neuron acts on $X$ as follows : $$\begin{equation} + X \mapsto \Phi ~ \gamma(\bar{X}) ~ \Psi^\top, +\end{equation}$$ where $\gamma(\bar{X})$ is a diagonal matrix with the non-linearity $\gamma$ applied entrywise to the diagonal of $\bar{X}$. + +
+ +
Visualization of the anatomy of an SNN block and a depth-D SNN architecture. Each SNN block takes as input K matrices and outputs one matrix, both the input and output matrices are of size ℝd1 × d2. In layer i of the SNN, there are Li blocks, which aggregate matrices from the previous layers to produce Li output matrices as inputs for the next layer. The number of input matrices to a block equals the number of neurons in the previous layer. For example, blocks in layer 1 have K = L0, blocks in layer 2 have K = L1, and blocks in layer i have K = Li − 1.
+
+ +A *block* in the SNN architecture comprises of $K \ge 1$ elemental neurons as above, taking in $K$ matrix-valued inputs $X_1,\ldots,X_K$. Each input matrix $X_i$ is then individually operated upon by an elemental neuron, and finally, the resulting matrices are aggregated linearly to produce a matrix-valued output for the block. The coefficients of this linear combination are also parameters in the SNN architecture, and are to be learned during the process of training the network. + +The comprehensive SNN architecture is finally obtained by combining such blocks into multiple layers of a deep network, as illustrated in [1](#fig:snn-architecture){reference-type="ref" reference="fig:snn-architecture"}. + +For the purposes of theoretical analysis, in the present article, we specialize the SNN architecture to focus on the setting of (quasi-) commuting measurement matrices $A_i$ and spectral near zero initialization; c.f. Assumptions [\[assump:1\]](#assump:1){reference-type="ref" reference="assump:1"} and [\[assump:2\]](#assump:2){reference-type="ref" reference="assump:2"} below. Similar settings have attracted considerable attention in the literature, including the foundational works of [@gunasekar2017implicit] and [@arora2019implicit]. Furthermore, our analysis holds under very general analytical requirements on the activation function $\gamma$; see Assumption [\[assump:3\]](#assump:3){reference-type="ref" reference="assump:3"} in the following. + +::: assump +[]{#assump:1 label="assump:1"} The measurement matrices $A_1, \dots, A_m$ share the same left and right singular vectors. Specifically, there exists two orthogonal matrices $\Phi \in \RR^{d_1 \times d_1}$ and $\Psi \in \RR^{d_2 \times d_2}$, and a sequence of (rectangular) diagonal matrices [^1] $\Abar_1, \dots, \Abar_m \in \RR^{d_1 \times d_2}$ such that for any $i \in [m]$, we have $$\begin{equation} +\label{eq:assump-1a} + A_i = \Phi \Abar_i \Psi^\top. +\end{equation}$$ Let $\sigma^{(i)}$ be the vector containing the singular values of $A_i$, i.e., $\sigma^{(i)}=\diag(\Abar_i)$. Furthermore, we assume that there exist real coefficients $a_1,\dots,a_m$ that $$\begin{equation} +\label{eq:assump-1b} + a_1 \sigma^{(1)} + \dots + a_m \sigma^{(m)} = \openone. +\end{equation}$$ We let $\Xbar^{\star}=\Phi^\top X^{\star}\Psi$ and $\sigma^{\star}$ be the vector containing the diagonal elements of $\Xbar^{\star}$, i.e., $\sigma^\star=\diag(\Xbar^\star)$. Without loss of generality, we may also assume that $\sigma^\star$ is coordinatewise non-zero. This can be easily ensured by adding the rectangular identity matrix (c.f. Appendix D) $c I_{d_1 \times d_2}$ to $X^*$ for some large enough positive number $c$. +::: + +Eq. [\[eq:assump-1a\]](#eq:assump-1a){reference-type="ref" reference="eq:assump-1a"} postulates that the measurement matrices share the same (left- and right-) singular vectors. This holds if and only if the measurement matrices pair-wise *quasi-commute* in the sense that for any $i, j \in [m]$, we have $$\begin{equation} +\label{eq:quasi-comm} + A_i A_j^\top = A_j A_i^\top, \quad + A_i^\top A_j = A_j^\top A_i. +\end{equation}$$ A natural class of examples of such quasi-commuting measurement matrices comes from families of commuting projections. In such a scenario Eq. [\[eq:assump-1b\]](#eq:assump-1b){reference-type="ref" reference="eq:assump-1b"} stipulates that these projections cover all the coordinate directions, which may be related conceptually to a notion of the measurements being sufficiently informative. For example, in this setting, Eq. [\[eq:assump-1b\]](#eq:assump-1b){reference-type="ref" reference="eq:assump-1b"} would entail that the trace of $X^\star$ can be directly computed on the basis of the measurements. + +Eq. [\[eq:assump-1b\]](#eq:assump-1b){reference-type="ref" reference="eq:assump-1b"} acts as a regularity condition on the singular values of the measurement matrices. For example, it prohibits peculiar scenarios where $\sigma^{(i)}_1 = 0$ for all $i$, i.e., all measurement matrices have $0$ as their smallest singular values, which makes it impossible to sense the smallest singular value of $X^\star$ from linear measurements. + +Note that $$\begin{align} + y_i &= \langle A_i, X^{\star} \rangle = \Tr(A_i^\top X^{\star})\\ + &= \Tr(\Psi\Abar_i\Phi^\top X^{\star}) = \Tr(\Abar_i\Phi^\top X^{\star} \Psi) = \langle \Abar_i, \Phi^\top X^{\star} \Psi \rangle, +\end{align}$$ where in the above we use the fact that $\Abar_i=\Abar_i^\top$ (since $\Abar_i$ is diagonal) and the cyclic property of trace. We have $$\begin{align} + \label{eq:y_i} + y_i = \langle \Abar_i, \Xbar^{\star} \rangle = \langle \sigma^{(i)}, \sigma^{\star} \rangle = \sigma^{(i)\top} \sigma^{\star}, +\end{align}$$ where the second equality is due to $\Abar_i$ being diagonal. + +We further define vectors $z^{(k)}$ and three matrices $B$, $\Cc$, and $H$ as follows $$\begin{align*} + z_i^{(k)} &= [\Ubar_k]_{ii} [\Vbar_k]_{ii} \\ + B &= \begin{bmatrix} \sigma^{(1)} \, | \, \dots \, | \, \sigma^{(m)} \end{bmatrix} \in \RR^{d_1 \times m} \\ + \Cc &= B B^\top \in \RR^{d_1 \times d_1} \\ + H &= \begin{bmatrix} \gamma(z^{(1)}) \, | \, \dots \, | \, \gamma(z^{(K)}) \end{bmatrix} \in \RR^{d_1 \times K}. +\end{align*}$$ Under these new notations, we can write the label vector $y$ as $y = B^\top \sigma^{\star}$. + +::: assump +[]{#assump:2 label="assump:2"} **(Spectral Initialization)** Let $\Phi$ and $\Psi$ be the matrices containing the left and right singular vectors of the measurement matrices from Assumption [\[assump:1\]](#assump:1){reference-type="ref" reference="assump:1"}. Let $G \in \RR^{d \times d}$ is any arbitrary orthogonal matrix. We initialize $X(0)$ such that the following condition holds: for any $k=1,\dots,K$, we have + +1. $U_k(0) = \Phi \Ubar_k(0) G$ and $V_k = \Psi \Vbar(0) G$, and + +2. $\Ubar_k(0)$ and $\Vbar_k(0)$ are diagonal, and + +3. $\sum_{k=1}^{K} \alpha_k \gamma(\Ubar_k(0)_{ii}\Vbar_k(0)_{ii}) \leq \sigma^\star_i$ for any $i=1,\dots,d_1$. +::: + +Assumption [\[assump:2\]](#assump:2){reference-type="ref" reference="assump:2"}, especially part (c) therein, may be thought of as a much more generalized counterpart of the "near-zero" initialization which is widely used in the related literature ([@gunasekar2017implicit; @li2018algorithmic; @arora2019implicit]). A direct consequence of Assumption [\[assump:2\]](#assump:2){reference-type="ref" reference="assump:2"} is that at initialization, the matrix $X(0)$ has $\Phi$ and $\Psi$ as its left and right singular vectors. As we will see later, this initialization imposes a distinctive structure on the gradient flow dynamics, allowing for an explicit analytical expression for the flow of each component of $X$. + +::: assump +[]{#assump:3 label="assump:3"} The function $\gamma: \RR \rightarrow \RR$ is bounded between $[0, 1]$, and is differentiable and non-decreasing on $\RR$. +::: + +Assumption $\ref{assump:3}$ imposes regularity conditions on the non-linearity $\gamma$. Common non-linearities that are used in deep learning such as Sigmoid, ReLU or tanh satisfy the differentiability and non-decreasing conditions, while the boundedness can be achieved by truncating the outputs of these functions if necessary. + +Our first result provides a compact representation of the gradient flow dynamics in suitable coordinates. The derivation of this dynamics involves matrix differentials utilizing the Khatri-Rao product $\Psi \boxtimes \Phi$ of the matrices $\Psi$ and $\Phi$ (see Eq. [\[eq:khatri-rao\]](#eq:khatri-rao){reference-type="ref" reference="eq:khatri-rao"} in Appendix [\[app:proof-thm-gradient-flow\]](#app:proof-thm-gradient-flow){reference-type="ref" reference="app:proof-thm-gradient-flow"} of the supplement). + +::: thm +[]{#thm:gradent-flow label="thm:gradent-flow"} Suppose Assumptions [\[assump:1\]](#assump:1){reference-type="ref" reference="assump:1"}, [\[assump:2\]](#assump:2){reference-type="ref" reference="assump:2"}, [\[assump:3\]](#assump:3){reference-type="ref" reference="assump:3"} hold. Then the gradient flow dynamics in [\[eq:gradient-flow\]](#eq:gradient-flow){reference-type="ref" reference="eq:gradient-flow"} are $$\begin{align*} + \balphadot &= H^\top \Cc (\sigma^\star - H \alpha), \\ + \Udot_k &= \Phi L_k \Psi^\top V_k, \text{ and} \\ + \Vdot_k &= (\Phi L_k \Psi^\top)^\top U_k, +\end{align*}$$ where $L_k \in \RR^{d_1 \times d_2}$ is a diagonal matrix whose diagonal is given by $$\begin{align} + \nonumber + \diag(L_k) = \lambda^{(k)} &= [\lambda_1^{(k)}, \dots, \lambda_{d_1}^{(k)}]^\top \\\label{eq:lambda} + &= \alpha_k \gamma^\prime(z^{(k)}) \circ \Cc (\sigma^\star - H \alpha). +\end{align}$$ +::: + +::: proof +*Proof.* (Main ideas -- full details in Appendix [\[app:proof-thm-gradient-flow\]](#app:proof-thm-gradient-flow){reference-type="ref" reference="app:proof-thm-gradient-flow"}). We leverage the fact that the non-linearity $\gamma(\cdot)$ only changes the singular values of the product matrix $U_k V_k^\top$ while keeping the singular vectors intact. Therefore, the gradient flow in [\[eq:gradient-flow\]](#eq:gradient-flow){reference-type="ref" reference="eq:gradient-flow"} preserves the left and right singular vectors. Furthermore, by Assumption [\[assump:2\]](#assump:2){reference-type="ref" reference="assump:2"}, $U_k V_k^\top$ has $\Phi$ and $\Psi$ as its left and right singular vectors at initialization, which remains the same throughout. This property also percolates to $\nabla_{U_k V_k^\top} \ell$. Mathematically speaking, $\nabla_{U_k V_k^\top} \ell$ becomes diagonalizable by $\Phi$ and $\Psi$, i.e., $$\begin{equation*} + \nabla_{U_k V_k^\top} \ell = \Phi \Lambda_k \Psi^\top +\end{equation*}$$ for some diagonal matrix $\Lambda_k$. It turns out that $\Lambda_k = L_k$ as given in the statement of the theorem. In view of Eq. [\[eq:gradient-flow\]](#eq:gradient-flow){reference-type="ref" reference="eq:gradient-flow"}, this explains the expressions for $\Udot_k$ and $\Vdot_k$. Finally, since $\alpha_k$ is a scalar, the partial derivative of $\ell$ with respect to $\alpha_k$ is relatively straightforward to compute. ◻ +::: + +Theorem [\[thm:gradent-flow\]](#thm:gradent-flow){reference-type="ref" reference="thm:gradent-flow"} provides closed-form expressions for the dynamics of the individual components of $X$, namely $\alpha_k, U_k$ and $V_k$. We want to highlight that the compact analytical expression and the simplicity of the gradient flow dynamics on the components are a direct result of the spectral non-linearity. In other words, if we use the conventional element-wise non-linearity commonly used in deep learning, the above dynamics will be substantially more complicated, containing several Hadamard products and becoming prohibitively harder for theoretical analysis. + +As a direct corollary of Theorem [\[thm:gradent-flow\]](#thm:gradent-flow){reference-type="ref" reference="thm:gradent-flow"}, the gradient flow dynamics on $\Ubar_k$ and $\Vbar_k$ are $$\begin{equation} + \label{eq:dynamics-uvbar} + \Ubardot_k = L_k \Vbar_k, \quad + \Vbardot_k = L_k^\top \Ubar_k. +\end{equation}$$ Under Assumption [\[assump:2\]](#assump:2){reference-type="ref" reference="assump:2"}, $\Ubar_k(0)$ and $\Vbar_k(0)$ are diagonal matrices. From the gradient flow dynamics in Eq. [\[eq:dynamics-uvbar\]](#eq:dynamics-uvbar){reference-type="ref" reference="eq:dynamics-uvbar"}, and recalling that the $L_k$'s are diagonal, we infer that $\Ubardot_k(0)$ and $\Vbardot_k(0)$ are also diagonal. Consequently, $\Ubar_k(t)$ and $\Vbar_k(t)$ remain diagonal for all $t\geq 0$ since the gradient flow dynamics in Eq. [\[eq:dynamics-uvbar\]](#eq:dynamics-uvbar){reference-type="ref" reference="eq:dynamics-uvbar"} does not induce any change in the off-diagonal elements. Thus, $\Ubar_k(t) \Vbar(t)_k^\top$ also remains diagonal throughout. + +A consequence of the spectral initialization is that the left and right singular vectors of $X(t)$ stay constant at $\Phi$ and $\Psi$ throughout the entire gradient flow procedure. To this end, the gradient flow dynamics is completely determined by the evolution of the singular values of $X(t)$, i.e., $H \alpha$. The next result characterizes the convergence of the singular values of $X(t)$. + +::: thm +[]{#prop:convergence-rate label="prop:convergence-rate"} Under Assumptions [\[assump:1\]](#assump:1){reference-type="ref" reference="assump:1"} and [\[assump:2\]](#assump:2){reference-type="ref" reference="assump:2"}, for any $i=1,\dots,d_1$, there are constants $\eta_i,C_i>0$ such that we have : $$\begin{equation*} + 0 \le \sigma^\star_i - (H(t)\alpha(t))_i \le C_i e^{- \eta_i t}. +\end{equation*}$$ On the other hand, we have the lower bound $$\begin{equation*} + \|\sigma^\star - (H(t)\alpha(t)) \|_2 \ge C e^{-\eta t}, +\end{equation*}$$ for some constants $\eta,C>0$. +::: + +::: proof +*Proof.* (Main ideas -- full details in Appendix [\[app:proof-prop-convergence-rate\]](#app:proof-prop-convergence-rate){reference-type="ref" reference="app:proof-prop-convergence-rate"}). By part (c) of Assumption [\[assump:2\]](#assump:2){reference-type="ref" reference="assump:2"}, at initialization, we have that $H(0)\alpha(0) \leqentry \sigma^{\star}$, in which the symbol $\leqentry$ denotes the element-wise less than or equal to relation. Therefore, to prove Theorem [\[prop:convergence-rate\]](#prop:convergence-rate){reference-type="ref" reference="prop:convergence-rate"}, it is sufficient to show that $H(t)\alpha(t)$ is increasing to $\sigma^{\star}$ element-wise at an exponential rate. To achieve that, we show that the evolution of $H \alpha$ over time can be expressed as $$\begin{align*} + \dot{(H \alpha)} = 4\sum_{k=1}^{K} \alpha_k^2 \cdot \Big(\gamma^\prime(&z^{(k)})^2 \circ \Cc (\sigma^\star - H \alpha) \\ + &\circ \big(\int \lambda^{(k)} \circ z^{(k)} dt + + C^{(k)}\big)\Big) \\ + &\quad\,\,\qquad+ H H^\top \Cc (\sigma^\star - H \alpha). +\end{align*}$$ By definition, the matrix $B$ contains the singular values of the $A_i$'s, and therefore its entries are non-negative. Consequently, since $\Cc = BB^\top$, the entries of $\Cc$ are also non-negative. By Assumption [\[assump:3\]](#assump:3){reference-type="ref" reference="assump:3"}, we have that $\gamma(\cdot)\in[0,1]$, thus the entries of $H$ are non-negative. Finally, by Assumption [\[assump:2\]](#assump:2){reference-type="ref" reference="assump:2"}, we have $H(0)\alpha(0) < \sigma^\star$ entry-wise. For these reasons, each entry in $\dot{(H\alpha)}$ is non-negative at initialization, and indeed, for each $i$, the quantity $(H \alpha)_i$ is increasing as long as $(H \alpha)_i < \sigma^*_i$ . As $H \alpha$ approaches $\sigma^\star$, the gradient $\dot{(H \alpha)}$ decreases. If it so happened that $H \alpha = \sigma^\star$ at some finite time, then $\dot{(H \alpha)}$ would exactly equal $0$, which would then cause $H \alpha$ to stay constant at $\sigma^*$ from then on. + +Thus, each $(H \alpha)_i$ is non-decreasing and bounded above by $\sigma^*_i$, and therefore must converge to a limit $\le \sigma^*_i$. If this limit was strictly smaller than $\sigma^*_i$, then by the above argument $(H \alpha)_i$ would be still increasing, indicating that this cannot be the case. Consequently, we may deduce that $$\begin{align*} + \lim_{t \rightarrow \infty} H(t) \alpha(t) = \sigma^\star. +\end{align*}$$ + +It remains to show that the convergence is exponentially fast. To achieve this, we show in the detailed proof that each entry of $\dot{H\alpha}$ is not only non-negative but also bounded away from $0$, i.e., $$\begin{align*} + \dot{(H \alpha)}_i \geq \eta_i (\sigma^\star_i - (H \alpha)_i), +\end{align*}$$ for some constant $\eta_i > 0$. This would imply that $H \alpha$ converges to $\sigma^{\star}$ at an exponential rate. ◻ +::: + +The limiting matrix output by the network is, therefore, $\Phi \Diag(\sigma^*) \Psi^\top$, and given the fact that $\sigma^*=\diag(\Phi^\top X^* \Psi)$, this would be the best approximation of $X^*$ among matrices with (the columns of) $\Phi$ and $\Psi$ as their left and right singular vectors. This is perhaps reasonable, given the fact that the sensing matrices $A_i$ also share the same singular vectors, and it is natural to expect an approximation that is limited by their properties. In particular, when the $A_i$ are symmetric and hence commuting, under mild linear independence assumptions, $\Phi \Diag(\sigma^*) \Psi^\top$ would be the best approximation of $X^*$ in the algebra generated by the $A_i$-s, which is again a natural class given the nature of the measurements. + +We are now ready to rigorously demonstrate the phenomenon of implicit regularization in our setting. To this end, following the gradient flow dynamics, we are interested in the behavior of $X_\infty$ in the limit when time goes to infinity. + +::: thm +[]{#thm:implicit-regularization label="thm:implicit-regularization"} Let $X_\infty=\lim_{t \rightarrow \infty} X(t)$. Under Assumptions [\[assump:1\]](#assump:1){reference-type="ref" reference="assump:1"} and [\[assump:2\]](#assump:2){reference-type="ref" reference="assump:2"}, the following hold: + +1. $\ell(X_\infty) = 0$, and + +2. $X_\infty$ solves the optimization problem $$\begin{equation} + \label{eq:nuclear-norm-opt} + \min_{X \in \RR^{d_1 \times d_2}} \lVert X \rVert_* \quad \textrm{subject to} \quad y_i = \langle A_i, X \rangle \,\,\forall i \in [m]. + \end{equation}$$ +::: + +::: proof +*Proof.* (Main ideas -- full details in Appendix [\[app:proof-thm-implicit-regularization\]](#app:proof-thm-implicit-regularization){reference-type="ref" reference="app:proof-thm-implicit-regularization"}). A direct corollary of Theorem [\[prop:convergence-rate\]](#prop:convergence-rate){reference-type="ref" reference="prop:convergence-rate"} is that $$\begin{align*} + H(\infty)\alpha(\infty) = \sigma^{\star}. +\end{align*}$$ By some algebraic manipulations, we can show that the limit of $X$ takes the form $$\begin{align*} + X_{\infty} = \Phi \Big[\Diag( + \sigma^\star)\Big] \Psi^\top. +\end{align*}$$ Now, let us look at the prediction given by $X_\infty$. For any $i = 1,\dots,m$, we have $$\begin{align*} + \langle A_i, X_\infty \rangle &= \langle \Phi \Abar_i \Psi^\top, \Phi \Diag(\sigma^\star) \Psi^\top \rangle = \langle \sigma^{(i)}, \sigma^\star\rangle = y_i, +\end{align*}$$ where the last equality holds due to [\[eq:y_i\]](#eq:y_i){reference-type="ref" reference="eq:y_i"}. This implies that $\ell(X_\infty) = 0$, proving (a). + +To prove (b), we will show that $X_\infty$ satisfies the Karush-Kuhn-Tucker (KKT) conditions of the optimization problem stated in Eq. [\[eq:nuclear-norm-opt\]](#eq:nuclear-norm-opt){reference-type="ref" reference="eq:nuclear-norm-opt"}. The conditions are $$\begin{align*} + &\exists \nu \in \RR^{m} \textrm{ s.t. } \quad \forall i \in [m], \langle A_i, X \rangle = y_i \quad \textrm{and} \\ + & \nabla_X \lVert X \rVert_* + \sum_{i=1}^{m} \nu_i A_i = 0. +\end{align*}$$ The solution matrix $X_\infty$ satisfies the first condition as proved in part (a). As for the second condition, note first that the gradient of the nuclear norm of $X_\infty$ is given by $$\begin{equation*} + \nabla \lVert X_\infty \rVert = \Phi \Psi^\top. +\end{equation*}$$ Therefore the second condition becomes $$\begin{align*} + \Phi \Psi^\top + \sum_{i=1}^{m} \nu_i A_i = 0 + \\ \Leftrightarrow\quad \Phi \Big( I - \sum_{i=1}^{m} \nu_i \Abar_i \Big) \Psi^\top = 0 + \quad\Leftrightarrow\quad B \nu = \openone. +\end{align*}$$ However, by Assumption [\[assump:1\]](#assump:1){reference-type="ref" reference="assump:1"}, the vector $\openone$ lies in the column space of $B$, which implies the existence of such a vector $\nu$. This concludes the proof of part (b). ◻ +::: + +In this section, we present numerical studies to complement our theoretical analysis. + +We highlight that gradient flow can be viewed as gradient descent with an infinitesimal learning rate. Therefore, the gradient flow model serves as a reliable proxy for studying gradient descent when the learning rate is sufficiently small. Throughout our experiments, we shall consider gradient descent with varying learning rates, and demonstrate that the behavior suggested by our theory is best achieved using small learning rates. + +The synthetic data used in our experiment is generated by $X^{\star} = \tilde{X}\tilde{X}^{T} \in\mathbb{R}^{10\times10}$ where each entry of $\tilde{X}\in\mathbb{R}^{10\times 6}$ is randomly sampled from uniformly on $[0,1]$ and normalize so that $\| X^{\star} \| =1$. + +In the first experiment, we follow Assumption [\[assump:1\]](#assump:1){reference-type="ref" reference="assump:1"} and Assumption [\[assump:2\]](#assump:2){reference-type="ref" reference="assump:2"} to generate measurement matrices and initialize our network. For every $k=1,K=2$, we initialize $\Ubar(0)$ and $\Vbar(0)$ to be diagonal matrices, whose diagonal entries are sampled uniformly from $[0,1]$, sorted in descending order, and set $U(0)=\Phi\Ubar(0)G,V(0)=\Psi^{\top}\Vbar(0)G$ where $G$ is a random orthogonal matrix. The rest of the parameters are sampled uniformly from $[0,1]$, and we use the hyperbolic tangent $\gamma(x) = (e^{x}-e^{-x})/(e^{x}+e^{-x})$ as activation. + +For every measurement matrix $A_i$ where $i = 1, \dots, m$, with $m = 50$, we sample each entry of the diagonal matrix $\Abar_i$ from the uniform distribution on $[0, 1]$, sort them in decreasing order, and set $A_i = \Phi \Abar_i \Psi^\top$. The measurements are recorded as $y_i = \langle A_i, X^\star \rangle +10^{-2}\epsilon_i$, where $\epsilon_i$ is sampled from the standard Gaussian distribution, $i = 1, \ldots, m$. + +We illustrate the numerical results in Figure [2](#fig:exp-1){reference-type="ref" reference="fig:exp-1"}. One can observe that singular values of the solution matrices converge to $\sigma^\star$ at an exponential rate as suggested by Theorem [\[prop:convergence-rate\]](#prop:convergence-rate){reference-type="ref" reference="prop:convergence-rate"}, while the nuclear norm values also converge to $\| X^{\star} \| = 1$. We re-emphasize that our theoretical results are for gradient flow, which only acts as a good surrogate to study gradient descent when the learning rates are infinitesimally small. As a result, our theory cannot characterize the behavior of gradient descent algorithm with substantially large learning rates. We show the evolution of the nuclear norm over time. Interestingly, but perhaps not surprisingly, the choice of the learning rate dictates the speed of convergence. Moderate values of the learning rate seem to yield the quickest convergence to stationarity. + +
+ +
The evolution of training error (left panel), nuclear norm (middle panel), and 3 leading singular values (right panel) over time with learning rates (lr) of different magnitudes.
+
diff --git a/2403.14155/main_diagram/main_diagram.drawio b/2403.14155/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..c2101a81f6436bdc9ddb34d82f95aceb579b1f90 --- /dev/null +++ b/2403.14155/main_diagram/main_diagram.drawio @@ -0,0 +1,70 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/2403.14155/paper_text/intro_method.md b/2403.14155/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..cd6cdb1b2dcd273769430323f62dddc692c1c897 --- /dev/null +++ b/2403.14155/paper_text/intro_method.md @@ -0,0 +1,75 @@ +# Introduction + +Recent advancements in text-to-image (T2I) generation, especially diffusion models [@rombach2022ldm; @balaji2023ediffi; @saharia2022imagen; @ramesh2022dalle2; @nichol2022glide] have opened up a new era of image creation. Subject-driven generation [@ruiz2023dreambooth; @gal2022textual_inversion; @tewel2023perfusion; @qiu2023controlling; @voynov2023p] aims to generate novel images featuring a specific subject provided by the user. The common approach represents the subject as a new pseudo-word ($\text{S}^*$) in the textual embedding space of the text encoder. They optimize a pseudo-word by updating the textual embedding of the pseudo-word [@gal2022image] or the diffusion model's parameters [@ruiz2023dreambooth; @tewel2023key]. However, both approaches require per-subject optimization, posing challenges for real-time applications and multi-subject customization. In response to this challenge, zero-shot customization methods [@li2024blip; @wei2023elite; @chen2023photoverse; @shi2023instantbooth; @ma2023subjectdiffusionopen; @jia2023tamingencoder; @yuan2023customnet; @xiao2023fastcomposer] have been proposed. + +They adopt the mappers (, MLP network [@wei2023elite; @yuan2023customnet; @xiao2023fastcomposer], adapter [@chen2023photoverse; @shi2023instantbooth; @ma2023subjectdiffusionopen] or multi-modal encoder [@li2024blip]) to transform a subject's image into the visual embedding, which is subsequently utilized along with the textual embedding. The visual embedding provides representative information about the subject's identity from the input image, eliminating the need for additional training processes. To improve the user's convenience, most zero-shot customization methods [@li2024blip; @wei2023elite; @chen2023photoverse; @ma2023subjectdiffusionopen; @jia2023tamingencoder; @yuan2023customnet; @xiao2023fastcomposer] encode a single image as the input. In line with this practice, we focus on single-image-based zero-shot customization in this paper. + +
+ +
Our method deals with the challenges of pose variation in zero-shot customization methods , (i) strong adherence to the pose in the input image and (ii) loss in the subject’s identity. Our method paves the way for a more diverse, efficient, and lively subject-driven generation.
+
+ +When dealing with a single image, the existing methods focus on separating the subject's identity from other irrelevant details within an image, as the input image typically encompasses more than just the subject's identity. To address this challenge, they employ the subject's segmentation mask [@li2024blip; @wei2023elite; @ma2023subjectdiffusionopen; @jia2023tamingencoder; @xiao2023fastcomposer] or the visual embedding from the deepest layers of an encoder [@wei2023elite]. However, while those techniques effectively separate the subject from the background or other objects within the image, disentangling the subject's pose from its identity is almost impossible as these aspects are tightly intertwined within the same pixels. We term this inevitable phenomenon as the *pose-identity entanglement* within visual embedding, which becomes apparent when attempting to change the pose of the subject while preserving the subject's identity. Regarding this, we have identified two significant problems when modifying the subject's pose with a text prompt, namely the (i) *pose bias* and (ii) *identity loss*: + +1. The generated images tend to either maintain the original pose of the subject or exhibit some unnaturalness that falls between the pose derived from the text prompt and the one presented in the input image. + +2. The subject in the generated images partially loses its identity, appearing with a different color or body shape. + +In this paper, we focus on resolving the pose bias and identity loss described above. They are imperative tasks for advancing towards a more diverse and desirable customization, while highly challenging due to the pose-identity entanglement. We start from unveiling the more direct and addressable cause of the two problems. We found that using only the textual embedding generates the subject in various poses faithfully complying with the text prompt (without *pose bias*) while using only the visual embedding results in the images where the subject retains its whole identity (without *identity loss*). This leads us to assume that a conflict exists between the visual and textual embeddings, from which the two problems mainly originate. We provide the result images in the supplementary materials. Regarding pose-identity entanglement, the input image carries the visual features of *the subject in a specific pose*. Therefore, when generating a novel image of the subject in a different pose, the input image and text prompt provide two concurrent but conflicting embeddings to the model with different pose information. As a result, the visual embedding readily interferes with the textual embedding, causing the *pose bias*. Conversely, textual embedding can also interfere with visual embedding, affecting the subject's identity and resulting in *identity loss*. + +To alleviate this conflict, we propose **contextual embedding orchestration** and **self-attention swap** that effectively resolve each problem. The former involves adjusting the visual embedding to align better with the textual embedding. This adjustment is achieved by orthogonalizing the visual embedding vector to the subspace of the textual embedding vectors, essentially reducing the interference of the visual embedding. The latter adopts another denoising process guided by the visual-only embedding that fundamentally evades the collision between the visual and textual embedding and aggregates clean information about the subject. By swapping the self-attention between the original and the visual-only-guided diffusion models, we could successfully maintain the subject's identity in newly generated images. Our method is generic and easily applicable to any zero-shot customization method that utilizes visual and textual embeddings, as it does not require an additional tuning process. We demonstrate that our method significantly improves the pose variation of the baseline without compromising its performance in pose-irrelevant scenarios, , placement of novel objects or changing the subject's texture. [1](#fig:intro){reference-type="ref+label" reference="fig:intro"} illustrates that our method effectively resolves the pose bias. We also provide the results regarding identity loss in [\[fig:qualitative,fig:qualitative_elitr\]](#fig:qualitative,fig:qualitative_elitr){reference-type="ref+label" reference="fig:qualitative,fig:qualitative_elitr"}, and supplementary materials. + +Our contributions are summarized as: + +- We unveil the problem of pose bias and identity loss in zero-shot customization and shed light on the discord among the visual and textual embeddings. + +- Our proposed method effectively resolves the pose bias and identity loss, offering highly diverse and pose-variant subject generation. + +- Our method is readily applicable to any zero-shot customization methods that incorporate visual and textual embeddings. + +# Method + +In this work, we employ text-to-image latent diffusion model (LDM) [@rombach2022ldm]. The denoising process is implemented in the latent space using the autoencoder structure with the encoder $\mathcal{E}(\cdot)$ and the decoder $\mathcal{D}(\cdot)$. Specifically, an image $x$ is projected to a latent representation $z=\mathcal{E}(x)$, and decoded back to the image space giving $\Tilde{x}=\mathcal{D}(z)=\mathcal{D}(\mathcal{E}(x))$, reconstructing $x$, , $\Tilde{x} \approx x$. Given the pre-trained autoencoder, the latent diffusion model $\epsilon_\theta(x_\tau,\tau); \tau=1 \cdots T$ is trained with the following objective where $\textbf{c}$ represents the contextual embedding of textual/visual condition generally obtained from the pre-trained CLIP encoder [@radford2021clip]: $$\begin{equation} + \label{eq:ldm_train} + L_{LDM} = \mathbb{E}_{x \sim p(x), \epsilon \sim \mathcal{N}(0,I),\textbf{c},\tau \sim \text{uniform}(1,T)}[\| \epsilon - \epsilon_\theta(z_\tau(x),\tau,\textbf{c}) \|^2_2]. +\end{equation}$$ During inference, $z_T \sim \mathcal{N}(0,I)$ is iteratively denoised to the initial representation $z_0(x) = \mathcal{E}(x)$. + +Text prompts act as a condition on LDM through the cross-attention mechanism. The latent spatial feature $f \in \mathbb{R}^{l \times h}$ is projected to produce the *query* $Q= f \cdot W_Q \in \mathbb{R}^{l \times d}$ while the text prompts are first encoded into the text embedding $\textbf{c} \in \mathbb{R}^{l_c \times h_c}$ and projected to yield the *key*, $K=\textbf{c} \cdot W_K \in \mathbb{R}^{l_c \times d}$, and *value*, $V=\textbf{c} \cdot W_V \in \mathbb{R}^{l_c \times d}$. Then the cross-attention is calculated as follows: $$\begin{equation} + \label{eq:attention} + \text{Attention}(Q,K,V) = \text{softmax}(QK^T / \sqrt{d}) \cdot V +\end{equation}$$ where $l$, $l_c$, $d$ and $h$ are spatial sequence length, context sequence length, dimension of key/query/value, and dimension of spatial feature/context respectively. Self-attetntion mechanism uses the same [\[eq:attention\]](#eq:attention){reference-type="ref+label" reference="eq:attention"} where the key and value are attained from latent spatial features $f$ instead of the contextual embedding $\textbf{c}$. + +
+ +
Visualization of discord among contextual embeddings. (a) The contextual embedding of zero-shot customization methods is generally composed of visual and textual embedding, each of which provides the subject’s identity and a novel context, respectively. (b) However, pose information in the visual embedding interferes with the textual embedding, resulting in biased images that are closely attached to the input image. Meanwhile, our orchestration resolves the conflict, generating images that are both identity-conservative and faithful to the text prompt.
+
+ +Single-image-based zero-shot customization methods [@li2024blip; @wei2023elite; @chen2023photoverse; @ma2023subjectdiffusionopen; @jia2023tamingencoder; @yuan2023customnet; @xiao2023fastcomposer] compose the contextual embedding by concatenating two heterogeneous embeddings; visual embedding $\textbf{v} \in \mathbb{R}^{M \times h_c}$ and textual embedding $\textbf{t} \in \mathbb{R}^{N\times h_c}$ ($\textbf{c} = [\textbf{v};\textbf{t}] = [v_1 \cdots v_M; t_1 \cdots t_N]$). Image features of a given subject's image are transformed into the visual embedding using the pre-trained mapper. The visual embedding is then combined with the textual embedding, engaging in the spatial features $z$ via cross-attention layers as in [\[eq:attention\]](#eq:attention){reference-type="ref+label" reference="eq:attention"} or additional adapters. We depict the overall zero-shot pipeline in  [2](#fig:discord){reference-type="ref+label" reference="fig:discord"}(a). + +Meanwhile, when modifying the subject's pose with a text prompt, the visual embedding, which includes pose information from the input image, conflicts with the textual embedding, rendering the images confined to the specific pose or partially losing the subject's identity. identity of the subject's concept. To verify the interference between the visual and textual embedding, we conduct an experiment utilizing three types of contextual embedding respectively: 1) contextual embedding composed solely of the visual embedding ($\textbf{c}_{\textbf{v},\varnothing} = [v_1 \cdots v_M; \varnothing]$), 2) solely of the textual embedding ($\textbf{c}_{\varnothing,\textbf{t}} = [\varnothing; t_1 \cdots t_N]$), and 3) the original combined contextual embedding ($\textbf{c} = [v_1 \cdots v_M; t_1 \cdots t_N]$). We generate 500 images conditioned on the respective contextual embedding and obtain their representations from the highest layer of VGG-16 [@simonyan2014very] as visualized in [2](#fig:discord){reference-type="ref+label" reference="fig:discord"}(b). Ideally, the generated images conditioned on the original contextual embedding should include the subject's identity and, at the same time, faithfully follow the text prompts. However, their image features exhibit a strong bias toward outputs from the visual-only embedding. The pre-existing pose information in visual embedding interferes with the textual embedding, restricting the generated images to be closely tied to the visual embedding, , maintaining the pose of the subject in the input images instead of constructing a novel pose as the textual embedding instructs. The result above clarifies that the visual embedding interferes with the textual embedding, confining the generated images away from text guidance. Conversely, the textual embedding also affects the visual embedding, harming the subject's identity. We include the analysis regarding the identity loss in supplementary materials. These results underline the necessity of adjusting the visual embedding to resolve this discord. + +To resolve the interference between the visual embedding $\textbf{v}$ and the textual embedding $\textbf{t}$ as presented in [4.1](#subsec:discord){reference-type="ref+label" reference="subsec:discord"}, we propose **orthogonal visual embedding** $\textbf{v}^\bot$ as follows: $$\begin{equation} + \label{eq:orthogonal} +v^\bot = v- v^{||} = v - \! \sum_{j=1, j \neq \text{sbj}}^{N} \langle \bar{t}_j, v \rangle \bar{t}_j , \quad \bar{t}_j = \text{normal}\left(t_j - \! \sum_{i=1, i\neq \text{sbj}}^{j-1} \langle \bar{t}_i, t_j \rangle \bar{t}_i \right) +\end{equation}$$ where $t_{\text{sbj}}$ represents the textual embedding corresponding to the pseudo-word (, 'S$^*$' in Fig. [1](#fig:intro){reference-type="ref" reference="fig:intro"}) and $\text{normal}(\cdot)$ means $l_2$ normalization[^1]. $\{\bar{t}_j \}_{j\neq \text{sbj}}$ are the basis vectors of the textual subspace, obtained by the Gram-Schmidt orthogonalization process. [\[eq:orthogonal\]](#eq:orthogonal){reference-type="ref+label" reference="eq:orthogonal"} can be viewed as breaking down a vector $v$ into two components, $v^\bot$ and $v^{||}$, of which $v^{||}$ resides in the textual subspace which is spanned by $\{\bar{t}_j \}_{j\neq \text{sbj}}$. + +We argue that $v^{||}$ causes the interference with $\textbf{t}$ when presented concurrently. The other component, $v^\bot$, is perpendicular to the textual subspace, embodying the essential information about the subject's identity alleviating the interference with $\textbf{t}$. Using $v^\bot$ instead of $v$, we could establish the new axes in the visual embedding that interplays more effectively with $\textbf{t}$, helping resolve the interference. The new contextual embedding $\textbf{c}^\bot = [v^\bot_1 \cdots v^\bot_M; t_1 \cdots t_N]$ is then incorporated with latent spatial features $f$ via cross-attention. Note that we also exclude the textual embedding of articles or the subject's class name from [\[eq:orthogonal\]](#eq:orthogonal){reference-type="ref+label" reference="eq:orthogonal"} when such tokens are used.  [3](#fig:method){reference-type="ref+label" reference="fig:method"}(a) illustrates the contextual embedding orchestration where the textual embedding $\textbf{t}$ effectively provides guidance with alleviated interference from the visual embedding $v^\bot$. + +We found that the conflict between visual and textual embeddings results in the loss of the subject's identity, when the textual embedding affects the visual embeddings. This phenomenon is still visible even after orchestration, due to the inevitable pose-identity entanglement; changing the visual embedding inevitably results in a change in identity. To alleviate the loss in the subject's identity, we propose a self-attention swap to retain essential information about the subject. Our intuition is based on the observation that when the visual-only embedding is used as the contextual embedding ($\textbf{c}_{\textbf{v},\varnothing} = [v_1 \cdots v_M; \varnothing]$), the output images are clean, effectively preserving the subject's identity, as it is free of the interference of the textual embedding. + +
+ +
The overall pipeline of our method. (a) To alleviate the pose-bias due to the visual embedding, we conduct Orchestration of the contextual embedding , adjust the visual embedding to be orthogonal to the textual embedding. Orchestration accurately guides the denoising process toward a direction that follows the pose directed by the text prompt. (b) Self-attention Swap obtains self-attention key and value from another denoising process guided by visual-only embedding, which offers the subject’s clean identity. After swapping, our method retains the subject’s identity while changing its pose faithfully following the text prompt.
+
+ +To effectively attain the subject's identity using the visual-only embedding while simultaneously maintaining a novel pose of the subject, we delve into the image editing methods. The image editing methods generally employ two denoising processes, one for a source image reconstruction and another for generating a new, edited image. The properties of the source image obtained during its reconstruction are transferred to the process of generating the edited image by swapping self-attention [@cao_2023_masactrl; @tumanyan2022plugandplay] or cross-attention [@hertz2022prompt; @parmar2023pix2pix_zero; @couairon2022diffedit]. + +Inspired by them, we adopt the second denoising process $\{z'_\tau\}_{\tau=1 \cdots T}$ where the visual-only embedding is provided as the contextual embedding ($\textbf{c}_{\textbf{v},\varnothing} = [v_1 \cdots v_M; \varnothing]$). The original denoising process is denoted as $\{z_\tau\}_{\tau=1 \cdots T}$ that utilizes orthogonal visual embedding in [\[eq:orthogonal\]](#eq:orthogonal){reference-type="ref+label" reference="eq:orthogonal"} along with the textual embedding ($\textbf{c}^\bot = [v^\bot_1 \cdots v^\bot_M; t_1 \cdots t_N]$). Then, we modify the self-attention layers of $\{z_\tau\}_{\tau=1 \cdots T}$, to swap key and value with those from $\{z'_\tau\}_{\tau=1 \cdots T}$. Our proposed self-attention swap can be formulated as follows: $$\begin{equation} + \label{eq:swap} +\text{AttnSwap}(z, z') = M'_S V', \hspace{0.4em} \text{where} \hspace{0.4em} M'_S = \text{softmax}(QK'^T/\sqrt{d}). +\end{equation}$$ Here, $Q \in \mathbb{R}^{l \times d}$ is the query from $\{z_\tau\}_{\tau=1 \cdots T}$ and $K', V' \in \mathbb{R}^{l \times d}$ are the key and the value from $\{z'_\tau\}_{\tau=1 \cdots T}$ where $l$ and $d$ are spatial sequence length and features dimension, respectively. $M'_S \in \mathbb{R}^{l \times l}$ evaluates how closely connected latent spatial features from $z_t$ and $z'_t$ are, assigning higher scores to latent pixels when they have a strong relationship. Using [\[eq:swap\]](#eq:swap){reference-type="ref+label" reference="eq:swap"}, we inject $V'$ in each latent pixel from $\{z'_\tau\}_{\tau=1 \cdots T}$ based on $M'_S$. In other words, while generating a novel pose of the subject, we incorporate *values* from the clear identity of the subject into the location where the corresponding latent pixels exist. + +As we aim to preserve the subject's identity while allowing flexibility in the remaining aspects, we restrict self-attention swap within latent pixels assigned to the subject. Inspired by [@hertz2022prompt; @cao_2023_masactrl], we utilize the cross-attention maps $M_C$ from the original denoising process to create a mask distinguishing the subject from the background. $M_C \in \mathbb{R}^{l \times l_c}$ indicates how much information of text tokens is associated with each latent pixel where $l_c$ is the context sequence length. We could obtain a binary mask $m \in \{0,1\}^l$ for the subject token with fixed thresholds. We ultimately adopt self-attention swap as follows: $$\begin{equation} + \label{eq:swap2} +\text{MaskedAttnSwap}(z, z') = \text{AttnSwap}(z, z') * m + M_S V * (1 - m), +\end{equation}$$ where $M_S = \text{softmax}(QK^T / \sqrt{d})$ and $K, V$ are key and value features from the original process $\{z_\tau\}_{\tau=1 \cdots T}$. [3](#fig:method){reference-type="ref+label" reference="fig:method"}(b) shows the overall process of self-attention swap. We also provide more details in the supplementary materials. diff --git a/2404.16551/main_diagram/main_diagram.drawio b/2404.16551/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..9b9ccff7e7cb49f451dfe7b3230c12374a8528ff --- /dev/null +++ b/2404.16551/main_diagram/main_diagram.drawio @@ -0,0 +1,116 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/2404.16551/main_diagram/main_diagram.pdf b/2404.16551/main_diagram/main_diagram.pdf new file mode 100644 index 0000000000000000000000000000000000000000..7379aea84c7c42fd54620805d0950fb20861fdb3 Binary files /dev/null and b/2404.16551/main_diagram/main_diagram.pdf differ diff --git a/2404.16551/paper_text/intro_method.md b/2404.16551/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..b3918dcf37ddcb1c1aa5d425fa677643f43343d8 --- /dev/null +++ b/2404.16551/paper_text/intro_method.md @@ -0,0 +1,41 @@ +# Introduction + +With the rising popularity of deep learning with applications across different domains, finding a well-performing neural network architecture for a given application has become a key problem. The field of Neural Architecture Search (NAS) automatizes the process and has gained noticeable attention in the past years [@ElskenMH19; @insight1000]. Due to the high costs of neural network training, speedup techniques are a key component of the NAS process. One of these techniques is based on performance predictors -- models that help us to reduce the number of trained architectures by predicting their performance based on a small number of sampled and trained networks [@naslib-predictors]. Since these prediction models still require some networks to be trained and in some cases even include an overhead when fitting on the sampled train set, other, even more efficient techniques were needed. As an answer, *zero-cost proxies* have been recently proposed as a variant that does not require any network training [@lightweight; @TanakaKYG20; @MellorTSC21]. For every network, these proxies need mostly a single mini-batch of data to compute a score that correlates with the true performance of the network after full training on the downstream task of interest. + +However, the reason for the correlation is still underexplored. Furthermore, @nbsuitezero even discovered biases between these proxies and network properties like the number of skip-connections. On the large number of various tasks that a NAS method should be able to solve nowadays [@tnb101], zero-cost proxies show inconsistent performances and most of these proxies even have a lower average correlation with the network performance than simple metrics, such as `flops` [@colin2022adeeperlook; @nbsuitezero]. In addition, most performance prediction methods lack interpretability. This is in particular an important property that would eventually allow us to determine what architectural properties drive the performance for a given task. + +To address these limitations of zero-cost proxies, we first examine in an in-depth analysis the reason behind the zero-cost proxy biases, showing that it is in some cases a direct dependency rather than an actual correlation. Inspired by that, we take advantage of these biases and propose in this paper *neural graph features* -- simple to compute network properties like operation counts or node degrees. We show that using the proposed neural graph features as an input to several (simple) prediction models provides strong and interpretable performance prediction on several tasks, ranging from the common accuracy prediction to hardware metrics prediction and lastly to (multi-objective) robustness prediction. + +Our main contributions are summarized as follows: + +1. We examine biases in zero-cost proxies (ZCP) and show that some of them directly depend on the number of convolutions. We also demonstrate that they are poor at distinguishing structurally similar networks. + +2. We propose *neural graph features* (GRAF), interpretable features that outperform ZCP in accuracy prediction tasks using a random forest, and yield an even better performance when combined with ZCP. + +3. Using GRAF's interpretability, we demonstrate that different tasks favor diverse network properties. + +4. We evaluate GRAF on tasks beyond accuracy prediction, and compare with different encodings and predictors. The combination of using ZCP and GRAF as prediction input outperforms most existing methods at a fraction of the compute. + +# Method + +
+
+
+ +
nwot against cifar-10 val. acc.
+
+
+ +
l2_norm against cifar-10 val. acc.
+
+
+ +
nwot against class_scene val. acc.
+
+
+
ZCP score of all networks against the validation accuracy - colors indicate the number of conv3x3 and 1x1
+
+ +In our work, we will use precomputed zero-cost proxy scores from NAS-Bench-Suite-Zero [@nbsuitezero]. Refer to Section [10.3](#app-zcp){reference-type="ref" reference="app-zcp"} in the appendix for more details about the provided zero-cost proxies, and Section [10.1](#app-benches){reference-type="ref" reference="app-benches"} for the used benchmarks, datasets, and abbreviations. For NB101 and NB301, zero-cost proxies were computed only for a fraction of the search space, and all subsequent experiments are evaluated on these samples. For NB201 and TNB101-micro, we also exclude networks with unreachable branches due to zero operations. Detailed information about the process is in Section [10.2](#appendix-unreachable){reference-type="ref" reference="appendix-unreachable"} in the appendix. + +To motivate our approach, we first examine the limitation of many zero-cost proxies, namely their correlation with the layer count. Based on these insights, we then propose our simple-to-compute graph properties which we name neural graph features (GRAF). diff --git a/2404.18585/main_diagram/main_diagram.drawio b/2404.18585/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..e9385032f37a63605ada8a6ddfdb9d820aa4796a --- /dev/null +++ b/2404.18585/main_diagram/main_diagram.drawio @@ -0,0 +1,409 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/2404.18585/main_diagram/main_diagram.pdf b/2404.18585/main_diagram/main_diagram.pdf new file mode 100644 index 0000000000000000000000000000000000000000..08f16d621bb81492741113ff55aa3babf79f3c52 Binary files /dev/null and b/2404.18585/main_diagram/main_diagram.pdf differ diff --git a/2404.18585/paper_text/intro_method.md b/2404.18585/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..204cae09b71b202358cdbbe3a386d90cd32f7aac --- /dev/null +++ b/2404.18585/paper_text/intro_method.md @@ -0,0 +1,122 @@ +# Introduction + +Table Question Answering (TQA) deals with answering natural language questions related to information organized in a table. TQA systems serve as a fundamental component for interacting with relational databases (Zhong et al., 2017; Yu et al., 2018) through natural language and for processing information across diverse domains, e.g., science (Desai et al., 2021) and finance (Zhu et al., 2021). + +Processing tabular knowledge poses notable challenges. While tables are structured, there are no standard table layouts for representing a particular type of data. Table cells may contain various data types, including text and numbers. Moreover, tables can have nested structures. Thus, TQA systems should be able to robustly integrate textual + +Q: Who received the least amount of votes? + +Olsson + +4 + +361 + +Olsson + +Figure 1: Our benchmark addresses three aspects of robustness shown in the yellow boxes. Answers are bold in tables. Original tables (left) are what exist in a TQA dataset and changed tables (right) show tables after perturbations. We demonstrate three perturbations in this figure (top to bottom): table transposing, removing relevant cells and modifying values to change answers. + +commonsense understanding with numerical reasoning applied to structured data. + +Recent benchmarks for TQA systems (Zhao et al., 2023; Yang et al., 2022) reveal that current TQA systems generate inconsistent responses by performing coarse-grained changes in tables and questions. However, these benchmarks contain a mix of questions requiring different levels of retrieval and aggregation capabilities. + +As a result, these benchmarks fail to differentiate between various aspects of robustness, such as errors stemming from incorrect cell retrieval versus those arising from failures in value aggregation steps. Pinpointing the exact failures of TQA systems, however, is a necessary diagnostic step towards enhancing their robustness. + +In this paper, we lay the groundwork for a more fine-grained systematic evaluation of TQA systems. + +&lt;sup>1https://github.com/boschresearch/FREB-TQA + +To foster the development of robust TQA systems, we address three[2](#page-1-0) important desiderata, as illustrated in Figure [1.](#page-0-1) With these aspects, we intend to separate several steps required when answering questions based on tabular data, i.e., identifying the relevant table cells and aggregating over or comparing them. In particular, we propose to evaluate TQA systems according to three aspects: (1) Retrieval robustness in case of table structure changes: Systems should retrieve the correct relevant cells regardless of table structure changes; (2) Attention to relevant cells: The system should use the relevant cell values from retrieval instead of exploiting shortcuts, such as model-internal knowledge or positional biases, when composing answers; and (3) Aggregation/comparison robustness in case of value changes: The system should aggregate the values of relevant cells correctly regardless of cell value changes. Aggregation can involve different types of reasoning. In this study, we focus on numerical operations, such as counting and comparisons, as they are the most common reasoning types in TQA [\(Zhu et al.,](#page-11-0) [2021\)](#page-11-0). + +To evaluate TQA systems with regard to these aspects, we create FREB-TQA, a new benchmark based on four well-studied TQA datasets using seven novel automatic perturbations, one perturbation from previous work [\(Zhao et al.,](#page-10-2) [2023\)](#page-10-2), and extensive manual annotations. + +In our extensive experiments, we study pipeline and end-to-end state-of-the-art TQA systems [\(Liu](#page-10-4) [et al.,](#page-10-4) [2021;](#page-10-4) [Jiang et al.,](#page-9-1) [2022a;](#page-9-1) [Herzig et al.,](#page-9-2) [2020;](#page-9-2) [Cheng et al.,](#page-9-3) [2022\)](#page-9-3), as well as large language models (LLMs) [\(Touvron et al.,](#page-10-5) [2023\)](#page-10-5) using our benchmark. The fine-grained results uncover shortcomings of these systems that have not been previously explored. For example, we show that model performance diminishes substantially when column order changes, or when cells containing the answer are positioned at the bottom of the table. This finding indicates a strong positional bias in TQA systems, motivating more informative table encodings. + +Our main contributions are: (1) We propose a novel benchmark of 8,590 selected and partially manually annotated questions and tables, resulting in a total of 75,205 instances using seven perturbation methods; (2) using several inter-annotator agreement studies, we demonstrate the solidness of our benchmark; and (3) using our benchmark, we + +experiment extensively with state-of-the-art TQA systems, discovering that for almost all robustness tests, system performances drop, demonstrating the difficulty of our benchmark for TQA systems. + +# Method + +FREB-TQA is a Fine-grained Robustness Evaluation Benchmark for Table Question Answering. From four TQA datasets (Table [1\)](#page-2-0), we first classify questions with regard to whether questions merely require cell value retrieval or further reasoning. We then generate perturbations for evaluating each aspect of robustness by making use of seven perturbation methods and collecting human annotations. + +For building FREB-TQA, we leverage the development sets of four well-studied TQA datasets: WikiTableQuestions (WTQ, [Pasupat and Liang,](#page-10-14) [2015\)](#page-10-14), WikiSQL [\(Zhong et al.,](#page-10-0) [2017\)](#page-10-0), Sequential Question Answering (SQA, [Iyyer et al.,](#page-9-8) [2017\)](#page-9-8), and Tabular And Textual dataset for Question Answering (TAT, [Zhu et al.,](#page-11-0) [2021\)](#page-11-0). The first three datasets feature tables from Wikipedia, TAT addresses the financial domain. See Appendix [A.1](#page-12-0) for detailed statistics. For all source datasets, we eliminate questions that relate to the table structure, such as "What is the name of the actor in the first row?" To identify such questions, we use a word list provided in Appendix [A.3.](#page-12-1) Around 10% of the questions in WTQ and 3% of TAT questions are filtered out by this criterion. + +To decouple the robustness aspects in TQA, we group questions from the source datasets into extraction questions (EQs) and reasoning questions (RQs) by applying heuristics and classification models. The answer to an EQ can be retrieved from a single cell of the table. The answer to an RQ additionally requires aggregating over several cell values. In Figure [1,](#page-0-1) the question "How many floors does the Fosh Tower have?" is an EQ since it only requires retrieving a cell value. The question + + + +| Model | EQ | RQ | +|------------|-------|-------| +| LLaMA2 | 74.27 | 57.02 | +| Rule-based | 75.23 | 83.42 | +| Combined | 93.81 | 60.17 | + +Table 2: The precision of examined models for question type classification on 200 questions from WTQ. + +"Who received the least amount of votes?" is an RQ as it requires comparing values of several cells. + +Question type classification for WTQ. WTQ provides a diverse set of questions. We group them into EQs and RQs using two different methods. Each method is tuned to identify either EQs or RQs with high precision, as our final benchmark will only contain the set of questions identified as EQ or RQ by the respective method.[3](#page-2-1) + +Given a table, a question, and an answer, the lexical rule-based method conducts a string match between the answer and the table's cell values. If there is no match, the corresponding question is labeled as RQ. Otherwise, we detect if the question contains any comparative or superlative words using POS tags.[4](#page-2-2) Since these words are signals for aggregation over table cells, we mark the question as RQ. Otherwise, the question's type is set to EQ. We also prompt LLaMA2-13b [\(Touvron](#page-10-5) [et al.,](#page-10-5) [2023\)](#page-10-5) to label the question type (The prompt is provided in Appendix [A.4\)](#page-12-2). Finally, we combine the lexical rule-based and LLaMA2 models. If the lexical rule-based model labels a question as EQ, we obtain the LLaMA's prediction in addition. If they agree, the question type is set to EQ. Otherwise, it is set to RQ. To estimate the quality of these methods, we manually annotate questions sampled randomly from WTQ (100 EQs and 100 RQs). Details are given in Appendix [A.2.](#page-12-3) + +Table [2](#page-2-3) shows the results achieved by these models. As we aim for a high-precision question-type classification, we apply the combined model for identifying EQs and the lexical rule-based method for identifying RQs on all questions in WTQ. + +Question type classification for WikiSQL. Questions in WikiSQL do not require complex reasoning [\(Zhao et al.,](#page-10-2) [2023;](#page-10-2) [Lin et al.,](#page-10-12) [2023\)](#page-10-12). In WikiSQL, questions are labeled with regard to the operations required to answer them. We select the questions that can be answered without performing + +3Around 20% of questions are selected by neither of the models and thus eliminated from our benchmark. We also found that the final set of EQ and RQ cases do not overlap. + +4 + +aggregation operations and label them as EQs. + +Question type classification for SQA. The SQA dataset consists of dialogues in the form of questions and answers related to the information in a table. The answers to all questions except for the first question in a dialog rely on the dialog history. Thus, we select only the first question from each sequence, which usually asks for retrieving information from the table [\(Iyyer et al.,](#page-9-8) [2017\)](#page-9-8). Hence, we mark them as EQs. + +Question type classification for TAT. TAT consists of questions about tables and text snippets and requires numerical reasoning. The dataset also features annotations about how an answer is derived, whether the answer is based on the table or the text, and which operations are required to answer a question. We leverage these annotations to identify EQs and RQs. More specifically, we first select questions that can be answered only based on tables. Next, if a question needs derivation or comparison, we classify them as RQ. Otherwise, we classify them as EQ. + +A robust TQA system should answer an EQ by retrieving the answer from a table, regardless of table structure changes. We perturb the structure of tables associated with extraction questions. We replicate one perturbation type from previous work and introduce two new perturbation types for measuring this robustness aspect. + +Shuffle all rows (columns). Following [Zhao et al.](#page-10-2) [\(2023\)](#page-10-2), we randomly shuffle all rows (columns) in a table. This perturbation allows us to study if TQA systems are robust against changes in re-arranging all rows (columns). However, it does not reveal which biases may impact the robustness of TQA systems. Thus, we introduce the following two new perturbation methods for this aspect. + +Shift target rows (columns). For each EQ, the *target* row (column) in a table contains the cell that corresponds to the answer. This perturbation type shifts target rows (columns) either to the top, to the middle, or to the bottom part of a table. We identify the target cell by applying exact match between the answer and table cell value. For shifting target rows, we partition a table into three equal-length parts, referred to as top, middle, and bottom. For shifting target columns, we partition a table into + +two equal-length parts: front and back. We use more partitions for rows because on average tables include more rows than columns (Table [3\)](#page-4-0). For our benchmark, we remove the target rows (columns) from the table and re-insert them at a random position in each partition. This perturbation method allows us to study whether TQA systems exhibit any positional biases. + +Transpose. This perturbation type transposes the table, i.e., it rotates the table by 90 degrees and turns rows into columns and columns into rows. This perturbation allows us to study if TQA systems have a bias towards particular table layouts. + +TQA systems may answer a question by exploiting shortcuts (model-internal knowledge or positional biases) without paying attention to relevant cells (cells that are important to compose an answer). We propose three new perturbation methods for reasoning questions (RQs) to study this aspect of TQA robustness. + +Remove relevant cells. We perturb a table by removing relevant cells from tables. The relevant cell annotations associated with RQs in our benchmark have been created by [Zhu et al.](#page-11-0) [\(2021\)](#page-11-0) and [Ye et al.](#page-10-9) [\(2023\)](#page-10-9) for TAT and WTQ, respectively. We test the validity of the latter since the annotations are gathered from LLMs (see Appendix [A.2\)](#page-12-3). This perturbation lets us investigate to what extent TQA systems bypass relevant cells to derive their answers. We observe that 70% of relevant cells contain non-numerical values for WTQ. For TAT, all relevant cells contain numerical values. + +Remove table. TQA systems may bypass the whole table and use their internal knowledge to answer a question. To test for this behavior, for any RQs, we replace the table with a dummy table, consisting of one cell with a "None" value. + +Shift relevant rows. This perturbation evaluates to what extent TQA systems bypass table cell values and rely on the position of relevant cells. For instance, to answer the question "Who received the least amount of votes?" in Figure [1,](#page-0-1) a TQA system may exploit a shortcut between the last row and the question since in most cases, rows in tables from TQA datasets are sorted. Because of the correlation between cell values and positions, this type of + +![](_page_4_Figure_0.jpeg) + +Q: how many album entries are there? + +Figure 2: An example of aggregation/comparison robustness in case of string change. ① illustrates shortening an original table to table cells on which numerical aggregations or comparisons operate. ② illustrates modifications on a shortened table, leading either to a change in answer or not. The orange parts mark changed strings. + +shortcut mostly occurs for RQs that require comparing cell values. In this perturbation type, for each RQ, we remove relevant rows and re-insert them at a random position of the table. + +TQA systems should compose a correct answer to an RQ by aggregating over or comparing values in the provided table independent of value changes. To evaluate this capability, for RQs, we manually modify cell values that should be aggregated to answer questions. We ask three human annotators to first identify the table cells that numerical aggregations or comparisons operate on (step 1) in Figure 2). We refer to the part of the table relevant to answering the question as the *shortened table*. We use the shortened table versions in our experiments to minimize the effect of table length in this test. Then, annotators are asked to change one or two cell values per question, once resulting in an answer change and once not resulting in an answer change (step 2) in Figure 2). We define two types of perturbations. + +Modify values to change answers. This perturbation allows us to study to what extent TQA systems perform correct aggregations and adapt answers to the value changes accordingly. An example of numeric value changes is shown in Figure 1. The votes for *Olsson* is changed from 4 to 361, resulting in a change of the answer (from *Olsson* to *Leslie*). Figure 2 shows an example of string value changes. The first entry *Go West Young Man* is removed, changing the answer from 2 to 1. + + + +| Aspect | #P | #Q | #QT | #R | #C | #A | +|---------|-------|------|-------|-------|------|------| +| RR-TSC | 64890 | 6489 | 10.05 | 12.14 | 7.18 | 1.02 | +| Rel-Cel | 6303 | 2101 | 14.75 | 21.44 | 5.77 | 1.05 | +| ACR-VC | 4012 | 2006 | 14.62 | 21.70 | 5.61 | 1.05 | +| Total | 75205 | 8590 | 10.28 | 14.42 | 6.85 | 1.03 | + +Table 3: Benchmark statistics grouped by robustness aspects: retrieval robustness in case of table structure changes (RR-TSC), attention to relevant cells (Rel-Cel), and aggregation/comparison robustness in case of value changes (ACR-VC). #P and #Q show the number of perturbations and questions, respectively. # QT, #R, #C, and #A show the average question length in tokens, the number of table rows, the number of table columns and the number of cells for composing an answer. + +Modify values without changing answers. This perturbation aims to study if systems articulate correct answers because of their biases to certain values. For instance, in Figure 1, systems might be capable of comparing 15 and 4. However, if we change the votes for *Leslie* from 15 to 1500, without changing the answer, systems might face difficulties as 1500 might not fall into the cell values distribution during pre-training. Figure 2 shows an example of string value changes: *West* is changed to *East* without changing the answer (2). + +We filter out instances that annotators find unanswerable. In the case of TAT, all modified values are numeric. In the case of WTQ, 50% of the modified cell values are numeric, the others are string-based. + +To assess the quality of the output of these perturbations, for each perturbation type, we randomly sample 50 instances created using this type from each dataset, resulting in 200 instances in total. Then, we ask two annotators who were not involved in the perturbation creation to provide the answers given the changed tables and questions. We compute the exact match accuracy of answers provided by the two annotators, which amount to 92.5% and 93.5%. We find that most wrong cases are related to questions asking for percentage changes: here, annotators sometimes neglected to add the minus symbol to negative percentage changes. diff --git a/2405.00313/main_diagram/main_diagram.drawio b/2405.00313/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..40a771648faa7945a8f6e32c7e920fd9d099cf89 --- /dev/null +++ b/2405.00313/main_diagram/main_diagram.drawio @@ -0,0 +1,245 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/2405.00313/paper_text/intro_method.md b/2405.00313/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..97efeaffa38a54b8df488b9d6759439e00a64883 --- /dev/null +++ b/2405.00313/paper_text/intro_method.md @@ -0,0 +1,289 @@ +# Introduction + +Image editing is a continuously evolving area of research in image processing, which has seen significant advancements in recent years with the advent of advanced text-to-image (T2I) techniques. T2I methods are a group of machine learning models that are able to generate images from textual descriptions and include Generative Adversarial Networks (GANs), Variational Autoencoders (VAEs), and Denoising Diffusion Models (DMs) [@zhang2023text]. Among these, DMs [@ho2020denoising] have garnered considerable attention due to various characteristics, including a simple training scheme, great performance in generating high-quality images with a wide range of styles, not being prone to mode collapse, and being easy to employ in different tasks. + +Despite showing exceptional results in different applications, e.g. image synthesis, super-resolution [@saharia2022image], inpainting, etc., DMs models are prone to several disadvantages, such as being sensitive to the choice of the original noise distribution, and being difficult to control [@wang2020state]. These limitations can lead to a high degree of stochastic behaviour, often requiring many "generations" before a desirable result is achieved. Additionally, due to the nature of the LDM generation technique, the results are often hard to fine-tune. Therefore, a region-targeted image editing tool has the potential to significantly expand the capabilities of these models. + +In this work, we introduce Layered Diffusion Brushes, a novel Latent Diffusion Model-based (LDM) [@rombach2022high] editing and fine-tuning tool that users, including artists, can utilize for efficiently making real-time region-specific adjustments and modifications to images. The core of our technique, Layered Diffusion Brushes, uses a user-specified mask to introduce targeted random noise patterns into the latent image at specific steps during the reverse diffusion process, which are then denoised using both mask and prompt guidance into new image content over subsequent steps. This produces targeted changes to the specified regions while preserving the original context from the rest of the image. Our technique results in changes that are well-integrated with the rest of the generated image, without affecting unmasked areas. By employing a latent caching technique, our system is capable of rendering edits in as little as 140 ms on consumer hardware, enabling rapid feedback and refinement. In particular, we provide for a "scroll" gesture that allows users to rapidly adjust the random noise seed and thus explore variations of the desired edit. + +We initially developed an early prototype of this work, with the goal of re-randomizing targeted regions for fine-tuning, and conducted a user study to assess its usability and features. Based on the feedback received from this first study, we made significant improvements and revamped the tool. These enhancements included caching in the backend to improve inference speed, a completely redesigned user interface (UI) featuring improved masking controls, and more streamlined adjustments. + +On top of the core Layered Diffusion Brushes technique, we implement a layering system which allows multiple edits to be stacked on top of each other. Edit layers can be switched, hidden, or deleted without affecting other layers, despite all operating on the same latent image. Multiple edit layers with their corresponding masks and prompts are shown in Figure 1. + +To evaluate the performance, usability, and effectiveness of Layered Diffusion Brushes, we conducted a user study with artists who are experienced in editing images and also have basic familiarity with AI image editing and generative tools. Our user study aimed to assess the efficacy of Layered Diffusion Brushes in improving the experience of users in fine-tuning real AI-generated images and providing artistic control, and the creativity support and functionality of our tool in comparison to two other well-known existing image editing tools, InstructPix2Pix [@brooks2022instructpix2pix] and SD-Inpainting [@rombach2022high]. []{#intro label="intro"} + +# Method + +As mentioned in section [1](#sec:intro){reference-type="ref" reference="sec:intro"}, we use an LDM-based variant of image generative models. Similar to [@lugmayr2022repaint; @avrahami2023blended], we make intermediate adjustments to the latent space using the pre-trained LDM. Therefore, Layered Diffusion Brushes does not require any additional training and only modifies the latents during the reverse diffusion process. + +[\[alg:store\]](#alg:store){reference-type="ref" reference="alg:store"} presents the algorithm for generating intermediate latents. This part simply caches two latents, $Z_{r}$ and $Z_{t}$. Here, $Z_{r}$ serves as the initial latent for editing. In this formulation, $r = N-n$, where $N$ is the total number of steps to create the initial image, and $n$ is the total number of steps for editing. $Z_{L1}$ represents the latent used for blending the edited latent in layer 1. + +We first initialize the sample $Z_0 = \epsilon_0$ and noise level $\sigma_0$ ($i = 0$) if the image is generated using the DM. If a real image is used, we obtain the initial noise latent for the image using the Direct Inversion method presented in [@ju2023direct]. Based on our testing, this method is significantly faster in terms of speed and comparable in terms of performance to other methods in the literature, including null text inversion [@mokady2023null] and Negative-Prompt Inversion [@miyake2023negative] methods. + +Algorithm [\[alg:regen\]](#alg:regen){reference-type="ref" reference="alg:regen"} contains the main backbone of the system, which handles both editing and layering functionalities. + +To initiate an edit, the algorithm begins by generating a new noise pattern $Z^{\prime}_0 = \epsilon^{\prime}k$ sampled from $\mathcal{N}(0, S^{\prime2}{noise}I)$. Here, we generate a new noise pattern from a different seed (set per-layer). Then, $Z^{\prime}0$ is scaled to match the variance of $Z{r}$ obtained from the previous step. This ensures that the additive noise is not too different from the latent for editing, which would result in visual artifacts. The new noise is added to the original latent, controlled by the mask and $\alpha_k$. The effect of using different values for $n$ and $\alpha$ will be discussed in detail in [2](#ui){reference-type="ref" reference="ui"}. + +During the edit loop, at step $t$, in each layer, the latent of the new seed $S'$ and the previous layer's latent are merged using the mask, thus confining the influence of the new latent to the masked region. Subsequently, the new latent region is progressively denoised and integrated into the existing image contents from steps $t$ through $N$. To keep the UI simple and reduce the number of user-controlled parameters, we fix the value of $t$ in our implementation. Preliminary tests suggested that the value of $t = N - 2$ works well for many applications. + +Finally, there is separate logic that tracks all the layers and their corresponding parameters, which is triggered whenever a new layer is created or if the user switches between the layers. Consequently, all the layers are created and stacked one by one, and the appropriate $Z_{L_{k}}$ is provided to the Algorithm [\[alg:regen\]](#alg:regen){reference-type="ref" reference="alg:regen"} to perform the edit. + +When creating a new layer, the user specifies a mask $m$. For each layer and the corresponding mask $m_k$, the user specifies an integer $n_k$ (per [1](#fig:overview){reference-type="ref" reference="fig:overview"}), which is the total number of steps that the noise is introduced to the latent, and a tunable parameter $\alpha_k$ to control the additive noise strength. + +We show the overview of the core algorithm of the proposed method in [1](#fig:overview){reference-type="ref" reference="fig:overview"} and also present the updated pseudocode for the denoising process: + +
+ +
Overview of the proposed method: The top row shows the typical DM composition of an image from a noisy latent Z0 and a prompt P. The bottom shows the overview of the proposed method: the algorithm takes a new seed S′, and combines it with the original latent at step n using a mask m, and a strength control alpha. The diffusion process moves forward and at a certain step t (we fix t = N − 2) the original latent and the new modified latent are merged together using the previous layer’s intermediate latent and masking and finally the edited image is generated.
+
+ +::: algorithm +**Input:** A source prompt **$\mathcal{P}$**, a random seed **$S$** and a number of steps $N$ for generating original\ +[]{#alg:store label="alg:store"} **Output:** Generated image $Z_{N}$, latent for regeneration $Z_{N-n}$, and latent for blending $Z_{t}$.\ +$Z_0 \sim \mathcal{N} (0,S^2 I) || Z_0 \sim Inv(I)$ **Return** $Z_N, Z_{r}, Z_{L_0}$ +::: + +::: algorithm +**Input:** An edit prompt **$\mathcal{P'}$**, a fine-tuning mask **$m$**, a random seed **$S'$**, a brush hardness control variable **$\alpha$**, a step number **$n$** to introduce additive noise, a latent for regeneration $Z_{r}$, and a latent for blending with the input image $Z_{L_{K-1}}$\ +**Output:** Edited image $Z_{N}$.\ +$Z^{\prime}_0 \gets \epsilon^{\prime}_{n_k} \sim \mathcal{N}(0, S^{\prime2}I)$\ +$Z^{\prime}_0 \gets \sqrt{Var( Z_{r})} \cdot Z^{\prime}_0$\ +$Z^{\prime}_0 \gets Z_{r} + \alpha \cdot Z^{\prime}_0 \cdot m$ **Return** $Z^{\prime}_n$ +::: + +Our goal with Layered Diffusion Brushes was to design a practical tool for artists and graphic designers to use. In particular, we aimed to design a set of controls that would be intuitive and flexible, without overwhelming users with excessive choices. Figure [2](#ui){reference-type="ref" reference="ui"} provides an overview of the user interface. As illustrated in the figure, the UI comprises the following primary sections (each section highlighted with the corresponding number on the image): + +![ Design of the Layered Diffusion Brushes UI: The name and functionality of each section are described in the text. In this example, the user has created three layers, visualized on the image canvas, along with the mask and edit prompt. The selected layer in this picture is Layer 1.](ui2.PNG){#ui width="99%"} + +1. Model Section + + - This section in the UI enables users to load various model combinations, including pre-trained models, schedulers, and LoRA [@hu2021lora]. + +2. Generation section + + - This section allows users to either generate a new image using a seed and prompt combination, or upload a real image that will be inverted. + +3. Image Canvas + + - This canvas serves as the workspace where users interact with and make edits to images. + +4. Editing Section + + - This section empowers users to create and access different layers and adjust various controls to make desired edits. + +Two types of interactions are provided for performing edits, as shown in Figure [3](#fig:brush){reference-type="ref" reference="fig:brush"}: + +**Box Mode:** Users can click or drag on the image to move a resizable square mask around the image. This option enables users to quickly and interactively explore how various parts of the image will change in response to a given set of editing settings (prompt and strength), simply by moving the cursor. + +**Custom Mask Mode:** Users can draw the mask in any shape they like using a resizable paintbrush, clicking and dragging to add to the mask. Users can navigate between new generation samples by scrolling the mouse up or down while hovering over the image, allowing them to rapidly explore variations on their edit. + +We envision that users could use the box option first to choose a good location to place their edit and refine the prompt/strength settings, followed by the custom mask to refine the shape and appearance of the edit. + +
+
+ +
Box option with moving cursor
+
+
+ +
Custom mask option with mouse scroll
+
+
Box and Custom Mask options: When utilizing the box option, the user clicks on the center of the targeted region to obtain an edit within the specified area. Additionally, the user can drag the box to a new position to explore different outcomes. Alternatively, when the custom mask option is selected, the user initially draws the mask on the desired region. Subsequently, the user can use mouse scroll to generate new variations.
+
+ +Proper utilization of the layering capability enables the tool to achieve more complex effects: users can stack, hide and unhide layers and adjust the strength of each layer. Furthermore, each layer can be edited independently without substantially affecting the appearance of edits in higher or lower layers. + +Here are the main hyperparameters used in the Layered Diffusion Brushes editing method: + +- **Number of regeneration steps ($n$)**: An integer value that specifies the number of steps that the Layered Diffusion Brushes will run to make the edit + +- **Brush Strength ($\alpha$)**: A float number that corresponds to the $\alpha$ value in equation ([\[alpha\]](#alpha){reference-type="ref" reference="alpha"}) which controls how strong the noise should be applied at the targeted region. Initial $\alpha^*$, specified by the user has a value between 0 and 100, which will be scaled later on in the algorithm using the following equation: $$\begin{equation} + \alpha = \frac{\sqrt{\left| \frac{\alpha}{100} \cdot \left( \sigma - 2 \cdot \frac{\text{Cov}(Z_{n_k}, x^{\prime}_0)}{\text{Var}(Z_{n_k})} \right) \right|}}{\sqrt{\frac{\sum_{i=1}^{m} \sum_{j=1}^{n} [m_{ij} \neq 0]}{W}}} + \label{alpha} + \end{equation}$$ + + where $x^{\prime}_0$ and $Z_{n_k}$ are the new noise latent and latent for regeneration respectively (as notes on \...), $\sigma$ is the acceptable range for the variance of the $Z_{n_k}$(we used $\sigma$=0.25), $m$ is the corresponding mask, and $W$ is the width of the $Z_{n_k}$ (W=512) + +- **Seed Number ($s'$)**: An integer number that will be used for generating the Gaussian noise pattern in the specified region. As with normal image generation, the UI provides buttons to randomize the seed or reuse the previous seed. Moving the box around (in box mode) or using the scroll wheel (in custom mask mode) will adjust the seed automatically. + +- **Brush Size $r$**: An integer value that dictates the radius of the box when utilized in box mode, or the size of the brush in custom mask mode. The formulation of $\alpha$ in Equation [\[alpha\]](#alpha){reference-type="ref" reference="alpha"} aims to mitigate the impact of varying brush sizes on the quality of edits by considering the number masked pixels and adjusting accordingly. + +The number of edit steps ($n$) and the brush hardness ($\alpha$) parameters jointly control the magnitude of the intermediate additive noise and ultimately the amount of applied change to the latent image. Higher values for the brush hardness will result in a bigger change. Figure [4](#stren){reference-type="ref" reference="stren"} (top row) shows the effect of a wide range of different hardness values while keeping other parameters the same. As shown, if the mask strength is too high, the LDM cannot fully denoise, and the additive noise causes artifacts. Conversely, if the $\alpha$ value is too small, Layered Diffusion Brushes will not be able to make sufficient adjustments. We use a generalized version of strength in order to make the parameter more intuitive, as discussed above. + +Similarly, we tested the effect of using different $n$ values while keeping the rest of the parameters stationary. As demonstrated, if the editing is run for fewer steps (lower $n$) , the LDM cannot recover from it and will produce artifacts, while if $n$ has a higher value, the area inside the mask will be denoised too much and cannot properly blend well with the rest of the image. + +Additionally, there is a correlation between the two parameters. When the noise is introduced in later stages of diffusion (higher value of $n$), the model has less time to recover from the additive noise. Therefore, it is recommended to adjust the $alpha$ value to a higher number. Conversely, if the noise is introduced in the early stages of diffusion, and if the value of $alpha$ is not large enough, the additive noise will be dissolved into the input image latent and several stages of diffusion will eliminate its effect. Therefore, if $n$ is small, the $alpha$ value should be larger to achieve optimal results. + +
+
+ +
+

  

+
+ +
+
+ +
+
+ +
+
+ +
+
+ +
+
+ +
+

  

+
+ +
+
+ +
+
+ +
+
+ +
+
+ +
+
Effect of changing different controlling parameters of the Layered Diffusion Brushes. On the top row, the mask strength (α) is altered while the mask, the seed, and the intermediate number for denoising (n) have remained stationary. On the bottom row, we changed n and the rest of the parameters remained intact. In all examples S′ = 2, edit prompt: “balloons” original prompt: “Photo of a city with balloons”, and the latent for edit (Z = 6)
+
+ +We conducted a user study in order to evaluate the effectiveness of Layered Diffusion Brushes for providing targeted image fine-tuning, in comparison with InstructPix2Pix and the SD-Inpainting methods. + +We recruited a cohort of seven expert participants with extensive experience in using image editing software. The group comprised four females and three males, with an average age of 30.4 years. As part of our participant selection criteria, we ensured that all participants possessed at least a basic level of familiarity with AI image generation techniques and were regular users of editing software such as Adobe Photoshop for creating visual art. Two participants were proficient in image generative models and Stable Diffusion, while the remaining five were graphic design students who used Adobe Photoshop and Illustrator on a daily basis. + +The study was conducted remotely; participants were provided a link to access the tool. Each user engaged in two types of tasks: free-form tasks where users generated an image for editing using a fixed prompt and seed (type 1), and pre-determined tasks where the user worked with existing real images from the MagicBrush dataset [@zhang2024magicbrush] (type 2). + +The study started with a brief introduction to each of the methods. Following this, participants received a short tutorial on how to navigate the user interface (UI). Subsequently, they were provided with a 5-minute window to explore the various options and sections of the tool, becoming familiar with the use of each section. + +A dedicated task section was incorporated into the user interface (UI) specifically for the user study. Each type of task comprised three rounds of edits using the three methods: Layered Diffusion Brushes, InstructPix2Pix, and SD-Inpainting. + +Each user was assigned a unique user ID, and tasks were randomly selected and pre-assigned to users. Throughout the study, users interacted with the task table to load, select, and save each task. An example of the task section is illustrated in Figure [5](#tasks){reference-type="ref" reference="tasks"}. + +![Overview of the tasks section, where users can interact to load, select, and save each task. Tasks that are selected are highlighted in blue, while those completed and saved are highlighted in green.](tasks.PNG){#tasks width="99%"} + +For the type 1 tasks, we selected specific types of edits that showcase various functionalities and capabilities of the system. Here are the description of edit types along with an example used during the user study: + +1. Stack layers and create sequential edits (draw with Layered Diffusion Brushes: + + - Input image: photo of a beautiful beach. + + - Layer 1: boat (Introduce a boat in the sea) + + - Layer 2: rocks (Scatter weathered rocks along the shoreline) + + - Layer 3: birds (Populate the sky above the boat with a flock of birds. + +2. Modify attributes and features of objects: + + - Input image: portrait of a young man + + - Layer 1: blond (Transform a person's hair color to blond). + + - Layer 2: joker (Perform facial manipulation by swapping one person's face with another's, reshaping identities.) + +3. Correct image imperfections and errors: + + - Input image: portrait of a man holding an umbrella + + - Layer 1: remove the rod that is mistakenly placed + + - Lyaer 2: fix the extra part on the side of the coat + +4. Enhance discernibility of similar objects through modification: + + - Input image: aerial photo of a pool table with balls + + - Layer 1: change the colour of s specific ball (third ball from the left) to red + +5. Target specific regions for style transfer, refining aesthetics: + + - Input image: Mona Lisa by Leonardo da Vinci + + - Layer 1: make the left part of the background similar to van gogh stary night style. + +In our study design, we strategically chose the combination of seeds and prompts to encompass and evaluate these functionalities. Each user was given three seed-prompt items and tasked with creating and editing up to three layers of edits. For the majority of the tasks, $N$, i.e. the total number of steps for image generation was set to $N=13$. All the images were generated using Dreamshaper-7 [@dreamshaper] and DDIM Scheduler. + +For the Layered Diffusion Brushes method, users started by selecting a layer with an existing edit instruction from the task table, then created the corresponding layer in the UI. They had the option of choosing either the box option or the custom mask option. The task was followed by drawing the mask, tweaking the controls or edit prompt if needed, and completing the edit. Once the task was complete, the user saved the edit and moves on to the next task. + +Users followed a similar procedure for the InstructPix2Pix and SD-Inpainting methods, with the exception of creating layers, as these methods do not incorporate layering capabilities. After completing each layer edit task, users saved the edits, and the user interface (UI) stacked subsequent edits onto the edited image. For the InstructPix2Pix method, users were required to write the instruction prompt and then adjust the image and text guidance scales and regeneration steps to finalize the edit. On the other hand, for the SD-Inpainting method, users drew a mask and controlled the edit using the strength control. Completion times for each task were recorded for both methods. + +Type 2 tasks, corresponding to the MagicBrush dataset [@zhang2024magicbrush], were more structured, with the mask, edit prompt, and input images provided by the dataset. MagicBrush utilized crowd workers to collect manual edits using DALL-E 2 [@ramesh2022hierarchical]. This process involved 5,313 editing sessions and 10,388 editing iterations, resulting in a robust benchmark for instructional image editing. Additionally, the dataset provides manually annotated masks and instructions for each edit. We selected a subset of 35 input images, each containing up to three layers of edits. Users selected each image, started with the provided mask, could modify the mask if necessary, adjusted the control parameters and prompt, and saved and completed the task for each method. [6](#fig:resultsofstudy){reference-type="ref" reference="fig:resultsofstudy"} provides examples of generated images for each of the methods during the user study. + +
+
+

“What if there was no plant life along the building” / “no plants”

+
+
+ +
+
+ +
+

  

+
+ +
+
+ +
+
+ +
+
+ +
+
+

“"Replace the wooden figure with a barbie doll.” / “parking meter with barbie doll on top”

+
+
+ +
+
+ +
+

  

+
+ +
+
+ +
+
+ +
+
+ +
+
+

“Have the woman wear a hat.” / “woman wearing a hat”

+
+
+ +
+
+ +
+

  

+
+ +
+
+ +
+
+ +
+
+ +
+
Results of editing MagicBrush dataset images using different methods: InstructPix2Pix, SD-Inpainting, and Layered Diffusion Brushes. These images were generated by users during the user study. While MagicBrush provides the input mask and editing prompts, users had the option of fine-tuning the mask, modifying the prompts, and selecting the best output. The last column corresponds to the provided MagicBrush ground truth images. Edit instructions (for InstructPix2Pix) and prompt (for SD-Inpainting and Layered Diffusion Brushes) are presented on top of each row.
+
diff --git a/2406.02462/main_diagram/main_diagram.drawio b/2406.02462/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..c69debd80a154cc8a966e77a8d94e4b9aa4e394e --- /dev/null +++ b/2406.02462/main_diagram/main_diagram.drawio @@ -0,0 +1,94 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/2406.02462/main_diagram/main_diagram.pdf b/2406.02462/main_diagram/main_diagram.pdf new file mode 100644 index 0000000000000000000000000000000000000000..66afd97630582835275b66454d8422a5a8957760 Binary files /dev/null and b/2406.02462/main_diagram/main_diagram.pdf differ diff --git a/2406.02462/paper_text/intro_method.md b/2406.02462/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..c3e649d3307efe19e02dc6c09dd48ed1acc12224 --- /dev/null +++ b/2406.02462/paper_text/intro_method.md @@ -0,0 +1,77 @@ +# Introduction + +![ Training the proposed Patch Diffusion Inverse Solver (PaDIS) method. Different sized patches are used in each training iteration. The X and Y position arrays have the same size as the patch and consist of the normalized X and Y coordinates of each pixel of the patch.](fig/training.png){#fig: training width="85%"} + +Diffusion models learn the prior of an underlying data distribution and can use the prior to generate new images [@song:19:gmb; @song:21:sbg; @ho:20:ddp]. By starting with a clean training images and gradually adding higher levels of noise, eventually obtaining images that are indistinguishable from pure noise, the score function of the image distribution, denoted $\s(x) = \nabla \log p(\x)$, can be learned by a neural network. The reverse process (sampling or generation) then starts with pure noise and uses the learned score function to iteratively remove noise, ending with a clean image sampled from the underlying distribution $p(\x)$. + +Inverse problems are ubiquitous in image processing, and aim to reconstruct an image from a measurement , where $\y = \mathcal{A}(\x) + \bepsilon$, $\mathcal{A}$ represents a forward operator, and represents random unknown noise. By Bayes rule, $p(\x | \y)$ is proportional to $p(\x) \cdot p(\y | \x)$. Hence, to recover , it is important to have a good estimate of the prior $p(\x)$, particularly when contains far less information than . Diffusion models are known for their ability to learn a strong prior, so there is a growing literature on using them to solve inverse problems  [@MCGchung; @chung:23:dps; @DDRM; @DDNM; @kawar2021snips]. + +However, diffusion models require large quantities of training data and vast computational power to be able to generate high resolution images; Song et al. [@song:21:sbg] and Ho et al. [@ho:20:ddp] used several days to weeks of training on over a million training images in the ImageNet [@imagenet] and LSUN [@lsun] datasets to generate $256 \times 256$ images. This high cost motivates the research on improved training efficiency for diffusion models, such as fine-tuning an existing checkpoint on a different dataset [@moon2022finetuning; @chung:22:s3i] to reduce the required training time and data. However, this strategy restricts the range of usable network architectures and requires the existence of a pretrained network, which limits the range of applications. Besides the demanding training data and computational cost, diffusion models also struggle in very large-scale problems, such as very high resolution images or 3D images. To address these challenges, latent diffusion models  [@rombach2022highresolution; @karras2022elucidating] have been proposed to learn the image distribution in a smaller latent space, but it is difficult to solve general inverse problems in the latent space [@resample]. Patch-based diffusion models have also been proposed to reduce computational cost. For example, Wang et al. [@wang2023patch] trained on patches of images, but for image generation, ultimately still relied on computing the score function of the whole image at once. Ding et al. [@ding2023patched] used patches in the feature space, requiring an additional encoding and decoding step. For 3D volumes, the most common method involves breaking up the volume into 2D slices [@chung:22:s3i], [@lee2023improving]. These methods add regularizers between slices to enforce consistency during sampling, and thus do not provide a self-contained method for computing the score function of the whole volume. These application-specific strategies make it difficult to adapt these methods to general purpose inverse problem solvers using diffusion models  [@chung:23:dps], [@chung:22:sbd], [@DDNM], [@song:22:sip]. + +![Overview of reconstruction process for the proposed Patch Diffusion Inverse Solver (PaDIS) method. Starting at $t=T$, at each iteration we choose a random partition of the zero padded image and use the neural network trained on patches to get the score function of the entire image. Due to the shifting patch locations, the output image has no boundary artifacts.](fig/reconstruction.png){#fig: reconstruction width="100%"} + +Our proposed method tackles these challenges in a unified way by training diffusion models on patches of images, as opposed to whole images (see ). We provide the location of the randomly extracted patch to the network to help it learn global image properties. Since each training image contains many patches, the required size of the training dataset is greatly reduced, from the millions usually needed to generate high quality images to only a couple thousand or even several hundred (see Tab. [\[datasetsize_table\]](#datasetsize_table){reference-type="ref" reference="datasetsize_table"}). The required memory and training time is also reduced because it is never necessary to backpropagate the whole image through the network. Our proposed method allows for a flexible network architecture and only requires it to accept images of any size, a property true of many UNets [@ho:20:ddp], so there is much more flexibility in the architecture design than fine-tuning methods. + +At inference time (see ), by first zero padding the image, the proposed approach partitions it into patches in many different ways (see ), eliminating the boundary artifacts between patches that would appear if non-overlapping patches were used. We develop a method to express the distribution of the whole image in terms of the patch distribution that is learned by the proposed patch-based network. By incorporating positional information of patches, this framework allows us to compute the score function of the whole image without ever inputting the whole image into the network. Unlike previous patch-based works that may be task-specific  [@xia2022patchbased; @xia2024parallel; @visionpatch], the prior defined by our approach may be treated in a black box manner and then paired with any other stochastic differential equation (SDE) solver to sample from the prior, or with any general purpose inverse problem solver to perform image reconstruction. We conduct experiments on multiple datasets and different inverse problems and demonstrate that the proposed method is able to synthesize the patches to produce reasonably realistic images and very accurate reconstructions for inverse problems. Furthermore, PaDIS provides a promising avenue for which generation and inverse problem solving of very large and high dimensional images may be tackled in the future. + +In summary, our main contributions are as follows: + +- We provide a theoretical framework whereby a score function of a high-resolution high-dimensional image is learned purely through the score function of its patches. + +- The proposed method greatly reduces the amount of memory and training data needed compared to traditional diffusion models. + +- The trained network has great flexibility and can be used with many existing sampling algorithms and is the first patch-based model that can solve inverse problems in an unsupervised manner. + +- We perform experiments on a variety of inverse problems to show superior image quality over whole image diffusion model methods while being far less resource heavy. + +# Method + +To be able to solve large 2D imaging problems as well as 3D and 4D inverse problems, our goal is to develop a model for $p(\x)$ that does not require inputting the entire image into any neural network. We would like to simply partition into nonoverlapping patches, learn the distribution of each of the patches, and then piece them together to obtain the distribution of . However, this would result in boundary artifacts between the patches. Directly using overlapping patches would result in sections of the image covered by multiple patches to be updated multiple times, which is inconsistent with the theory of diffusion models. Ideally, we would like to use nonoverlapping patches to update , but with a variety of different patch tiling schemes so that boundaries between patches do not remain the same through the entire diffusion process. + +To accomplish this task, we zero pad $\x_0$ and learn the distribution of the resulting zero padded image. More precisely, consider an $N \times N$ image $\x_0$ and let $1 \leq P 0$, $\zeta_i>0$, $P, M, \y$ Initialize $\x \sim \mathcal{N}(0, \sigma_T^2 \I)$ Sample $z \sim \mathcal{N}(0, \sigma_t^2 \I)$ Set $\alpha_t = \epsilon \cdot \sigma_t^2$ Randomly select integers $i, j \in [1, M]$ For all $1 \le r \le (k+1)^2$, extract patch $\x_{i,j,r}$ Compute $D_{i,j,r} = D_\theta({\x_{i,j,r}, \sigma_t})$ Set $\s_{i,j,r} = (D_{i,j,r} - \x_{i,j,r})/\sigma_t^2$ Apply [\[eq:patchupdate\]](#eq:patchupdate){reference-type="eqref" reference="eq:patchupdate"} to get $\s = \s(\x, \sigma_t)$ Set $\x$ to $\x - \zeta_t \nabla_{\x_t} \| \y - \mathcal{A}(D)\|_2^2$ Set $\x$ to $\x + \frac{\alpha_t}{2} \s + \sqrt{\alpha_t} \z$ +::: +:::: +::::: +:::::: + +For solving inverse problems with diffusion models, there are general algorithms e.g., [@chung:22:sbd] and [@chung:23:dps], as well as more model-specific algorithms, e.g., [@DDRM] and [@DDNM]. Here, we demonstrate that our method applies to a broad range of inverse problems, and opt to use generalizable methods that do not rely on any special properties (such as the singular value decomposition of the system matrix as in [@DDRM], [@DDNM]) of the forward operator. We found that DPS [@chung:23:dps] in conjunction with Langevin dynamics generally yielded the most stable and high quality results, so we use this approach as our central algorithm. Similar to [@chung:23:dps], we chose the stepsize to be $\zeta_i = \zeta/ \| \y - \mathcal{A}(D(\x))\|_2$. To the best of our knowledge, this is the first work that learns a fully patch-based diffusion prior and applies it to solve inverse problems; we call our method **Pa**tch **D**iffusion **I**nverse **S**olver (PaDIS). Computing the score functions of the patches at each iteration is easily parallelized, allowing for fast sampling and reconstruction. Alg. [\[alg:patch3d\]](#alg:patch3d){reference-type="ref" reference="alg:patch3d"} shows the pseudocode for the main image reconstruction algorithm; the appendix shows the pseudocode for the other implemented algorithms. + +Finally, we comment on some high-level similarities between our proposed method and [@lee2023improving]; in both cases, the image in question is partitioned into smaller parts in multiple ways. In [@lee2023improving], one of the partitions consists of 2D slices in the x-y direction, and the other partition consists of 2D slices in the x-z direction, whereas for our method, each of the partitions consists of $(k+1)^2$ square patches and the outer border region. For both methods, the score functions of each of the parts are learned independently during training. Then for each sampling iteration, both methods involve choosing one of the partitions, computing the score functions for each of the parts independently, and then updating the entire image by updating the parts separately. For our approach, the zero-padding method allows for many possible partitions of the image and eliminates boundary artifacts between patches. diff --git a/2408.01091/main_diagram/main_diagram.drawio b/2408.01091/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..c3ecb120b88f8c5f0fcc096f4aae94584504d879 --- /dev/null +++ b/2408.01091/main_diagram/main_diagram.drawio @@ -0,0 +1,19 @@ + + + + + + + + + + + + + + + + + + + diff --git a/2408.01091/main_diagram/main_diagram.pdf b/2408.01091/main_diagram/main_diagram.pdf new file mode 100644 index 0000000000000000000000000000000000000000..2f214e659ec6eb9d82ea367ed03adef55b877172 Binary files /dev/null and b/2408.01091/main_diagram/main_diagram.pdf differ diff --git a/2408.01091/paper_text/intro_method.md b/2408.01091/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..e1e9101013268b85c70a6ee813a6f6117bfc2d64 --- /dev/null +++ b/2408.01091/paper_text/intro_method.md @@ -0,0 +1,39 @@ +# Introduction + +Large multimodal models (LMMs) have become prominent for their exceptional ability to follow human instructions [1,4,12,26,29,31,32,38]. Designed to process various data types, LMMs can generate and understand content in a human-like way, aligning closely with human cognition through extensive research and development [3,14,44,45]. This focus on following human instructions has led to high compliance, sometimes verging on sycophancy [9,36,40]. + +LMMs are also rapidly developing to expand context windows and strengthen multimodal interaction. The Claude 3 family of models [1] offers a 200K token context window. Gemini 1.5 Pro [12] comes with a standard context window size of 128K (even up to 1M tokens in a private preview phase). Both models have sophisticated vision capabilities and can process a wide range of visual formats, including photos, figures, graphs, and technical diagrams. New multimodal models are emerging at a fantastic speed, demonstrating unprecedented performance in tackling long-context and multimodal instructions [11, 12, 20, 22, 25, 32]. + +However, self-contradictory instructions may arise due to the increasing trend of multimodal interaction and context window expansion, which is particularly challenging for language beginners and vulnerable populations. As shown in Fig. 1, children or language beginners may not realize the potential multimodal conflicts when LMMs are used in translation and education. It is also difficult for users to remember all details in multi-round conversations to avoid in- + +struction contradiction, especially when the context window size grows to 1M tokens and beyond. Moreover, conflicts between modalities may occur as the number of modalities gradually increases. Such conflicts may compromise the performance of LMMs once they fail to own meta-awareness [2] and to recognize the dissonance. Such self-awareness raises attention from researchers who attempt to enhance from the model level, while instruction-level studies are overlooked [7, 23, 43, 47]. + +Hence, we propose a multimodal benchmark, Self-Contradictory Instructions (SCI), to evaluate the ability of LMMs to detect conflicted instructions4. It encompasses 20K conflicting instructions and 8 tasks, evenly distributed between language-language and vision-language paradigms (Fig. 2). SCI is constructed using our novel automatic dataset creation framework, AUTOCREATE (Fig. 3), which builds a multimodal cycle based on programs and large language models. We have rigorously guaranteed the quality of SCI and manually provide three levels of splits according to the occurring frequency of conflict types, SCI-CORE (1%), SCI-BASE (10%), and SCI-ALL (100%), to facilitate qualitative evaluation. AUTOCREATE expedites the dataset creation process and enables the inclusion of a wide array of instruction forms, complexities, and scopes. + +Based on SCI, we assess the capability to decipher self-contradictory instructions for current LMMs, including 5 language and 6 vision-language models. Experiments reveal that LMMs consistently fall short of accurately identifying conflicts despite remarkable performance in following instructions. Besides, we observe that such deficiency persists owing to a lack of self-awareness. Although the training process enables LMMs to handle information and knowledge but not to assess the reasonableness of user instructions and context, a capability we term *cognition*. Hence, we propose a plug-and-play prompting approach, Cognitive Awakening Prompting (CAP), to inject cognition from the external world, thereby largely enhancing dissonance detection even compared with advanced in-context learning techniques [5, 42, 46]. CAP is demonstrated to improve performance on both language-language and vision-language instruction conflicts. + +- We propose the SCI benchmark, a multimodal dataset designed to evaluate the capability of LMMs to comprehend conflicting instructions effectively. +- We design a novel LLM-based cyclic framework, AUTOCREATE, for automatic dataset creation, substantially accelerating the process and allowing for the integration of extensive knowledge. +- We present CAP, a prompting approach to enhance instruction conflict awareness of LMMs, significantly improving dissonance detection compared to advanced in-context learning techniques. + +Website: https://sci-jingao.pages.dev +Dataset: https:// huggingface.co/datasets/sci-benchmark/self-contradictory +Code: https://github.com/shiyegao/Self-Contradictory-Instructions-SCI + +# Method + +In this section, we delve into our exploration using in-context learning techniques, detailed in Section 4.1. Through experiments across various Large Multimodal Models (LMMs), we've pinpointed a crucial challenge where LMMs struggle to detect instruction conflicts. Additionally, we introduce our proposed Cognitive Awakening Prompting (CaP) approach, outlined in Section 4.2. + +We study three in-context learning techniques in SCI, including few-shot prompting [5], zero-shot chain-of-thoughts prompting [18], and self-consistency prompting [42]. Although few-shot prompting has been widely used in Large Language Models, its application in Large Multimodal Models (LMMs) remains limited. Recent research highlights challenges such as LMMs' inability to support multiple image inputs or comprehend sophisticated few-shot prompts [17,50]. Consequently, few-shot prompting is primarily employed within the language-language paradigm. Here, we detail the application of these prompting techniques in our SCI. + +Zero-shot Prompting refers to the ability of the model to perform a task without providing examples of how to perform a task correctly. We task the model with generating responses in SCI solely based on its general knowledge and understanding of language and vision. This capability underscores the model's innate capacity to detect self-contradictory conflicts. + +Zero-shot Chain-of-thoughts Prompting [46] (CoT) involves appending text like "Please think step by step" to user prompts, proven to enhance LMMs' inference ability. In our experiment, we incorporate this text into the prompt. + +Self-consistency Prompting [42] (SC) involves sampling multiple reasoning paths and selecting the most consistent answers. In this paper, we generate three replies for each instruction (3-SC) and determine the final result through majority voting. + +Our initial exploration reveals an intriguing phenomenon: the performance order in vision-language tasks is 0-Shot, CoT, and 3-SC across diverse LMMs, shown in the Appendix. While 3-SC provides additional experience through more attempts, CoT offers extra knowledge by stimulating reasoning capabilities through a chain of thought. However, neither surpasses the simplicity of zeroshot prompting, suggesting that both additional experience and extra knowledge derived from the model itself may be counterproductive. We hypothesize that the LMMs may not fully grasp restricted cognition in the self-contradictory instruction scenarios. + +Therefore, we propose a plug-and-play prompting approach to infuse cognition from the external world: Cognitive Awakening Prompting (CaP). The externally added cognition prompt reminds LMMs of potential inconsistencies hidden in their cognition, e.g., adding "Please be careful as there may be inconsistency in user input. Feel free to point it out." at the end of the prompt. The injected cognition does not impair the basic functioning of LMMs but fosters selfawareness of internal information and knowledge defects. Detailed experiments are presented in Section 5.3. + +While CaP stems from observation and analysis in vision-language tasks, it also demonstrates promise in language-language tasks, outperforming 3-Shot in over half of LMMs. Generally, the 3-Shot provides extra information since more question-answer pairs are provided. This underscores that cognition represents a higher level of existence than experience, information, and knowledge. CaP embodies a prompting technique standing on the cognition dimension, enabling the identification of LMMs' shortcomings and exploration of profound issues. Detailed experiments are outlined in Section 5.2. diff --git a/2409.02718/main_diagram/main_diagram.drawio b/2409.02718/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..de842664e4002843db97cf79a3cd8543b2c9e769 --- /dev/null +++ b/2409.02718/main_diagram/main_diagram.drawio @@ -0,0 +1,572 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/2409.02718/main_diagram/main_diagram.pdf b/2409.02718/main_diagram/main_diagram.pdf new file mode 100644 index 0000000000000000000000000000000000000000..bc054d4a76260798e2ab4de5d73c08d518d3e414 Binary files /dev/null and b/2409.02718/main_diagram/main_diagram.pdf differ diff --git a/2409.02718/paper_text/intro_method.md b/2409.02718/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..25e8b5333cb83aafcfca9c89070b48799e6f32b1 --- /dev/null +++ b/2409.02718/paper_text/intro_method.md @@ -0,0 +1,5 @@ +# Method + +**Query Detection.** One approach to effectively prevent the attack of LoRD is by detecting the distribution of query texts. This is because LoRD, similar to current MEA algorithms, makes no improvements to query samples, indicating that it can be detected by analyzing the statistical information of the adversary's queries, such as the number of queries, distribution of query contents, and so on. However, this defense is usually resource-consuming, as it requires the LLM provider to store all query texts of each user. Besides, the potential for false positives could adversely affect the user experience. + +**More Powerful Watermarks.** While we highlight the watermark resistance of LoRD, watermarking remains one of the most effective solutions to mitigate MEAs. For example, some model-level watermarks, such as backdoor-based watermarking [@backdoor-mea-wm; @mea-defender], can effectively certify the theft of DNNs. However, model-level watermarking on LLMs remains preliminary. Besides, this technique might not work when the adversary only steals a subset of knowledge in which no backdoor is embedded. diff --git a/2410.02506/main_diagram/main_diagram.drawio b/2410.02506/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..fd20be73020eae7075d86137289b6232315ac525 --- /dev/null +++ b/2410.02506/main_diagram/main_diagram.drawio @@ -0,0 +1,576 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/2410.02506/paper_text/intro_method.md b/2410.02506/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..8e8af8c8cb55f372f5d9d15b69e8f969ecba3432 --- /dev/null +++ b/2410.02506/paper_text/intro_method.md @@ -0,0 +1,107 @@ +# Introduction + +::: wrapfigure +r0.55 ![image](img/map_intro.drawio.png){width="\\linewidth"} +::: + +Large Language Model (LLM) based agents [@autogpt; @babyagi; @agentgpt] have demonstrated strong performance across a diverse range of tasks, including reasoning [@yao2023react], code generation [@reflexion], and even more complex applications like video gaming [@voyager] and autopilot systems [@jin2023surrealdriver]. Recent endeavors have shown that combining implicitly or explicitly different LLM-based agents into a team can outperform a single agent in handling complex tasks [@arXiv2023_MultiAgent-Debate; @arXiv2023_MultiAgent-Debate_2; @multi-persona; @blender; @reflexion; @PHPrompting; @autogen], which supports the presence of human-esque collaborative intelligence in multi-agent systems [@zhang2023exploring]. In practice, previous research has explored approaches in which instances of LLMs, referred to as agents [@FCS2024_Survey-Agent; @arXiv2023_Survey-Agent_2; @arXiv2023_Survey-Agent_3; @arXiv2024_Survey-Agent_4; @arXiv2024_Survey-Agents-CompExp], collaborate synergistically (*e.g.*, through debate or reflection) to complete tasks [@arXiv2023_Survey-MultiAgentCooperation; @arXiv2024_Survey-MultiAgent_2; @arXiv2024_Survey-MultiAgent; @arXiv2024_Survey-MultiAgent-System; @arXiv2024_Survey-MultiAgent-System_2] via diverse communication topologies (*e.g.*, chain [@cot], tree [@tot], complete graph [@qian2024scaling], random graph [@qian2024scaling], optimizable graph [@zhuge2024gptswarm], LLM-based network [@chatllm-network; @arXiv2023_Dynamic-LLM-Agent]). The exceptional performance of these cooperative agents significantly benefits from their interactive communication and collaboration, specifically how agents *transmit, exchange,* and *assimilate* information [@chateval; @multi-persona; @arXiv2023_Dynamic-LLM-Agent]. + +![(***Left***) The accuracy comparison on MMLU [@mmlu] among (1) a single , (2) three as agents equipped with intra-dialogue communication structures like chain and tree [@qian2024scaling], complete graph, and GPTSwarm [@zhuge2024gptswarm], and (3) those equipped with inter-dialogue communication structures like PHP [@PHPrompting], LLM-Debate [@arXiv2023_MultiAgent-Debate], DyLAN [@arXiv2023_Dynamic-LLM-Agent] and our **AgentPrune**. (***Middle***) The prompt token consumption comparison on MMLU between different methods. (***Right***) The overview of our proposed **AgentPrune**.](img/intro_final.pdf){#fig:intro_2 width="\\linewidth"} + +Taking a closer look into the communication mechanisms in existing multi-agent systems, they typically involve two key types (as shown in [\[fig:intro\]](#fig:intro){reference-type="ref+Label" reference="fig:intro"}): **Intra-dialogue communication**: For a given query/task, multiple agents interact--whether by cooperating [@arXiv2023_Survey-MultiAgentCooperation; @autogen; @rasal2024llm], teaching [@zhang2024classroom], or competing [@zhao2023competeai; @bargaining-feedback]--to produce a solution within a dialogue round; **Inter-dialogue communication**: In a specific manner---whether by summarizing [@chateval; @shen2024smallllms], replicating [@yin2023exchange; @arXiv2023_MultiAgent-Debate], or filtering [@arXiv2023_Dynamic-LLM-Agent]---the content of the current dialogue is passed to the next round of interaction as a reference, initiating a new cycle of collaborative efforts. + +To better illustrate the power of agent communication, [1](#fig:intro_2){reference-type="ref+Label" reference="fig:intro_2"} (*Left*) compares the performance of a single with three agents equipped with different inter/intra-dialogue communication structures. The results demonstrate that even the simplest communication framework significantly leads to a notable accuracy improvement, which vividly showcases the social intelligence and collaborative capabilities of LLMs [@PNAS2024_TuringTest_Chatbots-Humans]. However, the success of multi-agents comes at the cost of significantly increased token consumption, imposing substantial economic burdens [@wang2024tokeneconomy], which are detrimental to the widespread application of multi-agent systems, as deployment on edge devices does not accommodate excessively costly inference [@liu2023deja]. A piece of empirical evidence is in [1](#fig:intro_2){reference-type="ref+Label" reference="fig:intro_2"} (*Middle*), where various communication methods result in a $2\sim11.8\times$ increase in token consumption compared to the simple chain structure, severely undermining the **token economy** of existing multi-agent systems. + +In the light of this limitation, we *for the first time* identify a significant phenomenon of *Communication Redundancy* [@meyer2021entspann] within existing LLM-based multi-agent (LLM-MA) communication topologies, where a substantial portion of message passing does not contribute meaningfully to the collaborative intelligence. With this finding, we introduce *an economical and versatile communication pruning framework for LLM-powered multi-agent systems*, dubbed **AgentPrune**, which can be smoothly incorporated within various existing LLM-MA systems, offering comparable reasoning and planning performance as well as significantly lower token consumption. Practically, **AgentPrune** treats the entire LLM-MA framework as a spatial-temporal communication graph, in which each agent, along with its unique properties (*e.g.*, profile [@NeurIPS2023_Agent-SoM], external API tools [@zhuang2023toolchain], or knowledge base [@chen2024benchmarking]), is packaged as a node, communication between agents within the same dialogue forms spatial edges, and communication across dialogues forms temporal edges. By training a low-rank-principle-guided graph mask, **AgentPrune** efficiently identifies the important graph connectivities (*i.e.*, message passing through edges). This comes with a one-shot pruning to derive a sparse yet informative communication graph (in [1](#fig:intro_2){reference-type="ref+Label" reference="fig:intro_2"} (*Right*)), which is then fixed as the communication topology for subsequent token-economic and efficient reasoning. Our contributions can be summarized as follows: + +- ***System Discovery.*** We present a spatial-temporal graph paradigm to describe the communication topology of contemporary LLM-MA frameworks, and further identify and define the *Communication Redundancy* issue in current systems, wherein a significant portion of spatial and temporal edges, *i.e.*, communication, does not contribute to collaborative intelligence. + +- ***Pratical Solution.*** We propose **AgentPrune**, an economical, simple, and robust multi-agent communication pruning pipeline. By leveraging a trainable communication graph mask, **AgentPrune** identifies key message exchanges and prunes non-essential components in a one-shot manner, resulting in a sparse, token-economical, and highly informative communication graph. Notably, **AgentPrune** employs a low-rank principle to guide the graph mask training, successfully robustifying LLM-MA systems against various agent-targeted adversarial attacks. + +- ***Experimental Validation.*** Extensive experiments on six benchmarks show that **AgentPrune** is: **(1) high-performing**, achieving comparable performance on MMLU at $\$5.6$ cost, to that of state-of-the-art communication topologies at $\$43.7$; **(2) token-economical**, integrating seamlessly into popular multi-agent frameworks including AutoGen and GPTSwarm, reducing their token cost by $28.1\%\sim72.8\%\downarrow$; and **(3) adversarially robust**, successfully defending against two types of agent adversarial attacks, with a $3.5\%\sim10.8\%\uparrow$ performance improvement. + +We describe the whole multi-agent system as a graph $\mathcal{G}=(\mathcal{V},\mathcal{E})$, with $\mathcal{V}=\{v_1,v_2,\cdots,v_{|\mathcal{V}|}\}$ being the node set and $\mathcal{E}$ being the edge set. Each node $v_i\in \mathcal{V}$ represents an agent, which can be further interpreted as follows: $$\begin{equation} +v_i = \{\texttt{Base}_i, \texttt{Role}_i, \texttt{State}_i, \texttt{Plugins}_i\},\;\;\texttt{Plugins}_i=\{\texttt{F}_j,\texttt{C}_j\}_{j=1}^{P} , +\vspace{-0.2em} +\end{equation}$$ where an agent $v_i$ consists of: (1) $\texttt{Base}_i$, the language model instance used by $v_i$; (2) $\texttt{Role}_i$, the pre-defined role or responsibility of the agent; (3) $\texttt{State}_i$, the state of the agent, encapsulating the accumulated knowledge and experience from previous interactions; (4) $\texttt{Plugins}_i$, a set of $P$ external plugins available to agent $v_i$, where each plugin is defined by its functionalities $\texttt{F}_j$ (*e.g.*, web search, python compiler) and configurations $\texttt{C}_j$. + +For the edges, we divide them into two subsets: **intra-dialogue (spatial) edges** $\mathcal{E}^{\mathcal{S}}\subseteq\mathcal{V}^{(t)}\times\mathcal{V}^{(t)}$ and **inter-dialogue (temporal) edges** $\mathcal{E}^{\mathcal{T}}\subseteq \mathcal{V}^{(t-1)}\times\mathcal{V}^{(t)}$. For each spatial edge $e^{\mathcal{S}}_{ij} = (\mathbf{M}_{ij}, \mathbf{O}_{ij})$, it represents the information flow from agent $v_i$ to agent $v_j$ *within the same utterance*, composed of the message content $\mathbf{M}_{ij}$, as well as the associated operation $\mathbf{O}_{ij}$ (*e.g.*, task assignments, requests). For each temporal edge $e^{\mathcal{T}}_{ij}$, it denotes the message passing *between two utterance rounds*, *i.e.*, whether the output from agent $v_i$ in the $(t-1)$-th round should be passed on to agent $v_j$ in the $t$-th round. Further, we define the temporal/spatial (in-)neighbors for each agent as follows: $$\begin{equation} +\mathcal{N}^\mathcal{T}(v_i) = \{v_j\;|\;(j,i)\in\mathcal{E}^\mathcal{T}\},\;\mathcal{N}^\mathcal{S}(v_i) = \{v_j\;|\;(j,i)\in\mathcal{E}^\mathcal{S}\}. +\vspace{-0.5em} +\end{equation}$$ + +We provide a graph-based description of the reasoning process in task-oriented multi-agent systems. Given a query/task $q$, it is sequentially fed to each agent, which produces its output. To maintain an orderly sequence of agent interactions, we utilize topological ordering [@bondy1976graph] to ensure that each node is processed only after all its dependencies have been addressed. This necessitates that the spatial communication graph $\mathcal{G}^{\mathcal{S}} = (\mathcal{V}, \mathcal{E}^{\mathcal{S}})$ be structured as a directed acyclic graph (DAG). Formally, for $\mathcal{G}^{\mathcal{S}}$, the following condition holds: $$\begin{equation} +\forall (v_i ,v_j),\;\mathbb{I}(v_i) < \mathbb{I}(e_{ij}) < \mathbb{I} (v_j), +\end{equation}$$ where $\mathbb{I}(x)$ denotes the execution order of $x$. For each agent $v^{(t)}_i$ at round $t$, it produces its rationale or answers, uniformly denoted as $\mathbf{M}_{i}$, as follows: $$\begin{equation} +\label{eq:agent_action} +\mathbf{M}^{(t)}_i \sim \mathcal{P}_\theta\Biggl(\mathbf{M}_i\;|\;q,\; \texttt{Role}^{(t)}_i,\; \texttt{State}^{(t)}_i, \overbrace{\cup_{\scriptscriptstyle v_j \in \mathcal{N}^{\mathcal{T}}(v_i)}\!\!\!\!\!\mathbf{M}_{j}}^{{\text{temporal}}}, \overbrace{\cup_{\scriptscriptstyle v_j \in \mathcal{N}^{\mathcal{S}}(v_i)}\!\!\!\!\mathbf{M}_{ji}}^{{\text{spatial}} }\Biggl), +\vspace{-0.8em} +\end{equation}$$ where agent $v_i$ responds based on the query $q$, its current role and state, temporal and spatial messages, and certain prompting instruction $\mathcal{P}_\theta$. Typically, after $K$ rounds of dialogue, a summarizer agent or an answer aggregation mechanism (*e.g.*, voting) is employed to produce the final solution $a^{(K)}$ for the given query $q$. We conclude the general pipeline in [\[algo:llmma\]](#algo:llmma){reference-type="ref+Label" reference="algo:llmma"}. + +::: algorithm +$a^{(t)}$ as the final solution +::: + +# Method + +::: wrapfigure +r0.55 ![image](img/preliminary.pdf){width="\\linewidth"} +::: + +In this section, we explore and define the *communication redundancy* issue within existing multi-agent communication pipelines. Specifically, we examine two representative communication topologies: (1) for *spatial communication*, the fully-connected mesh graph from MacNet [@qian2024scaling], which exemplifies a densely structured intra-utterance communication, and (2) for *temporal communication*, the LLM-Debate [@arXiv2023_MultiAgent-Debate], where at the start of each dialogue round, an agent receives all responses from the previous round as input. Using four as agents, we assess system performance on MMLU after randomly pruning a certain proportion of connections. As illustrated in [\[fig:preliminary\]](#fig:preliminary){reference-type="ref+Label" reference="fig:preliminary"}, when randomly removing $10\%\sim30\%$ of the communication connectivity, the performance actually gains up to $2.83\%$ improvement. This suggests that, in both spatial and temporal information flow, a substantial portion of messages does not contribute to the task-solving process, which we formally define as follows: + +::: definition +**Definition 1** (Communication Redundancy). *For any LLM-based multi-agent communication graph $\mathcal{G} = (\mathcal{V}, \mathcal{E}^\mathcal{S} \cup \mathcal{E}^\mathcal{T})$, the following condition holds: $$\begin{equation} +\exists\; \mathcal{G}^\text{sub} = (\mathcal{V}, \mathcal{E}' \cup \mathcal{E}'') \subseteq \mathcal{G}, \; \operatorname{where} \; \mathcal{E}' \subseteq \mathcal{E}^\mathcal{S}, \; \mathcal{E}'' \subseteq \mathcal{E}^\mathcal{T}, \; \operatorname{s.t.} \; \phi(\mathcal{G}^\text{sub}) \geq \phi(\mathcal{G}), +\end{equation}$$ where $\phi(\cdot)$ represents a utility function that measures the solution quality achieved by the system. The redundant components in the communication topology, denoted as $(\mathcal{E}^{st} \setminus \mathcal{E}') \cup (\mathcal{E}^{tp} \setminus \mathcal{E}'')$, are referred to as the communication redundancy in LLM-MA systems.* +::: + +We further outline the objective of this study as follows: $$\begin{equation} +\label{eq:problem_definition} +{\arg \max}_{\mathcal{E}',\mathcal{E}''} \mathcal{G} \setminus \mathcal{G}^\text{sub},\;\text{s.t.} \; |\phi(\mathcal{G}^\text{sub}) - \phi(\mathcal{G})| \leq \epsilon, +\end{equation}$$ where $\epsilon$ represents the allowable threshold for performance variation. [\[eq:problem_definition\]](#eq:problem_definition){reference-type="ref+Label" reference="eq:problem_definition"} aims to minimize communication redundancy with performance guarantee. + +[2](#fig:framework){reference-type="ref+Label" reference="fig:framework"} illustrates how our method is applied within an LLM-MA system. Specifically, given an input query, **AgentPrune** first performs *spatial pruning* by eliminating redundant spatial messages within a dialogue round, followed by *temporal pruning* to discard unnecessary dialogue history. In the following sections, we will first explain how **AgentPrune** facilitates efficient multi-round communication based on an optimizable spatial-temporal communication graph ($\vartriangleright$ [\[sec:stgraph,sec:optimize_conn\]](#sec:stgraph,sec:optimize_conn){reference-type="ref+Label" reference="sec:stgraph,sec:optimize_conn"}), leverages one-shot pruning to derive a sparse interaction topology ($\vartriangleright$ [3.3](#sec:one-shot){reference-type="ref+Label" reference="sec:one-shot"}), and finally, detail the optimization paradigm for the entire framework ($\vartriangleright$ [3.4](#sec:opt){reference-type="ref+Label" reference="sec:opt"}). + +
+ +
The overview of our proposed AgentPrune.
+
+ +Given an arbitrary LLM-MA system and its corresponding spatial-temporal communication graph $\mathcal{G}$, the task of **AgentPrune** is to discover its sparse yet equally high-performing counterpart $\mathcal{G}^{\text{sub}}$. The objective is essentially a graph sparsification problem [@spielman2008graph; @chen2023demystifying], whose goal is to identify the essential graph connections and discard the less critical ones. To achieve this, following classical practices in graph sparsification [@chen2021unified; @wang2023brave; @zhang2024graph], we relax the original binary communication graph $\mathcal{G}$ by transforming its edge elements from binary values to continuous variables, denoted as $\Tilde{\mathcal{G}}$. We have: $$\begin{equation} +% \begin{aligned} +\mathbf{A}({\mathcal{G}}) = \{\mathbf{A}^\mathcal{S}, \mathbf{A}^\mathcal{T}\},\;\mathbf{A}(\Tilde{\mathcal{G}}) = \mathbf{A}(\{\Tilde{\mathcal{G}}^\mathcal{S},\Tilde{\mathcal{G}}^\mathcal{T}\}) = \{\mathbf{A}^\mathcal{S}\odot \mathbf{S}^\mathcal{S}, \mathbf{A}^\mathcal{T}\odot \mathbf{S}^\mathcal{T}\}, +% \end{aligned} +\end{equation}$$ where $\mathbf{A}(\mathcal{G})$ obtains the adjacency matrix of input graph $\mathcal{G}$, and $\mathbf{A}^{\mathcal{S}}, \mathbf{A}^{\mathcal{T}}\in \{0,1\}^{|\mathcal{V}|\times|\mathcal{V}|}$ represent the spatial and temporal adjacency matrices, respectively. Specifically, $\mathbf{A}^{x}[i,j] = 1$ indicates that $e_{ij} \in \mathcal{E}^{x}$, and $0$ otherwise. It is important to note that both $\mathbf{A}^{\mathcal{S}}$ and $\mathbf{A}^{\mathcal{T}}$ are predefined by the LLM-MA system. $\mathbf{S}^\mathcal{S}, \mathbf{S}^\mathcal{T}\in\mathbb{R}^{|\mathcal{V}|\times|\mathcal{V}|}$ are differentiable graph masks. As mentioned in [2.0.0.2](#sec:multi-agent-graph){reference-type="ref+Label" reference="sec:multi-agent-graph"}, we require the interaction topology to be a DAG to ensure that agent input/output (I/O) can be processed sequentially. Therefore, we leverage a `DAGSampling` function to transform the original $\Tilde{\mathcal{G}}^\mathcal{S}$ into a DAG: $\hat{\mathcal{G}}^\mathcal{S}\leftarrow \texttt{DAGSampling}(\Tilde{\mathcal{G}}^\mathcal{S})$, whose procedure is described in [\[algo:dag\]](#algo:dag){reference-type="ref+Label" reference="algo:dag"}. + +With $\hat{\mathcal{G}}^{\mathcal{S}}$ and $\Tilde{\mathcal{G}}^\mathcal{T}$ in hand, we aim to optimize them toward both high-performance and token efficiency. To this end, we introduce two optimization objectives for $\hat{\mathcal{G}}^{\mathcal{S}}$ and $\Tilde{\mathcal{G}}^\mathcal{T}$: *distribution approximation*, ensuring accurate estimation of their underlying probability distributions, and *low-rank sparsity*, which promotes a more efficient and sparse structure [@li2024gslb]. The first objective ensures that the magnitudes of the graph masks correctly reflect the importance of different communication channels, facilitating subsequent redundancy pruning, and the second ensures that the learned connectivity remains sparse and robust. Formally, we define the following objective: $$\begin{equation} +\label{eq:objective} +\underset{\mathbf{S}^\mathcal{S},\mathbf{S}^\mathcal{T}\in \mathbb{S}}{\arg \max}\; \overbrace{\mathbb{E}_{\hat{\mathcal{G}}^{\mathcal{S}},\mathcal{G}^\mathcal{T}\sim\mathbb{G}}\left[\phi\left(\{\hat{\mathcal{G}}^{\mathcal{S}},\Tilde{\mathcal{G}}^\mathcal{T}\}\right)\right]}^\text{distribution approximation}-\!\!\overbrace{\sum_{\scriptscriptstyle \mathcal{X}\in\{\mathcal{S},\mathcal{T}\}}\!\!\operatorname{rank}(\mathbf{S}^\mathcal{X})}^\text{low-rank sparsity},\; \operatorname{s.t.}\!\!\!\!\sum_{\scriptscriptstyle \mathcal{X}\in\{\mathcal{S},\mathcal{T}\}}\!\!\!||\mathbf{A}^\mathcal{X} - \mathbf{S}^\mathcal{X}||_F\leq \delta, +% ||\mathbf{A}^\mathcal{S} - \mathbf{S}^\mathcal{S}||_F + +\end{equation}$$ where $\mathbb{S}$ and $\mathbb{G}$ represent the viable parameter space, $\phi(\cdot)$ serves as the utility evaluator for the input multi-agent framework, $\operatorname{rank}(\cdot)$ calculates the rank of matrix, and $\delta$ is the noise level. Next, we will provide a detailed explanation of the implementation of these two optimization objectives. + +The first term in [\[eq:objective\]](#eq:objective){reference-type="ref+Label" reference="eq:objective"} encourages $\{\mathbf{S}^\mathcal{S},\mathbf{S}^\mathcal{T}\}$ towards the maximization of the system's utility. However, since $\phi(\cdot)$ often depends on external APIs [@li2023api] or compilers [@human-eval] for evaluation, it is generally non-differentiable. Therefore, we employ policy gradient [@williams1992simple] to make [\[eq:objective\]](#eq:objective){reference-type="ref+Label" reference="eq:objective"} tractable: $$\begin{gather} +\label{eq:policy} +\nabla_{\mathbf{S}}\; \mathbb{E}_{\hat{\mathcal{G}}^{\mathcal{S}},\Tilde{\mathcal{G}}^\mathcal{T}\sim\mathbb{G}}\left[\phi\left(\{\hat{\mathcal{G}}^{\mathcal{S}},\Tilde{\mathcal{G}}^\mathcal{T}\}\right)\right] \approx \frac{1}{M}\sum_{k=1}^M\phi\left(\{\hat{\mathcal{G}}^\mathcal{S}_k, \Tilde{\mathcal{G}}^\mathcal{T}_k\}\right)\nabla_{\mathbf{S}}\log \left(p_{\mathbf{S}}(\{\hat{\mathcal{G}}^\mathcal{S}_k, \Tilde{\mathcal{G}}^\mathcal{T}_k\})\right),\\ +p_\mathbf{S}\left(\{\hat{\mathcal{G}}^\mathcal{S}_k, \Tilde{\mathcal{G}}^\mathcal{T}_k\}\right) = \left(\prod \mathbbm{1}_{e_{ij}\in \mathcal{E}^\mathcal{S}} \mathbf{S}^\mathcal{S}[i,j]\right) \cdot \left(\prod \mathbbm{1}_{e_{ij}\in \mathcal{E}^\mathcal{T}} \mathbf{S}^\mathcal{T}[i,j]\right) +\end{gather}$$ where $\mathbf{S} = \{\mathbf{S}^\mathcal{S}, \mathbf{S}^\mathcal{T}\}$, $\{\hat{\mathcal{G}}^{\mathcal{S}}, \Tilde{\mathcal{G}}^{\mathcal{T}}\}_{k=1}^M$ are independently sampled from $\{\hat{\mathcal{G}}^{\mathcal{S}}, \Tilde{\mathcal{G}}^\mathcal{T}\}$, $p_\mathbf{S}(\{\hat{\mathcal{G}}^\mathcal{S}_k, \Tilde{\mathcal{G}}^\mathcal{T}_k\})$ calculates the probability of the sampled structure, and $\mathbbm{1}(\cdot)$ is an indicator function. + +The second term in [\[eq:objective\]](#eq:objective){reference-type="ref+Label" reference="eq:objective"} promotes the graph masks $\{\mathbf{S}^\mathcal{S},\mathbf{S}^\mathcal{T}\}$ to be low-rank, which not only filters out informative agent communications but also aids in removing redundant, noisy, and even malicious messages, which has been demonstrated in recent studies, showing that low-rank graphs are more robust to network attacks [@entezari2020all; @ennadir2024simple]. We will empirically validate **AgentPrune**'s ability to enhance multi-agent robustness in [4.4](#exp:robust){reference-type="ref+Label" reference="exp:robust"}. However, directly optimizing the rank minimization is NP-hard, so we replace the rank function with the nuclear norm as an alternative, reformulating this term as follows: $$\begin{equation} +\label{eq:rank_1} +\underset{\mathbf{S}^\mathcal{S},\mathbf{S}^\mathcal{T}\in \mathbb{S}}{\arg \min} {\textstyle\sum}_{\scriptscriptstyle \mathcal{X}\in\{\mathcal{S},\mathcal{T}\}}||\mathbf{S}^\mathcal{X}||_*,\;\operatorname{s.t.} ||\mathbf{A}^\mathcal{X} - \mathbf{S}^\mathcal{X}||_F\leq \delta, +\end{equation}$$ where $||\mathbf{S}||_* = \sum_i \sigma_i$, and $\sigma_i$ represents the $i$-th singular value of $\mathbf{S}$. Guided by [\[eq:objective\]](#eq:objective){reference-type="ref+Label" reference="eq:objective"}, we iteratively optimize the spatial-temporal connectivity in conjunction with the multi-agent conversation over $K'$ rounds, where $K' \ll K$. + +We dynamically optimize $\{\mathbf{S}^\mathcal{S}, \mathbf{S}^\mathcal{T}\}$ for only $K'$ iterations, rather than the full $K$ iterations, because prior work on Early-bird (EB) and Graph EB has demonstrated that limited training can also construct high-quality benchmarks reflecting the topology distribution [@achille2018critical; @you2019drawing; @zhang2024two], which also aligns with **AgentPrune**'s token-saving initiation. To eliminate redundancy in the current communication structure, we perform one-shot magnitude pruning on the optimized graph masks $\mathbf{S}$ (either $\mathbf{S}^\mathcal{S}$ or $\mathbf{S}^\mathcal{T}$): $$\begin{equation} +\label{eq:oneshot} +% \begin{aligned} +\mathbf{B} = \mathbbm{1}\biggl(\mathbf{A} \neq 0 \wedge \operatorname{TopK}\Bigl(\mathbf{S}, |\mathbf{A}|\times \left(1-p\%\right)\Bigl)\biggl), +\end{equation}$$ where $\operatorname{TopK}(S,x\%)$ return the largest $x\%$ elements in matrix $S$, and $p\%$ is the pruning ratio. By applying the binary masks to the original topology, we obtain sparse, compact, and communication-minimizing connectivity $\mathcal{G}^\text{sub}$, where $\mathbf{A}(\mathcal{G}^\text{sub}) += \{\mathbf{A}^\mathcal{S} \odot \mathbf{B}^\mathcal{S}, \mathbf{A}^\mathcal{T}\odot \mathbf{B}^\mathcal{T}\}.$ In the subsequent $(K-K')$ rounds, the entire framework's message passing pipeline is strictly constrained by $\mathcal{G}^\text{sub}$, and agents are continuously optimized to refine the solution for query $q$. + +As a plug-and-play module, **AgentPrune** can be harmoniously embedded in mainstream multi-agent frameworks to facilitate token-efficient communication, provided that the number of agents exceeds three and the communication structure is moderately organized (*e.g.*, chain or direct-output structures are too simple to be applicable). When combined with **AgentPrune**, multiple agents first undergo $K'$ rounds of interactions alongside trainable graph masks, which are then one-shot pruned to yield the sparse $\mathcal{G}^\text{sub}$, leveraged for the subsequent $(K-K')$ rounds of optimization. We summarize all the notations used in [8](#app:notation){reference-type="ref+Label" reference="app:notation"} and the comprehensive algorithmic workflow in [9](#app:workflow){reference-type="ref+Label" reference="app:workflow"}. + +For complex tasks like repository-level code generation [@software-dev; @liu2024evaluating], multi-turn dialogues are often inevitable. However, for simpler tasks that involve a large number of queries, such as multiple choice answering [@agashe2023evaluating; @qian2024scaling], typically only one or two dialogue rounds are needed, according to previous practices [@yin2023exchange]. Under such circumstances, optimizing the connectivity for each query independently can be unnecessarily costly. Therefore, we give a *multi-query training paradigm* for **AgentPrune**, which optimizes and prunes the spatial-temporal topology using merely $Q' (Q'\ll Q)$ queries, given a dataset composed of $Q$ queries. See details in [10](#app:multi_query){reference-type="ref+Label" reference="app:multi_query"}. + +In this section, we quantify the difference in token consumption between **AgentPrune** and the vanilla pipeline. Given a communication graph $\mathcal{G}$ and $K$ dialogue rounds, assuming that the average token count per spatial/temporal/query message is $c_\mathcal{S}, c_\mathcal{T}, c_q$, respectively, then the total token consumption of the vanilla system is $C_\mathcal{G} = K\left[ c_\mathcal{S} |\mathcal{E}^\mathcal{S}| + c_\mathcal{T} |\mathcal{E}^\mathcal{T}| + C_q|\mathcal{V}| \right]$. The token consumption after applying **AgentPrune** is divided into two stages. The first stage involves $MK'\left[ c_\mathcal{S} |\mathcal{E}^\mathcal{S}| + c_\mathcal{T} |\mathcal{E}^\mathcal{T}| + C_q|\mathcal{V}| \right]$, while in the second stage, after the topology is fixed, the consumption becomes $(K-K')\left[ (1-p\%)\cdot\left(c_\mathcal{S} |\mathcal{E}^\mathcal{S}|+c_\mathcal{T} |\mathcal{E}^\mathcal{T}|\right) + C_q|\mathcal{V}| \right]$. Therefore, the total token savings $\Delta$ achieved by **AgentPrune** can be expressed as: $$\begin{equation} +\Delta = \Bigl( \left(1+p\%\right)K - \left(M+p\%\right)K' \Bigl) \Bigl(c_\mathcal{S} |\mathcal{E}^\mathcal{S}|+c_\mathcal{T} |\mathcal{E}^\mathcal{T}| \Bigl) + (1-M)K' C_q|\mathcal{V}|. \vspace{-0.5em} +\end{equation}$$ We present the cost analysis of **AgentPrune** in multi-query training in [11](#app:cost_analysis){reference-type="ref+Label" reference="app:cost_analysis"}. We will empirically evaluate the substantial token savings gained by **AgentPrune** in [4.2](#exp:perf_cost){reference-type="ref+Label" reference="exp:perf_cost"}. diff --git a/2410.24005/main_diagram/main_diagram.drawio b/2410.24005/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..b7b6020e8337171572cfa29b8b2f08e0211e1791 --- /dev/null +++ b/2410.24005/main_diagram/main_diagram.drawio @@ -0,0 +1,172 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/2410.24005/main_diagram/main_diagram.pdf b/2410.24005/main_diagram/main_diagram.pdf new file mode 100644 index 0000000000000000000000000000000000000000..489426627d93ffa656ba39ad54363066bfd24730 Binary files /dev/null and b/2410.24005/main_diagram/main_diagram.pdf differ diff --git a/2410.24005/paper_text/intro_method.md b/2410.24005/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..4b19533b94666fffc6ad0db1e6198d3fe2b36c51 --- /dev/null +++ b/2410.24005/paper_text/intro_method.md @@ -0,0 +1,225 @@ +# Introduction + +The ability to rigorously test and validate machine learning (ML) models is crucial for their reliable deployment in real-world applications. Despite solid aggregate performance, ML models are unreliable in a variety of real-world scenarios, such as on different subgroups [@oakden2020hidden; @suresh2018learning; @goel2020model; @cabreradiscovery; @vanbreugel2023; @seedat2023check], or encountering data deviating from its training distribution [@pianykh2020continuous; @koh2021wilds; @patel2008investigating; @goetz2024generalization], often resulting in significant financial or societal consequences. These failures point to deficiencies in the way we test such ML models. To understand this in greater detail, let's address two questions: how are we currently testing ML models and can we do better? + +**How are we testing?** The predominant *de facto* paradigm of testing ML models relies on using only held-out data to evaluate the model either on average or by assessing performance on different subgroups within the dataset. Such *data-only methods* optimize a given objective function to find subgroups (or slices of data) within the dataset where a trained model underperforms relative to aggregate performance. We refer to this relative underperformance as a *model failure*. However, data-only methods operate under the restrictive assumption that the available empirical data is the sole input for testing ML model performance. In practice, this is almost always violated. It is common to have *a priori* knowledge of where models are likely to fail, given the problem context. + +The restrictive assumption of data-only methods comes at a significant cost. Specifically, data-only methods, which have been the go-to approach for ML testing, are required to iterate and test the performance across a large number of possible subgroups. Each subgroup test is equivalent to evaluating a separate hypothesis about the model's performance on that specific slice of data. This raises a subtle but important problem---from the perspective of hypothesis testing, such methods *implicitly test multiple hypotheses* (Sec. [3.1](#sec:multiple_testing){reference-type="ref" reference="sec:multiple_testing"}). The more subgroups we test, the more hypotheses we implicitly evaluate. Consequently, this results in critical problems associated with multiple testing: (i) ***high false positive*** and (ii) ***high false negative rates*** (Sec. [3.2](#sec:failure-data-only){reference-type="ref" reference="sec:failure-data-only"}). A third complementary challenge is that each data slice tested is (iii) ***not necessarily practically meaningful***. The practical implications of these drawbacks are quite concerning---they hinder our ability to accurately identify where the model performs poorly, thereby undermining the reliability of our model evaluations. + +**Can we do better?** To address these three challenges, we propose to *loosen the restrictive assumption of reliance only on data* and propose a new paradigm of testing called Context-Aware Testing (CAT) to offer an alternative to the dominant data-only view ([3.3](#sec:context-aware-testing){reference-type="ref+label" reference="sec:context-aware-testing"}). CAT provides a principled approach to incorporate external knowledge---or context--- to the ML testing process. With the view that each evaluated data slice corresponds to an implicit hypothesis test, we propose that ML evaluation on observational datasets can be achieved via a *context-guided sampling mechanism* (Definition [1](#def:cat){reference-type="ref" reference="def:cat"}). This mechanism is a sampling procedure that uses context as an inductive bias to prioritize specific data slices to test for which have a higher chance of surfacing meaningful model failures. Therefore, CAT fundamentally helps to answer the question of "what should we test for?". + +Let's consider an example of building an ML model to predict prostate cancer [@seer]. Data-only methods employ a search procedure over the dataset to find divergence across a large number of possible feature combinations which may lead to (i) ***high false positive rates*** by identifying spurious underperforming subgroups (e.g. based on eye color or patient ID); (ii) ***high false negative rates*** by failing to identify true underperforming subgroups due to the large number of combinations tested and applied testing corrections; and (iii) ***testing subgroups which are not practically meaningful*** (e.g. interaction between eye color and height). In contrast, a CAT-based approach would define and target task-relevant subgroups, limiting the number of tests conducted with better false positive control and greater statistical power. As we empirically show in [5](#sec: experiments){reference-type="ref+label" reference="sec: experiments"}, obtaining many false positives and false negatives is overwhelmingly common in current testing practices. + +In bringing CAT to reality, we develop the framework called **SMART [^2] Testing**, which performs automated ML model evaluation by actively identifying potential failure cases ([4](#sec:SMART){reference-type="ref+label" reference="sec:SMART"}). SMART uses large language models (LLMs) to generate contextually relevant failure hypotheses to test. We further introduce a *self-falsification mechanism*, to automatically validate the generated failure hypotheses using the available data, allowing efficient pruning of spurious hypotheses. Finally, SMART generates comprehensive model reports that provide insights into the identified failure modes, their impact, and potential root causes, enabling stakeholders to make informed decisions. + +::: mdframed +**Contributions.** **[1⃝]{style="color: ForestGreen"}** We identify critical gaps in predominant data-only ML testing, illustrating they miss important dimensions ([3](#sec:context-aware){reference-type="ref+label" reference="sec:context-aware"}). **[2⃝]{style="color: ForestGreen"}** We formalize the *Context-Aware Testing* paradigm, providing a principled framework to incorporate context in addition to data into the testing process, which is then used to guide the generation of targeted tests ([3](#sec:context-aware){reference-type="ref+label" reference="sec:context-aware"}). **[3⃝]{style="color: ForestGreen"}** We build the first context-aware testing system, *SMART* Testing, which employs LLMs to hypothesize likely and relevant model failures and empirically refutes them with data using a novel *self-falsification mechanism* ([4](#sec:SMART){reference-type="ref+label" reference="sec:SMART"}). **[4⃝]{style="color: ForestGreen"}** We demonstrate the value of context for effective testing, challenging the de facto data-only paradigm by showing how SMART identifies impactful model failures while avoiding false positives across diverse settings, when compared to data-only testing. Additionally, SMART identifies failures on important societal groups and generates comprehensive model reports ([5](#sec: experiments){reference-type="ref+label" reference="sec: experiments"}). +::: + +# Method + +The prevailing paradigm for testing ML models relies on *data-only* methods which exclusively use data to surface model failures. In this section, we explore the limitations of data-only methods by viewing ML testing as a multiple hypothesis testing problem. We explore why data-only methods are uniquely prone to finding *false positive* and *false negative* model failures. To address this, we introduce a new paradigm of testing called **context-aware testing** which relies on external knowledge, or *context*, as an inductive bias to better identify where models fail. + +**Preliminaries.** Denote the feature space by $\mathcal{X}$ and the label space by $\mathcal{Y}$, and $\mathcal{P}$ the joint probability distribution over $\mathcal{X} \times \mathcal{Y}$. We wish to test a fixed, trained black-box model $f: \mathcal{X} \rightarrow \mathcal{Y}$, using a finite dataset $\mathcal{D}\subset \mathcal{X} \times \mathcal{Y}$ usually split into $\mathcal{D}_{train} = \{ (x_i, y_i) \}_{i=1}^{N_{train}}$ and $\mathcal{D}_{test} = \{ (x_i, y_i) \}_{i=1}^{N_{test}}$. We assume the existence of a loss function $\ell: \mathcal{Y} \times \mathcal{Y} \rightarrow \mathbb{R}$ which measures the discrepancy between the model's prediction and the true labels point-wise. + +The primary goal of ML testing is to identify *meaningful failure modes*---subgroups (data slices) of the data distribution where the model's performance is significantly worse than its average behavior. Formally, let $\mathcal{S} \subseteq \mathcal{X} \times \mathcal{Y}$ denote a *data slice* and let $\mathcal{P}_\mathcal{S} = \mathcal{P}(\cdot|(x,y) \in \mathcal{S})$ be the conditional distribution induced by $\mathcal{S}$. We aim to identify slices where the slice-specific expected loss $\mu_\mathcal{S} = \mathbb{E}_{(x,y) \sim \mathcal{P}_\mathcal{S}}[\ell(f(x), y)]$ significantly exceeds the population-level expected loss $\mu_\mathcal{P} = \mathbb{E}_{(x,y) \sim \mathcal{P}}[\ell(f(x), y)]$[^3]. + +**Testing ML models is a multiple hypothesis testing problem**. We are interested in identifying failure modes that generalize beyond the training dataset. We can interpret the empirical dataset $\mathcal{D}$ as a sample from a broader distribution $\mathcal{P}$, and our goal is to make an *inferential* claim on the performance on the data slices with respect to $\mathcal{P}$. Suppose we have a candidate data slice $\hat{\mathcal{S}} \subseteq \mathcal{D}$ . We can evaluate the empirical slice-specific loss as $\hat{\mu}_\mathcal{S} = |\hat{\mathcal{S}}|^{-1} \sum_{(x,y) \in \hat{\mathcal{S}}} \ell(f(x), y)$ and compare it to the empirical loss over the entire dataset $\hat{\mu}_\mathcal{D} = |\mathcal{D}|^{-1} \sum_{(x,y) \in \mathcal{D}} \ell(f(x), y)$. To make an inferential claim about the model's performance on the data slice $\mathcal{S}$ w.r.t. $\mathcal{P}$, we can follow the frequentist testing paradigm and formulate a hypothesis test where we evaluate whether the performance is significantly different. Therefore, $H_0: \mu_\mathcal{S} = \mu_\mathcal{D}$ and alternative hypothesis $H_1: \mu_\mathcal{S} \neq \mu_\mathcal{D}$, where $\mu_{\mathcal{S}} = \mathbb{E}_{(x,y) \sim \mathcal{P}_\mathcal{S}}[\ell(f(x), y)]$ and $\mu_\mathcal{D} = \mathbb{E}_{(x,y) \sim \mathcal{P}}[\ell(f(x), y)]$ denote the *true* slice-specific and population-level losses. In practice, this evaluation could be done by running an appropriate frequentist statistical test and evaluating whether $p < \alpha$ for each slice, given some pre-defined $\alpha$. + +However, in realistic testing scenarios, we evaluate the model's performance not just on a single slice but on a large collection of candidates $\{\mathcal{S}_j\}_{j=1}^{m}$. This amounts to conducting many simultaneous hypothesis tests. Accounting for multiple testing is important. A naive testing procedure that does not adjust for multiplicity could surface a large number of spurious failure modes simply by chance (Type I error). Conversely, controlling the false discovery rate [@benjamini1995controlling] may involve adjusting the per-test significance threshold to $\alpha' \ll \alpha$, potentially sacrificing power to detect true failures (Type II error). + +::: mdframed +  The multiple hypothesis testing viewpoint reveals a key challenge in ML model evaluation: To reliably surface meaningful failures, we require a principled procedure for generating a relatively small number of promising hypotheses (candidate slices) to test. +::: + +Existing ML testing methodologies are *data-only* in that they only use the available empirical data to test for model failures and vary in their optimization objective [@oshingbesan2022beyond; @chung2019slice; @sagadeeva2021sliceline; @lemmerich2018pysubgroup; @pastor2021looking]. However, data-only methods operate under the restrictive assumption that the available empirical data is the sole input for testing ML model performance. In practice, this is almost never the case. It is common to have an *a priori* understanding of where models are likely to fail given the data distribution, model class, training algorithm, and deployment context. This restrictive assumption results in three challenges: $\blacktriangleright$ **(i) High false positive rate**: data-only methods search over a large space of data slices and each evaluation amounts to an implicit hypothesis test (Sec. [3.1](#sec:multiple_testing){reference-type="ref" reference="sec:multiple_testing"}). Therefore, the probability of observing a false failure increases with every test performed. $\blacktriangleright$ **(ii) High false negative rate**: The naive testing procedure can be made robust by correcting for the number of tests performed which reduces the statistical power to detect true failures. $\blacktriangleright$ **(iii) Lack of meaningful failures**: Data-only methods are fundamentally limited by the fact that not all statistically significant slices are practically meaningful. We empirically validate these claims in Sec. [5](#sec: experiments){reference-type="ref" reference="sec: experiments"}. + +::: mdframed +  The core limitation of data-only testing is the lack of a principled failure mode *discovery* procedure that can incorporate prior knowledge to guide the search for meaningful errors. +::: + +To address the limitations of data-only testing, we introduce **context-aware testing** (CAT), a principled framework for identifying meaningful model failures using context. This could be the context implicitly encoded in the dataset (i.e. via meaningful feature names) or available external input (i.e. external contextual knowledge, such as a string of input information from a human). Let $\mathcal{C}$ denote the space of all possible contextual information and $c \in \mathcal{C}$ be specific external input. Our core insight is that we can use $\mathcal{C}$ as an inductive bias to select which slices to test for. + +::: {#def:cat .definition} +**Definition 1** (Context-Aware Testing). *Let $\mathcal{X}$, $\mathcal{Y}$, $\mathcal{P}$, $f$, $\mathcal{D}=\{(x_i,y_i)\}_{i=1}^N$, and $\ell$ be defined as in the standard supervised learning setup. Let $\mathcal{C}$ be a space of contexts.* + +*Context-aware testing is defined by two procedures:* + +1. *A context-guided slice sampling mechanism $\pi: \mathcal{C} \times (\mathcal{X} \times \mathcal{Y}) \times \mathbb{N} \to 2^{\mathcal{X} \times \mathcal{Y}}$ such that $\pi(c, \mathcal{D}, m) = \{\mathcal{S}_1, \ldots, \mathcal{S}_m\}$, where $c$ is used as an inductive bias for function $\pi$ to prioritize slices likely to contain meaningful failures, and $m \in \mathbb{N}$ are the number of slices to evaluate.* + +2. *A multiple hypothesis testing procedure: $\forall\mathcal{S}_i\in\pi(c,\mathcal{D}, m)$, test $H_0:\mu_{\mathcal{S}_i}=\mu_\mathcal{D}$ vs. $H_1:\mu_{\mathcal{S}_i}\neq\mu_\mathcal{D}$, where $\mu_{\mathcal{S}_i}=\mathbb{E}_{(x,y)\sim\mathcal{P}_{\mathcal{S}_i}}[\ell(f(x),y)]$, $\mu_\mathcal{D}=\mathbb{E}_{(x,y)\sim\mathcal{P}}[\ell(f(x),y)]$* + +*A meaningful failure is characterized by statistical significance and practical relevance.* +::: + +The targeted sampling mechanism $\pi$ uses the context $\mathcal{C}$ as an inductive bias to prioritize testing of slices that are (i) *more relevant* to the deployment context, and (ii) *more likely* to exhibit significant performance gaps. + +This principled slice selection offers several key advantages over data-only methods: $\blacktriangleright$ *(i) Improved false positive control*: by limiting the number of tests conducted to $m$, CAT controls the risk of spurious discoveries that arise when naively testing all possible slices. $\blacktriangleright$ *(ii) Improved true positive rate*: The targeted selection of slices likely to contain failures maintains test power by avoiding the need to aggressively correct for multiple testing. $\blacktriangleright$ *(iii) Meaningful failures*: context-guided sampling identifies failures that are both statistically significant and practically relevant. + +::: mdframed +  Context-aware testing overcomes the limitations of data-only methods by employing a principled, context-guided slice sampling mechanism $\pi$ to prioritize the discovery of meaningful model failures. +::: + +The core technical challenge in realizing CAT is the development of an effective context-conditional sampling algorithm $\pi$. In the following section, we propose a concrete instantiation of $\pi$ using a large language model to generate plausible failure modes and guide testing of ML models. + +
+ +
+
Overview of the SMART Testing Framework showing the four steps. All steps are automatically executed by an LLM.
+
+ +We now instantiate the CAT framework outlined in Sec. [3.3](#sec:context-aware-testing){reference-type="ref" reference="sec:context-aware-testing"} with a method called **SMART** testing (**s**ystematic, **m**odular, **a**utomated, **r**equirements-responsive, and **t**ransferable). SMART generates relevant and likely hypotheses about potential model failures and empirically evaluates these hypotheses on available data. SMART follows a four-step procedure: (1) Hypothesis generation; (2) Operationalization; (3) Self-falsification; (4) Reporting. Table [\[tab:hypotheses\]](#tab:hypotheses){reference-type="ref" reference="tab:hypotheses"} provides an early illustration of what it practically means to generate hypotheses and test them with SMART. The procedure below details how the four steps are implemented. + +::: table* +[]{#tab:hypotheses label="tab:hypotheses"} +::: + +**Step 1: Hypothesis Generation.** Recall from [3.3](#sec:context-aware-testing){reference-type="ref+label" reference="sec:context-aware-testing"}, we wish to define a sampling mechanism $\pi$ to sample slices $\mathcal{S}$ which are both relevant and have a high relative likelihood of failure --- where $\pi$ should be both contextually-aware and able to integrate requirements to guide sampling. + +::: wrapfigure +r0.42 ![image](figures/t6_updated2.pdf){width="97.5%"} + +------------------------------------------------------------------------ + +[]{#fig:sampling label="fig:sampling"} +::: + +We posit that LLMs have the potential to satisfy these properties due to the following capabilities: $\blacktriangleright$ *Contextual understanding*: LLMs have been pretrained with a vast corpus of information and hence have extensive prior knowledge around different contexts and settings [@chowdhery2022palm; @singhal2023large; @cllm2024; @astorga2024]. $\blacktriangleright$ *Integrate requirements*: LLMs are adept at integrating requirements or additional information about the problem via natural language [@yang2023large; @cllm2024]. $\blacktriangleright$ *Hypothesis proposers*: In proposing likely failure modes, LLMs have also been shown to be "phenomenonal hypothesis proposers" [@qiu2023phenomenal]. + +An LLM is defined as a probabilistic mapping $l : \Sigma^* \rightarrow P(\Sigma)$, where $\Sigma$ denotes the vocabulary space, and $P(\Sigma)$ represents the probability distributions over $\Sigma$. The model processes input sequences $s \in \Sigma^*$, each a concatenation of tokens representing external (contextual) input $\mathcal{C}_e$ (which can be null) and dataset contextualization $\mathcal{C}_D$ is formalized as $s = (\mathcal{C}_e, \mathcal{C}_D)$. We extract the contextualized description $\mathcal{C}_D$ from the dataset $\mathcal{D}$ using an extractor function $\mathcal{E}: \mathcal{D} \rightarrow \mathcal{C}_D$, which captures essential dataset characteristics (e.g. feature relationships and high-level dataset information). Additionally, we highlight the the LLM will implicitly extract context based on the context encoded in the dataset (e.g. via meaningful feature names). Based on input $s$, the LLM predicts a distribution over $\Sigma$ from which hypotheses of model failure and corresponding justifications are sampled. + +As depicted in Fig. [\[fig:sampling\]](#fig:sampling){reference-type="ref" reference="fig:sampling"}, given $\mathcal{C}_e$ and $\mathcal{C}_D$, we sample the $N$ most likely hypotheses of failures, $\mathcal{H} = \{H_1, H_2, \ldots, H_N\}$, and corresponding justifications $\mathcal{J} = \{J_1, J_2, \ldots, J_N\}$, to provide explainability. This process is formalized using the LLM's mapping $l$ as follows: $$(H_i, J_i) \sim l(s), \text{where} \; s = (\mathcal{C}_e, \mathcal{C}_D), \; \forall i \in \{1, 2, \ldots, N\}.$$ + +**Step 2: Operationalization.** The process of operationalizing each hypothesis $H_i \in \mathcal{H}$ involves translating its natural language expression into a form that can directly operate on the training dataset $\mathcal{D}_{train}$ (an example is provided in Table [\[tab:hypotheses\]](#tab:hypotheses){reference-type="ref" reference="tab:hypotheses"}). To achieve this, we define an interpreter function $I: \mathcal{H} \to \{0, 1\}^\mathcal{X}$ that maps each natural language hypothesis $H_i$ to a corresponding binary-valued function $g_i: \mathcal{X} \to \{0, 1\}$ on the feature space $\mathcal{X}$, where $g_i(x) = 1$ if $x$ satisfies the criteria and $g_i(x) = 0$ otherwise. Each function $g_i$ induces a data slice $\mathcal{S}_i \subseteq \mathcal{D}_\text{train}$ consisting of data points that satisfy the criteria of hypothesis $H_i$, such that $\mathcal{S}_i = \{(x, y) \in \mathcal{D}_\text{train}: g_i(x) = 1\}$. Therefore, each hypothesis, after operationalization, corresponds to a specific slice that is being tested on. Steps 1 and 2 serve to practically instantiate $\pi$ from [3.3](#sec:context-aware-testing){reference-type="ref+label" reference="sec:context-aware-testing"}. + +::: wrapfigure +r0.45 ![image](figures/self-falsification.pdf){width="100%"} + +------------------------------------------------------------------------ + +[]{#fig:falsification label="fig:falsification"} +::: + +**Step 3: Self-falsification** []{#sec:self-falsify label="sec:self-falsify"} We introduce a novel self-falsification mechanism to empirically evaluate (or *refute*) the generated hypotheses[^4]. Specifically, for each feasible hypothesis $H_i \in \mathcal{H}$, we attempt to falsify the hypothesis with observed empirical data [^5]. This involves evaluating the model $f$ over the data slice $\mathcal{S}_{i}$ operationalized from $H_i$. We then assess whether the slice performance on $f$ has a significant deviation from the model's overall performance. For instance, in Table [\[tab:hypotheses\]](#tab:hypotheses){reference-type="ref" reference="tab:hypotheses"}, this is done by computing $|\Delta \text{Acc}|$ and the p-value. The significance of this deviation is determined through frequentist statistical testing, i.e. when $p < \alpha$ for any $\alpha$ which might also be adjusted for multiple hypothesis testing. This step effectively "reshuffles/reranks" the hypotheses based on their likelihood on the observed data. For example, when benchmarking we select the top $n$ hypotheses based on statistical significance: $\mathcal{H}^* = \underset{h_{i} \in \mathcal{H}_{\mathcal{T}}}{\text{arg min}_n} \; \{p_i < \alpha\}$. + +***Remark*:** As shown in Fig. [1](#fig:smart-test){reference-type="ref" reference="fig:smart-test"}, we can exclude the self-falsification from SMART in cases of small-sample sizes. We denote this ablation of SMART as $SMART_{NSF}$. + +**Step 4. Model Performance Evaluation and Reporting.** Finally, SMART automatically generates a report of the overall performance of the model under varying conditions generated by the LLM, including a summary report, a complete summary of the tests carried out, intermediate and final results, and potential failure modes of the ML model. + +::: mdframed +  SMART is a tabular CAT method which (i) directly *searches for model failures*, sampling targeted tests (Sec. [3.3](#sec:context-aware-testing){reference-type="ref" reference="sec:context-aware-testing"}) and (ii) *incorporates data, and context* into ML testing. +::: + +**Practical use of SMART.** + +We highlight the practical use of SMART testing and emphasize SMART's ease of use and minimal input requirements needed. In particular, as shown in the example below, users do not need any prior knowledge to use SMART. Rather, we make use of the context inherently encoded in the dataset, feature names and task. + +``` {.python language="Python" caption="Use of SMART Testing" basicstyle="\\footnotesize\\ttfamily"} +import SMART + +model_tester = SMART('gpt-4') + +context = "Find where a model fails for the cancer prediction task." + +model = XGBoost() +dataset_description = X.describe() + +model_tester.fit(X, context, dataset_description, model) +``` + +We now quantitatively evaluate SMART Testing [^6] and demonstrate the viability of this new ML testing paradigm (i.e. CAT), in contrast to data-only testing. Table [1](#table:summary){reference-type="ref" reference="table:summary"} summarizes our experiments and takeaways. + +**Baselines.** We compare SMART with a variety of data-only testing baselines namely: Autostrat [@oshingbesan2022beyond], Slicefinder [@chung2019slice], Sliceline [@sagadeeva2021sliceline], Pysubgroup [@lemmerich2018pysubgroup] and Divexplorer [@pastor2021looking]. We also evaluate $SMART_{NSF}$ (i.e. no self-falsification) as an ablation. Given space limits, we exemplify the LLM in SMART by GPT-4 [@openai2023gpt4] and our ML model to audit is logistic regression. We investigate other LLMs in the Appendix [4.3](#appendix:effects_of_LLM){reference-type="ref" reference="appendix:effects_of_LLM"} and other tabular models in Appendix [4.4](#appendix:tab_model_effects){reference-type="ref" reference="appendix:tab_model_effects"}, observing similar results. We provide additional experimental details in Appendix [3](#sec:experiment_details){reference-type="ref" reference="sec:experiment_details"}. + +:::: adjustbox +max width=0.99 + +::: {#table:summary} + **Sec.** **Experiment** **Takeaway** + -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ------------------------------------------------------------- -------------------------------------------------------------------------------------------------- + Figure [2](#fig:irrelevant_features){reference-type="ref" reference="fig:irrelevant_features"} Assess robustness to irrelevant features SMART consistently avoids irrelevant features, outperforming data-only methods. + Table [\[tab:ml_models\]](#tab:ml_models){reference-type="ref" reference="tab:ml_models"} Assess ability to identify significant model failures SMART discovers slices with larger performance discrepancies across models. + Table [\[tab:fnr_comparison\]](#tab:fnr_comparison){reference-type="ref" reference="tab:fnr_comparison"} Measure FNR in identifying underperforming subgroups SMART achieves the lowest false negative rates in all settings. + Table [\[tab:bias\]](#tab:bias){reference-type="ref" reference="tab:bias"} Assess robustness to potential LLM biases SMART effectively mitigates biases in identifying underperforming subgroups. + Table [\[tab:requirements\]](#tab:requirements){reference-type="ref" reference="tab:requirements"} Evaluate ability to satisfy testing requirements SMART satisfies most requirements while maintaining statistical significance. + Figure [6](#fig:sample_size){reference-type="ref" reference="fig:sample_size"} Assess how sample size affects irrelevant feature detection SMART consistently avoids irrelevant features regardless of sample size. + Table [\[tab:data-shift-main\]](#tab:data-shift-main){reference-type="ref" reference="tab:data-shift-main"} Evaluate performance in different deployment environments SMART identifies more significant failure slices in new environments. + Figure [7](#fig:sample_size_exp){reference-type="ref" reference="fig:sample_size_exp"} Assess impact of sample size on performance SMART consistently outperforms data-only methods across all sample sizes. + Table [\[tab:data-shift-divergent\]](#tab:data-shift-divergent){reference-type="ref" reference="tab:data-shift-divergent"} Evaluate tendency to flag non-existent failures SMART avoids spurious failures, unlike data-only methods. + App. [4.3](#appendix:effects_of_LLM){reference-type="ref" reference="appendix:effects_of_LLM"} Compare performance of GPT-3.5 and GPT-4 in SMART SMART using both GPT versions outperform benchmark methods. + Table [3](#tab:data-shift-models){reference-type="ref" reference="tab:data-shift-models"}, [\[table:dl-models\]](#table:dl-models){reference-type="ref" reference="table:dl-models"} Evaluate SMART across different tabular ML models SMART identifies larger performance discrepancies across various models. + Table [\[table:llm_cost\]](#table:llm_cost){reference-type="ref" reference="table:llm_cost"} Evaluate SMART the cost of using SMART SMART is cost efficient to generate hypothesis less than 0.5 USD. + Table [\[table:overlap\]](#table:overlap){reference-type="ref" reference="table:overlap"} Compare the overlap of hypotheses w/ open-weight models SMART using open-weight models has a significant overlap of hypotheses. + Table [\[tab:hidden_names\]](#tab:hidden_names){reference-type="ref" reference="tab:hidden_names"} Evaluate SMART' SMART identifies larger performance discrepancies across various models. + App. [4.10](#sec:model_report){reference-type="ref" reference="sec:model_report"} Showcase SMART's practical output SMART generates comprehensive model reports, providing clear justifications for each hypothesis. + + + : Summary of experiments and takeaways. +::: +:::: + +[]{#table:summary label="table:summary"} + +**Goal.** We aim to underscore the role of contextual awareness in preventing false positives in model testing. In particular, we consider the scenario when dealing with tabular data that may contain many irrelevant or uninformative features [@karim2023interpreting], persisting even post-feature selection [@vargas2013contingency; @grinsztajn2022tree]. We contrast SMART which explicitly accounts for context, to data-only approaches which are context-unaware and can only operate on the numerical data. + +**Setup.** We fit a predictive model to the training dataset, varying the number of irrelevant, synthetically generated features contained in the dataset. The irrelevant features are drawn from different distributions. We then quantify the proportion of conditions in the identified slices that *falsely* include the irrelevant synthetic features. We evaluate using five real-world tabular datasets spanning diverse domains, namely finance, healthcare, criminal justice and education: loan, breast cancer, diabetes, COMPAS recidivism [@compas] and OULAD eduction [@kuzilek2017open]. These datasets have varying characteristics, from sample size to number of features and are representative of different contexts pertinent to tabular ML, to demonstrate the effectiveness of SMART across various real-world contexts + +
+ +
+
Contextual-awareness in SMART reduces FPs, i.e. reducing the proportion of irrelevant features in slices. SMART is not sensitive to the number of irrelevant features, unlike data-only methods. ↓ is better
+
+ +**Analysis.** Fig. [2](#fig:irrelevant_features){reference-type="ref" reference="fig:irrelevant_features"} shows the proportions of irrelevant features included in slices for increasing numbers of irrelevant features. Data-only methods are unaware of context and are shown to spuriously include high proportions of irrelevant features in their slices; i.e. *false positives (FPs)*. Additionally, these false discoveries increase for data-only methods as the number of irrelevant features increases. Consequently, the sensitivity to FPs of data-only methods risks that slices identified are neither empirically relevant nor meaningful. In contrast, SMART, by virtue of contexual-awareness when generating hypotheses is *not* sensitive to extraneous and contextually irrelevant features. Importantly, SMART also maintains its robustness to FPs even with an increasing number of irrelevant features. + +***Remark.*** We also assess sensitivity to the number of data samples. Fig. [7](#fig:sample_size_exp){reference-type="ref" reference="fig:sample_size_exp"}, Appendix [4.6](#appendix:context_aware){reference-type="ref" reference="appendix:context_aware"} shows that SMART remains robust to variations irrespective of sample size. In contrast, data-only methods have variable performance with false discoveries sensitive to the sample size. + +**Avoiding non-existent failure slices.** It is important that model testing does not flag failures when there are none. In Appendix [4.5](#sec:inductive_bias_subgroups){reference-type="ref" reference="sec:inductive_bias_subgroups"}, we demonstrate that SMART's contextual awareness means that it is resistant to spurious failure slices that have no underlying relationships. In contrast, data-only approaches implicitly assume the existence of problematic slices, which we empirically show makes them prone to spuriously flagging non-existent failure slices. + +[**Takeaway 1.**]{style="color: ForestGreen"} SMART's contextual awareness ensures testing relevance, thereby reducing false positives across different scenarios, in contrast to data-only methods. + +**Goal.** We assess whether the identified failure slices persist when evaluated on new, unseen data across different tabular models. This evaluates the generalization of the identified model failures across multiple different ML models. + +::: wraptable +r0.55 + +[]{#tab:ml_models label="tab:ml_models"} +::: + +**Setup.** We use the prostate cancer dataset [@cutract] and aim to discover slices indicative of underperformance. Thereafter, we assess generalizability of the identified failures across four different ML models (logistic regression, SVM, XGBoost and MLP). Specifically, we compare the top identified slice from each method and compute the absolute difference between the accuracies of that slice and the remainder of the dataset. We posit that a testing framework should be able to identify slices with high-performance discrepancies on unseen data across multiple ML models. + +**Analysis.** Table [\[tab:ml_models\]](#tab:ml_models){reference-type="ref" reference="tab:ml_models"} shows that SMART slices exhibit the greatest discrepancies in model accuracies on unseen test data across different ML models --- indicating the discovered failure modes are generalizable. We find that SMART surfaces slices with greater performance differences compared to $SMART_{NSF}$ (ablation without self-falsification), highlighting the importance of the introduced self-falsification mechanism. In contrast, data-only methods fail to identify slices where the performance discrepancy is as large as SMART. This limitation can be attributed to the tendency of data-only methods to overfit the training data, leading to high false discovery rates. + +[**Takeaway 2.**]{style="color: ForestGreen"} SMART discovers failure slices where the model substantially underperforms, generalizing to unseen test data across different ML models. In contrast, data-only approaches fail to find slices where the difference in accuracies is as large, highlighting the lack of generalizability and reliability of their findings. + +In our setup, false negatives (FNs) are directly tied to true positives (TPs). i.e. the more true positives we find, the fewer FNs we miss. Across multiple experiments (Fig. [2](#fig:irrelevant_features){reference-type="ref" reference="fig:irrelevant_features"}, Table [\[tab:requirements\]](#tab:requirements){reference-type="ref" reference="tab:requirements"}, Table [\[tab:ml_models\]](#tab:ml_models){reference-type="ref" reference="tab:ml_models"}, Table [\[tab:data-shift\]](#tab:data-shift){reference-type="ref" reference="tab:data-shift"}), we consistently show that SMART identifies TPs at substantially higher rates than data-only. For example, in Table [\[tab:data-shift\]](#tab:data-shift){reference-type="ref" reference="tab:data-shift"}, SMART identifies an average of 9.6 out of 10 subgroups where the ML model significantly underperforms. The fact that data-driven methods discover fewer such subgroups implies that they are missing the ones SMART uncovers. + +**Goal.** That said, we conduct an additional experiment to directly assess the false negative rate, wherein we can control issues (as FNs are naturally unspecified in real data). + +**Setup.** We simulate a dataset to predict recidivism ($Y$) based on five covariates: gender, race, age, income, and education. $\log ( \frac{P(Y_i \mid X_i = j)}{P(Y_i \neq j)}) = \alpha_j - (\delta_1 X_{\text{age}} + \delta_2 C_{\text{income}} + \delta_3 C_{\text{education}}) + \epsilon, \text{ where } \epsilon \sim \mathcal{N}(\mu_0, \sigma_0).$ We then train a predictor function $\hat{f}$ on the data and synthetically introduce underperformance on certain corrupted subgroups. For an individual $i$, if they belong to corrupted subgroup $j$, the prediction $\hat{Y}_i$ is equal to $\hat{f}(X_i)$ with probability $1 - p$, and a random prediction sampled from a Bernoulli distribution with probability $p$. If the individual does not belong to subgroup $j$, the prediction is simply $\hat{f}(X_i)$. Finally, we measure how often each testing method identifies that the model $\hat{f}$ underperforms on a subgroup. + +::: wraptable +r0.55 + +[]{#tab: false-negative label="tab: false-negative"} +::: + +**Analysis.** Table [\[tab: false-negative\]](#tab: false-negative){reference-type="ref" reference="tab: false-negative"} demonstrates results where we synthetically manipulate/corrupt the performance of an ML model on a single subgroup (n=1), two subgroups (n=2), and three subgroups (n=3) out of a total of five. The results show the average of 20 runs, where the corrupted groups are randomly selected within each run. We find that SMART consistently is least susceptible to false negatives across all corrupted variables, when compared to data-only methods which struggle especially once more than one variable is corrupted. This serves to corroborate our earlier results. + +[**Takeaway 3.**]{style="color: ForestGreen"} SMART is less prone to FNs, compared to data-only methods across all settings. + +**Background.** In many settings, we want SMART to be part of a human-in-the-loop model evaluation, particularly to address challenges with LLMs, such as biases or missing dimensions. Let us first discuss how SMART addresses some issues by design and then perform an experimental assessment. $\blacktriangleright$ **Using data to mitigate LLM challenges**: We use data in two ways *(i) Data usage in the generation of hypotheses:* Before generating explicit hypotheses of where the model is likely to fail, we provide the LLM with additional information about the data description and model failures of the training dataset. The hypotheses sampled are therefore reflective of the inductive bias of the LLM as well as being conditioned on the data itself; *(ii) Data usage in falsifying hypotheses:* core to SMART is the self-falsification mechanism where we iteratively generate hypotheses and test them on a validation dataset. Data is therefore used to filter out hypotheses which are not supported by the data. Hence, even if "incorrect" hypotheses about group failures are proposed, this step ensures they are discarded. $\blacktriangleright$ **SMART provides clear and transparent testing**: This is done in two ways: *(i) SMART's model reports:* that document specific failure cases with natural language justifications (see Appendix [4.10](#sec:model_report){reference-type="ref" reference="sec:model_report"} for examples). Automatically generated reports can be a useful tool for humans-in-the-loop experts to audit and validate the testing process, such as evaluating whether tests should be added or removed. For example, a domain expert (e.g. a clinician) could review the report to assess whether the identified failure modes are truly relevant and concerning in that specific context. *(ii) Tests include justifications:* The justifications for the tests allow human users to inspect the model's testing procedures, understand the reasons, and audit for biases or missed dimensions. + +**Goal.** Going beyond the mitigation strategies by design, we also assess SMART's robustness to prior biases. Specifically, we assess common ethnicity related biases of LLMs. + +::: wraptable +r0.6 +::: + +**Setup.** We use the same data generating process as Sec. [5.3](#exp:FN){reference-type="ref" reference="exp:FN"}. + +We train a predictor function $\hat{f}$ and simulate a scenario where we intentionally corrupt a model's predictions for a proportion $\tau$ of a minority subgroup (\"white\" or \"black\" ethnicity). $\tau$ ranges from 0 (no corruption) to 1 (completely random predictions for the subgroup). SMART aims to identify the top underperforming subgroup without bias based on historical patterns. We measure the proportion of times the top subgroup contains the \"white\" ($P_{\text{white}}$) or \"black\" ($P_{\text{black}}$) minority, averaged over 20 runs and 5 seeds, separately corrupting \"white\" and \"black\" ethnicities. + +**Analysis.** We show in Table [\[tab:bias\]](#tab:bias){reference-type="ref" reference="tab:bias"} that SMART is able to identify where models underperform even in scenarios such as ethnicity bias, where LLMs exhibit prior biases from the training dataset. This links to the above discussion that SMART mitigates such biases by design both using the training dataset to guide hypothesis generation and using the self-falsification mechanism to empirically evaluate hypotheses (and discard those that aren't reflective of the data). + +[**Takeaway 4.**]{style="color: ForestGreen"} SMART mitigates potential biases in the LLM, both by using the real data to guide hypothesis generation, as well as using the self-falsification mechanism to filter spurious hypotheses. diff --git a/2411.08923/main_diagram/main_diagram.drawio b/2411.08923/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..80172e53c11403f24d689c841db4c3e1967e985f --- /dev/null +++ b/2411.08923/main_diagram/main_diagram.drawio @@ -0,0 +1,492 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/2411.08923/main_diagram/main_diagram.pdf b/2411.08923/main_diagram/main_diagram.pdf new file mode 100644 index 0000000000000000000000000000000000000000..f13a84aac228a3b3f184d97a872d16af38d281db Binary files /dev/null and b/2411.08923/main_diagram/main_diagram.pdf differ diff --git a/2411.08923/paper_text/intro_method.md b/2411.08923/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..b5a91eec0ce9b89794177b135d91f9c5e41cd408 --- /dev/null +++ b/2411.08923/paper_text/intro_method.md @@ -0,0 +1,145 @@ +# Introduction + +In recent years, Vision-Language Models (VLMs) such as CLIP (Radford et al., 2021) have revolutionized downstream vision-language tasks like classification (Conde & Turgutlu, 2021), object detection (Zhong et al., 2021), segmentation (Xu et al., 2022), and image generation (Saharia et al., 2022). These models are trained on large-scale web data, such as the 400 million text-image pairs used for training CLIP. However, despite their success, VLMs exhibit several vulnerabilities. For instance, adversarial attacks (Goodfellow et al., 2015; Moosavi-Dezfooli et al., 2016) and backdoor vulnerabilities (Bai et al., 2024; Liang et al., 2024a) can lead to erroneous classifications with high confidence. CLIP has also been shown to be vulnerable to typographic attacks, where text within images causes misclassification (Goh et al., 2021). Additionally, biases such as gender and racial bias + +∗Equal contribution. + +1The code is available on [GitHub.](https://github.com/Amirabbas-Afzali/Aligning-Visual-Contrastive-learning-models-via-Preference-Optimization) + +(Ruggeri & Nozza, 2023; Zhang et al., 2022; Bolukbasi et al., 2016; Jialu Wang et al., 2021) can be amplified by these models due to biases in their training datasets. Another dispreferred behavior occurs when models focus on tasks unrelated to the intended objective, as highlighted by (Menon et al., 2022). These challenges underscore the need for alignment methods in large pre-trained models. + +Generative models, including Large Language Models (LLMs), have also been widely adopted for solving complex problems across various domains. Examples in the literature include GPT-4 (Achiam et al., 2023), Mistral (Albert Q. Jiang et al., 2023), and LLaMA (Touvron et al., 2023). These models are trained on massive datasets of unlabeled text, learning general language representations. Some of these LLMs require additional Supervised Fine-Tuning (SFT) to specialize for specific tasks using labeled data. However, even with fine-tuning, LLMs can produce outputs that misalign with human values or safety expectations, such as the unreliability of LLMs in autonomous driving (Chen et al., 2024). To mitigate this, alignment techniques are applied, where models are further trained using a reward model that captures human preferences (Stiennon et al., 2022; Ouyang et al., 2022; Bai et al., 2022). Popular alignment approaches include Reinforcement Learning from Human Feedback (RLHF) (Christiano et al., 2023) and Direct Preference Optimization (DPO) (Rafailov et al., 2024), both of which extend the capabilities of these models beyond what SFT alone can achieve. These alignment techniques have improved the safety and relevance of LLM outputs by making them more aligned with human preferences. + +Alignment methods have been explored in VLMs (Sun et al., 2023; Gallego, 2023; Clark et al., 2023), but to the best of our knowledge, they have not been extensively applied to contrastive learning models like CLIP. In this work, we extend the preference optimization paradigm to nongenerative VLMs. Our approach builds on the contrastive learning framework while simultaneously aiming to preserve the pretrained knowledge—a concept known as continual learning (Garg et al., 2024; Wang et al., 2023). + +In this work, we extend the preference optimization paradigm to contrastive learning models, exemplified on enhancing robustness against typographic attacks and mitigating gender biases in image classification and retrieval. By systematically aligning the model's behavior with human preferences, we aim to preserve its pretrained knowledge while improving its performance in sensitive areas such as fairness and robustness. The main contributions of this paper are as follows: + +- We extend and compare three PO methods, originally developed for aligning generative models, to non-generative contrastive learning models. This way, we improve model alignment, preserving the pretrained model knowledge while optimizing for desired behaviors. +- We propose controlling model behavior by adjusting the singular values of a learnable linear transformation. This allows fine-tuning regarding specific concepts. +- Our experiments demonstrate the benefits of applying preference optimization in contrastive learning tasks, such as training for robustness against typographic attacks and disentangling gender information from the embedding space, mitigating gender bias. + +# Method + +Figure 1 gives an overview of our approach with a joint training objective: a preference-based contrastive optimization using a dataset of preference labels $\mathcal{D}_{pref}$ and regularization using a regularization dataset $\mathcal{D}_{reg}$ . Details of our methodology are given in the following. + +![](_page_3_Figure_1.jpeg) + +Figure 1: Overview of our proposed approach. On the left side, we calculate the preference optimization loss $\mathcal{L}_{\text{pref}}(\pi_{\theta}, \pi_{\text{ref}}; \mathcal{D}_{\text{pref}})$ using the preference dataset, and the output logits $\mathcal{T}(y_w)^T \mathcal{I}(x), \mathcal{T}(y_l)^T \mathcal{I}(x)$ from both models. On the right side, the regulatory loss $\mathcal{L}_{\text{reg}}(\pi_{\theta}, \pi_{\text{ref}}; \mathcal{D}_{\text{reg}})$ is calculated using the regularization dataset. The snowflake icons denote frozen encoders. + +To effectively apply preference optimization to contrastive learning models, we first need to frame the problem within the reinforcement learning paradigm, modeling it as a Markov decision process (MDP). Specifically, we define the contrastive learning task as a one-step MDP, denoted as $\mathcal{M}=(\mathcal{S},\mathcal{A},\rho_0,P,R,\gamma)$ , where $\mathcal{S}$ is the state space, $\mathcal{A}$ is the action space, $\rho_0$ is the initial state distribution, P is the transition function, P is the reward function, and P is the discount factor. Since it is a one-step process, the transition function P and the discount factor P0 are unnecessary, as states do not continue beyond the first step. In this setup, the model operates as a retrieval system: the input (either text or image) represents the state P1 and the most similar text or image is selected as the action P2. Additionally, P3 defines the distribution of initial states, and P3 represents the reward function + +The policy $\pi_{\theta}(a|s)$ is defined analogously to the contrastive learning objective, typically using a softmax function over the similarity score $f_{\theta}(x,y) = \mathcal{I}_{\theta}(x)^T \mathcal{T}_{\theta}(y)/\tau$ , where $\mathcal{I}_{\theta}: \mathcal{X}_{\text{image}} \to \mathbb{R}^d$ and $\mathcal{T}_{\theta}: \mathcal{X}_{\text{text}} \to \mathbb{R}^d$ are the encoders for the image and text, respectively. Here, $\mathcal{X}_{\text{image}}$ and $\mathcal{X}_{\text{text}}$ represent the input distributions of images and texts, and d is the dimensionality of the latent space where both image and text representations are embedded. $\theta$ represents the model parameters, and $\tau$ is the temperature. The action space is chosen from a fixed set of possible outputs $\mathcal{A} = \mathcal{Y} \triangleq \{y_i\}_{i \in \{1, \dots, K\}}$ , where K represents the total number of classes. The components of $\mathcal{M}(\mathcal{S}, \mathcal{A}, \rho_0, R)$ are defined as follows: + +$$\begin{cases} s \triangleq x \\ a \triangleq y \end{cases}, \quad \begin{cases} \rho_0(s) \triangleq p(x) \\ R(s,a) \triangleq r(x,y) \end{cases}, \quad \pi_\theta(a|s) \triangleq \frac{e^{f_\theta(y,x)}}{\sum_{y_i} e^{f_\theta(y_i,x)}}.$$ + +This MDP framework can be applied to different tasks and provides flexibility in designing policies. We specifically utilize methods like DPO and KTO due to their implicit reward modeling, which allows for more effective model alignment with desired behaviors. + +In our framework, the desired response of the model given an input x (e.g., an adversarial image) is considered the preferred output $y_w$ . This represents the correct or human-aligned response that we aim to optimize for. On the other hand, the dispreferred output $y_l$ refers to the model's potentially biased response, or the target of an adversarial attack, or an undesirable outcome (such as bias related + +to gender or other attributes). The model learns to be robust against adversarial attacks or specific biases by comparing these two outputs. + +We use two datasets to facilitate this learning: + +- **Preference Dataset** $\mathcal{D}_{pref}$ : This dataset contains pairs $(x, y_w, y_l)$ , where $y_w$ is the preferred output and $y_l$ is the dispreferred output. The model learns to differentiate between desired and undesired outcomes through this dataset, improving its robustness against adversarial attacks and biases. +- Regularization Dataset $\mathcal{D}_{reg}$ : This dataset consists of clean samples (images or text) without adversarial manipulation. It is used during training to ensure the model retains its accuracy on clean data. + +By combining these datasets, the model is trained to align with a specific behavior or to be robust against adversarial inputs while maintaining performance on clean data and other downstream tasks. + +Building on the proposed formulation, preference optimization can be effectively applied to contrastive learning models by guiding the optimization process toward desired behaviors. Moreover, we can shape the policy to learn specific behaviors or patterns by defining them as *preferred*. Furthermore, in Section 4.3, we highlight a significant advantage of this optimization paradigm as a fine-tuning method, helping retain the pre-trained knowledge of the model while adapting it to new preferences and support it with experiment results. + +Recalling Eq. (4), we first focused on policy ratio, $h_{\pi_{\theta}}(y_w, y_l, x)$ which we can write as: + +$$h_{\pi_{\theta}}(y_w, y_l, x) = \left(\log \pi_{\theta}(y_w|x) - \log \pi_{\theta}(y_l|x)\right) - \left(\log \pi_{\text{ref}}(y_w|x) - \log \pi_{\text{ref}}(y_l|x)\right)$$ +(8) + +Given that the objective of the model $\pi_{\theta}$ is to simultaneously increase the probability of the preferred label $\pi_{\theta}(y_w|x)$ and decrease the probability of the dispreferred label $\pi_{\theta}(y_l|x)$ , the term $h_{\pi_{\theta}}(y_w,y_l,x)$ is positive. Note that $\pi_{\text{ref}}$ in this framework is kept frozen. + +**Lemma 3.1** Under the assumption that the text encoder is frozen, i.e., $\mathcal{T}_{ref} = \mathcal{T}_{\theta} = \mathcal{T}$ , the policy ratio for models using the contrastive learning policy, in the methods such as DPO or IPO can be expressed as:2 + +$$h_{\pi_{\theta}}(y_w, y_l, x) = \frac{1}{\tau} (\mathcal{I}_{\theta}(x) - \mathcal{I}_{ref}(x))^T (\mathcal{T}(y_w) - \mathcal{T}(y_l)). \tag{9}$$ + +Based on Lemma 3.1, the term $h_{\pi_{\theta}}$ depends only on the term $(\mathcal{I}_{\theta}(x) - \mathcal{I}_{\text{ref}}(x))^T (\mathcal{T}(y_w) - \mathcal{T}(y_l))$ . This indicates that the loss objective is designed to adjust $\mathcal{I}_{\theta}(x)$ to align more closely with the preferred text embedding difference, i.e., $\mathcal{T}(y_w) - \mathcal{T}(y_l)$ , while maintaining appropriate proximity to the reference embedding $\mathcal{I}_{\text{ref}}(x)$ . Having this in mind, we now raise the following question: + +First, we need to formulate the gradient of loss correctly in order to analyze it. The following Corollary provides a more insightful way of writing the gradient. + +**Corollary 3.2** Using Lemma 3.1, we can obtain the gradient of the loss as the following.3 + +$$\nabla_{\theta} \mathcal{L}_{pref}(\pi_{\theta}, \pi_{ref}) = -\underbrace{w_{pref}(y_{w}, y_{l}; x)}_{(I)} \cdot \underbrace{\left[\frac{\partial \mathcal{I}_{\theta}}{\partial \theta}\right]^{T} (\mathcal{T}(y_{w}) - \mathcal{T}(y_{l}))}_{(II)}, \tag{10}$$ + +where the gradient weight (I) in DPO is $w_{\text{pref}}(y_w, y_l; x) = \sigma(-\beta h_{\pi_{\theta}}(y_w, y_l, x))$ , and in IPO, $w_{\text{pref}}(y_w, y_l; x) = \left(\frac{1}{2\beta} - h_{\pi_{\theta}}(y_w, y_l, x)\right)$ . Intuitively, the term (II) shows that adjusting the image embedding $\mathcal{I}_{\theta}(x)$ in the direction of $\mathcal{T}(y_w) - \mathcal{T}(y_l)$ , also encourages the model to better align + +&lt;sup>2See Appendix B.1 for the proof. + +&lt;sup>3Further details are provided in Appendix B.3. + +with the preferred outputs. Importantly, if the distribution of $\pi_{\theta}$ deviates significantly from the reference distribution $\pi_{\text{ref}}$ , the absolute value of the *policy ratio* increases. As noted previously, this effect reduces the update rate in the direction of (II), acting as a mechanism to control the model and keep it closer to the reference policy, preventing excessive divergence from the reference distribution. + +In the KTO method, since there is no direct comparison between $y_w$ and $y_l$ , we do not get the same equation as in 9. However, the main idea still applies. The model adjusts the image embedding $\mathcal{I}_{\theta}(x)$ to match human preferences, but it does so through a different optimization approach. + +In all our training methods, we explicitly constrain the model to remain proximal to the reference model on the preference dataset $\mathcal{D}_{pref}$ . However, the preference dataset $\mathcal{D}_{pref}$ is typically small and often fails to capture the entire data manifold. In many cases, it primarily consists of adversarial and out-of-distribution samples. To address this limitation, we introduce additional regularization terms into our training process, ensuring the trained model maintains proximity to the reference model. This motivates the introduction of a regularization term $\mathcal{L}_{reg}$ , designed to maintain proximity between the trained model $\pi_{\theta}$ and the reference model $\pi_{ref}$ as follow: + +$$\mathcal{L}_{\text{reg}}(\pi, \pi_{\text{ref}}; \mathcal{D}_{\text{reg}}) = D_{\text{KL}}(\pi(y|x) \| \pi_{\text{ref}}(y|x)) = \mathbb{E}_{x \sim \mathcal{D}_{\text{reg}}} \mathbb{E}_{y \sim \pi(y|x)} \left[ \log \frac{\pi(y|x)}{\pi_{\text{reg}}(y|x)} \right], \tag{11}$$ + +where $\mathcal{D}_{reg}$ is the set of clean images; for example, in the typographic attack task, we construct $\mathcal{D}_{reg}$ using clean images from the training dataset. By adding this term to the loss function, the model can better maintain its accuracy on true labels, even when facing adversarial challenges, by using the regularization dataset $\mathcal{D}_{reg}$ to ensure robustness. + +A shortcoming of these preference losses also suggests introducing additional regularization terms. From this point on, for the sake of simplicity, we denote $\frac{\pi_{\theta}(y_w|x)}{\pi_{\text{ref}}(y_w|x)} = x_1$ and $\frac{\pi_{\theta}(y_t|x)}{\pi_{\text{ref}}(y_t|x)} = x_2$ . To better understand the effect of regularization, we will build up on top of the following theorem. + +**Theorem 3.3** (Feng et al., 2024) The partial derivatives of Eq. (5) with respect to $x_1$ and $x_2$ are given by: + +$$\left| \frac{\partial \mathcal{L}_{DPO}(x_1; x_2)}{\partial x_1} / \frac{\partial \mathcal{L}_{DPO}(x_1; x_2)}{\partial x_2} \right| = \frac{x_2}{x_1} < 1.$$ + (12) + +We find a similar result in the case of IPO:4 + +**Proposition 3.4** The partial derivatives of Eq. (6) with respect to $x_1$ and $x_2$ are given by: + +$$\left| \frac{\partial \mathcal{L}_{IPO}(x_1; x_2)}{\partial x_1} / \frac{\partial \mathcal{L}_{IPO}(x_1; x_2)}{\partial x_2} \right| = \frac{x_2}{x_1} < 1.$$ + (13) + +In our problem formulation in Section 3.1, we define the true label as $y_w$ and the adversarial label (e.g., typographic label) as $y_l$ . According to Theorem 3.3 and Proposition 3.4, the rate of increase in the probability of the preferred response $y_w$ is lower than the rate of decrease in the probability of the dispreferred response $y_l$ , Therefore, as suggested by (Feng et al., 2024), our model primarily focuses on reducing the likelihood of the dispreferred response. This also motivates us to introduce more regulatory terms to ensure proximity. + +Finally, for training, we propose using a total loss function from the dataset $\mathcal{D} = (\mathcal{D}_{pref}, \mathcal{D}_{reg})$ as follows, where $\mathcal{L}_{reg}$ is the regularization loss and $\mathcal{L}_{pref}$ is the preference optimization objective, computed using one of Eqs. 5, 6, or 7: + +$$\mathcal{L}(\pi_{\theta}, \pi_{\text{ref}}; \mathcal{D}) = \mathcal{L}_{\text{pref}}(\pi_{\theta}, \pi_{\text{ref}}; \mathcal{D}_{\text{pref}}) + \lambda_{\text{reg}} \mathcal{L}_{\text{reg}}(\pi_{\theta}, \pi_{\text{ref}}; \mathcal{D}_{\text{reg}}). \tag{14}$$ + +For example, consider giving images as input to CLIP, where the model chooses the corresponding caption from a fixed caption set. The selection process follows the policy we defined, using the softmax function over the similarity scores between the image and each caption in the set, as shown in Figure 1. The training process using the proposed loss (Eq. (14)) is outlined in Algorithm 1. + +&lt;sup>4See Appendix B.2 for the proof. + +``` +Require: dataset \mathcal{D} = (\mathcal{D}_{pref}, \mathcal{D}_{reg}), Model \pi_{\theta}, Reference model \pi_{ref}, Regularization coef. \lambda_{reg} + 1: for each batch b \in \mathcal{D} do + b_{\text{pref}}, b_{\text{reg}} \leftarrow b + ▶ Get batch of preference / regularization data + Compute \pi_{\Psi}(y|x) \triangleq \operatorname{Softmax}(f_{\Psi}(y,x)) for \Psi \in \{\theta, \operatorname{ref}\} + 3: + ⊳ Compute model and ref. distributions + 4: + l_{\text{pref}} \leftarrow \mathcal{L}_{\text{po}}(\pi_{\theta}, \pi_{\text{ref}}; b_{\text{pref}}) + Description > Compute preference loss using one of Eqs. (5), (6), or (7) + 5: + l_{\text{reg}} \leftarrow \mathcal{L}_{\text{reg}}(\pi_{\theta}, \pi_{\text{ref}}; b_{\text{reg}}) + \triangleright Compute regularization loss as in Eq. (11), zero if b_{reg} = \{\} + l_{\text{tot}} \leftarrow l_{\text{pref}} + \lambda_{\text{reg}} \cdot l_{\text{reg}} + + 7: + Update model \pi_{\theta} by minimizing l_{\text{tot}} + 8: end for +``` + +In many cases, fine-tuning the entire model may be unnecessary and require more interpretation and control. Therefore, we employ a linear head on top of the encoders. This linear projection layer offers a more interpretable approach, providing a plug-and-play framework where the projection can be learned during training and then applied as needed. + +This linear layer also introduces a linear-algebraic perspective to the adaptation process. Suppose the learned transformation is represented by the learnable matrix W. This matrix transforms both vectors $\mathcal{I}(x)$ and $\mathcal{T}(y)$ to $W\mathcal{I}(x)$ and $W\mathcal{T}(y)$ respectively, thereby transforming the similarity function $f(y,x) = \mathcal{I}(x)^T \mathcal{T}(y)/\tau$ into a new function $\tilde{f}(y,x) = \mathcal{I}(x)^T W^T W \mathcal{T}(y)/\tau$ . Utilizing the Singular Value Decomposition (SVD), we can express $W = U \Sigma V^T$ . Therefore, we have: + +$$\tilde{f}(y,x) = \mathcal{I}(x)^T W^T W \mathcal{T}(y) / \tau = (V^T \mathcal{I}(x))^T \Sigma^2 (V^T \mathcal{T}(y)) / \tau.$$ +(15) + +In this formulation, the unitary matrix V selects the important directions, and the diagonal matrix $\Sigma = \mathrm{diag}(\sigma_1,\ldots,\sigma_n)$ determines the degree of attenuation or amplification along each direction $v_i$ . Specifically, each $\sigma_i$ determines whether the corresponding direction $v_i$ is strengthened ( $\sigma_i > 1$ ) or weakened ( $\sigma_i < 1$ ). As detailed in Appendix E, our training process keeps the model close to the reference, as expected. Consequently, each $\sigma_i$ remains relatively close to 1, and the overall matrix $W^TW$ stays close to the identity matrix. + +This observation enables post-training adjustments, allowing us to strengthen or weaken each $v_i$ as needed. One approach to achieve this is by directly tuning each singular value $\sigma_i$ , either through linear interpolation or by applying matrix powers. In our work, we utilize matrix powers to modify the singular values via a transformation scaling parameter $t \in \mathbb{R}$ , implementing the transformation $\Sigma \to \Sigma^t$ as follows: + +$$W_t = U\Sigma^t V^T, (16)$$ + +where $\Sigma^t = \text{diag}(\sigma_1^t, \dots, \sigma_n^t)$ . This approach provides flexibility in adapting the model, enabling fine-grained control over the transformation applied to the encoded representations. + +For large positive or negative values of t, we recommend normalizing the output embeddings, as done in the original CLIP model. This normalization prevents certain directions in the embedding space from becoming excessively large. While normalization still allows for the strengthening or weakening of specific directions, it ensures that the length of the vector remains bounded, instead of becoming arbitrarily large. This approach maintains the intended alignment or misalignment with certain directions without causing instability in the embedding space. diff --git a/2412.00175/main_diagram/main_diagram.drawio b/2412.00175/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..7d934d4a61813c9f819a4a560c4e9e728f1b60d1 --- /dev/null +++ b/2412.00175/main_diagram/main_diagram.drawio @@ -0,0 +1,302 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/2412.00175/main_diagram/main_diagram.pdf b/2412.00175/main_diagram/main_diagram.pdf new file mode 100644 index 0000000000000000000000000000000000000000..e4950d776f7297694d445313871a982029d86974 Binary files /dev/null and b/2412.00175/main_diagram/main_diagram.pdf differ diff --git a/2412.00175/paper_text/intro_method.md b/2412.00175/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..e28c28f3d20b467d454461295c0b082a6bc3febe --- /dev/null +++ b/2412.00175/paper_text/intro_method.md @@ -0,0 +1,87 @@ +# Introduction + +Manipulated videos represent a threat to society as they have the potential of misleading people into believing actors with malicious intents. By spreading misinformation on social media platforms, people may be exposed to scams (e.g., identity theft operations), conspiracy theories and political misinformation. Therefore, deepfake detection methods are essential tools on a global scale. + +The progress in automated deepfake detection is fueled by the datasets developed by the research community. Good quality data is essential for both training and benchmarking the progress of these methods. Over the last years, numerous datasets have been proposed. Among the audio-video deepfake datasets, there have been released datasets that alter both streams [@khalid2022fakeavcelebnovelaudiovideomultimodal; @cai2024avdeepfake1mlargescalellmdrivenaudiovisual] or only one [@wang2020asvspoof2019largescalepublic; @muller2024mlaad; @Rossler_2019_ICCV] ; datasets with full-video [@khalid2022fakeavcelebnovelaudiovideomultimodal] or only local [@cai2024avdeepfake1mlargescalellmdrivenaudiovisual; @cai2023glitch] manipulations. + +However, care has to be taken to ensure a good deepfake detection dataset. Any asymmetry in the preparation of fakes and reals can result in biases that correlate spuriously with the groundtruth label. For example, in the image domain, different preprocessing pipelines [@chai2020eccv] or types of resizing [@rajan2024] between real and fake samples paint an overly optimistic picture. In the audio domain, the very popular ASVSpoof19 dataset [@wang2020asvspoof2019largescalepublic] has been shown to leak information about the labels in the form of silence duration [@muller2021speech] or bitrate information [@borzi2022synthetic]. Since many of the deepfake detection models are high capacity, they can easily use such artifacts as shortcuts for learning, consequently greatly impacting their ability to generalize in a real case scenario. + +
+ +
Audio-visual deepfake detection datasets have a silence bias: fake samples start with a brief moment of silence, which is not the case for real samples. Here we show the first 62.5 ms of the audio waveform for a real and the corresponding fake sample from the AV-Deepfake1M dataset .
+
+ +In this work, we first expose a bias present in two widely-adopted audio-video deepfake detection datasets, FakeAVCeleb [@khalid2022fakeavcelebnovelaudiovideomultimodal] and AV-Deepfake1M [@cai2024avdeepfake1mlargescalellmdrivenaudiovisual]: a short moment of silence present at the beginning of manipulated videos (see Figure [1](#fig:real_vs_fake){reference-type="ref" reference="fig:real_vs_fake"}). Based on this information, a very simple silence classifier can reach a near-perfect performance of over 98% on both FakeAVCeleb and AV-Deepfake1M. Moreover, we observe that the silence distribution is similar across datasets, implying that prior methods may have potentially over estimated the generalization performance based on this shortcut. + +Introducing such biases when creating the datasets is unavoidable and there might be others that are more subtle and harder to reveal. To circumvent this problem, we show that one solution is a shift in learning paradigm, from the more common supervised setup to the unsupervised one [@feng2023self; @reiss2023detectingdeepfakesseeing]. Removing fake samples from training eliminates the focus on the asymmetries induced by spurious artifacts. Moreover, this also has the potential to improve generalization among different manipulation techniques, since supervised detection methods tend to over-rely on generator-specific fingerprints [@marra2019gans; @yu2019attributing]. By limiting to real data, we can also naturally leverage self-supervised representations, which have been shown to improve generalization [@ojha2023towards]. + +To this end, we propose an approach, named AVH-Align, that learns a frame-level audio--video alignment score on top of AV-HuBERT features [@shi2022learning]. Since tampered videos are expected to have greater audio-video desynchronizations, this alignment score can effectively differentiate fake and real videos. We show that this approach is robust with respect to the identified shortcut of leading silence, and also outperforms other methods that do not use the silence bias, while not seeing any fake samples at training time. + +To summarize, our work makes the following contributions: 1. We expose a previously unknown spurious feature in two of the most widely adopted deepfake detection datasets. 2. We analyze the impact of this shortcut on various state-of-the-art models. 3. We show that a way to mitigate such shortcuts is by training on real data only and we introduce a new method in this direction. + +# Method + +We propose a method that aligns AV-HuBERT [@shi2022learning] features on real data. First, we extract audio and visual frame-level features with a pretrained AV-HuBERT model. Then, on top of these features we learn a network to better align them. The alignment network is learnt on real samples by matching each video frame to its corresponding audio frame [@feng2023self]. We call our method AVH-Align(AV-Hubert Aligned) and show its depiction in Figure [4](#fig:method){reference-type="ref" reference="fig:method"}. + +**Self-supervised features.** We use AV-HuBERT to represent both the audio and visual content of a video. AV-HuBERT is a Transformer network trained in a self-supervised way to predict iteratively refined centroids from masked inputs. The features extracted by AV-HuBERT encode audio information as proved by its strong performance on tasks such lip reading or noisy audio-visual speech recognition. We extract the audio and visual representations independently: to extract audio features we mask the visual input, to extract visual features we mask the audio input. For a video we obtain audio features $\mathbf{a}_i$ and visual features $\mathbf{v}_i$ features for each time step $i$. Both representations are 1024 dimensional and have a temporal resolution of 25 frames per second. + +**Alignment network.** To tell how well the audio and visual features match each other, we first independently L2 normalize the feature vectors and then feed them into a network $\Phi$. This is implemented as a multi-layer perceptron (MLP) over the concatenated normalized audio and visual features: $$\begin{align} + \Phi_{ij} = \text{MLP}\left(\left[\mathbf{a}_i; \mathbf{v}_j\right]\right). +\end{align}$$ The MLP has four layers, which progressively reduce the feature dimensionality, with layers mapping from the AV-HuBERT feature size of 1024 to 512, 256, 128, and finally to a single output. Each hidden layer includes Layer Normalization and ReLU activation. + +**Loss function.** To learn the alignment network $\Phi$ we maximize the probability of an audio frame $\mathbf{a}_i$ to match the corresponding video frame $\mathbf{v}_i$; this probability is defined as: $$\begin{align} + p(\mathbf{v}_i | \mathbf{a}_i) = \frac{\exp\Phi_{ii}}{\sum_{k\in\mathcal{N}(i)}\exp\Phi_{ik}}, + \label{eq:loss} +\end{align}$$ where $\mathcal{N}(i)$ represents the temporal neighborhood around the frame $i$; in our case we choose the 30 neighboring frames around frame $i$. We define the loss across the entire video as the average: $$\begin{align} + \mathcal{L} = -\frac{1}{T} \sum_{i=1}^T \log p(\mathbf{v}_i|\mathbf{a}_j). +\end{align}$$ The final loss is similar to the contrastive loss InfoNCE [@oord2018representation] which was also used for deepfake detection [@feng2023self; @oorloff2024avff]. + +**Inference.** Once $\Phi$ is learned, we can estimate the alignment score $\Phi_{ii}$ for each audio-video frame pair in a video. Then we compute an overall alignment score for the entire video by pooling the per-frame scores using log-sum-exp. We estimate the fakeness score as the complement of the alignment probability (Eq. [\[eq:loss\]](#eq:loss){reference-type="ref" reference="eq:loss"}). + +**Supervised variant.** To understand the impact of the silence bias on the standard supervised learning paradigm, we design a supervised variant, AVH-Align/sup, that uses the same features and alignment network as AVH-Align, but a classification loss. In this setup we assume that apart from the real videos, we also have access to fake videos in the training set, with corresponding labels $y$. We obtain a per-video alignment score by pooling the per-frame scores $\Phi_{ii}$ using the log-sum-exp function and optimize the binary cross-entropy (CE) loss: $$\begin{align} + \mathcal{L}_{\text{sup}} = \text{CE}(\log\text{sum}\exp(\Phi_{ii}), y) +\end{align}$$ + +::: table* ++----------------------+----------+--------------+---------------+-----------------------------------------------------------------------------------------------+---+------------------------------------------------------------------------------------------------+ +| | | | | FakeAVCeleb | | AV-Deepfake1M | ++:=====================+:========:+:============:+:=============:+=============================:+==================:+===========================================:+:=:+=============================:+==================:+============================================:+ +| 5-7 (lr)9-11 Method | Modality | Train type | Train data | Trim: []{style="color: red"} | Trim: []{style="color: my_green"} | | Trim: []{style="color: red"} | Trim: []{style="color: my_green"} | ++----------------------+----------+--------------+---------------+------------------------------+-------------------+--------------------------------------------+---+------------------------------+-------------------+---------------------------------------------+ +| AVH-Align/sup | AV | supervised | FakeAVCeleb | 99.5 | 99.4 | [$\cong$ 0.1$\%$]{style="color: my_green"} | | 88.8 | 63.6 | [$\downarrow$ 28.4$\%$]{style="color: red"} | ++----------------------+----------+--------------+---------------+------------------------------+-------------------+--------------------------------------------+---+------------------------------+-------------------+---------------------------------------------+ +| AVH-Align/sup | AV | supervised | AV-Deepfake1M | 75.9 | 72.2 | [$\downarrow$ 4.9$\%$]{style="color: red"} | | 99.9 | 84.2 | [$\downarrow$ 15.7$\%$]{style="color: red"} | ++----------------------+----------+--------------+---------------+------------------------------+-------------------+--------------------------------------------+---+------------------------------+-------------------+---------------------------------------------+ +| AVAD [@feng2023self] | AV | unsupervised | LRS | 66.8 | 67.4 | [$\cong$ 0.9$\%$]{style="color: my_green"} | | 54.2 | 54.3 | [$\cong$ 0.2$\%$]{style="color: my_green"} | ++----------------------+----------+--------------+---------------+------------------------------+-------------------+--------------------------------------------+---+------------------------------+-------------------+---------------------------------------------+ +| AVH-Align | AV | unsupervised | VoxCeleb2 | 96.3 | 96.4 | [$\cong$ 0.1$\%$]{style="color: my_green"} | | 85.9 | 86.2 | [$\cong$ 0.3$\%$]{style="color: my_green"} | ++----------------------+----------+--------------+---------------+------------------------------+-------------------+--------------------------------------------+---+------------------------------+-------------------+---------------------------------------------+ +::: + +
+ + + + + + + + + + + + + + + + + + + + + + + +
imageimageimage
imageimageimage
imageimageimage
imageimageimage
+
Per frame fakeness probabilities for AVH-Align and AVH-Align/sup on AV-Deepfake1M. AVH-Align/sup always marks the first frame—corresponding to the leading silence—as fake, thus confirming that it uses the bias to distinguish between real and fake videos. AVH-Align is not affected by the presence of the leading silence. The fakeness probabilities for AVH-Align can be interpreted as misalignment probabilities, which is why they are higher during or after the manipulated region: the separation and re-composition of the audio and video streams causes misalignment throughout the tampered video.
+
diff --git a/2412.03092/main_diagram/main_diagram.drawio b/2412.03092/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..e1d5121a6a132efaf4065890325e12f554710c3e --- /dev/null +++ b/2412.03092/main_diagram/main_diagram.drawio @@ -0,0 +1,583 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/2412.03092/paper_text/intro_method.md b/2412.03092/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..1e902797f882fce2ef28c8b3fa0d376076a98029 --- /dev/null +++ b/2412.03092/paper_text/intro_method.md @@ -0,0 +1,94 @@ +# Introduction + +In recent years, Large Language Models (LLMs) have dramatically advanced AI's ability to handle complex tasks through natural language processing, enabling LLM-based systems, often referred to as language agents, to interact with external tools and solve problems previously considered out of reach (Brown, 2020; Achiam et al., 2023; Team et al., 2023; Anthropic, 2023; Touvron et al., 2023; Jin et al., 2024; Zheng et al., 2024). However, the development of these language agents still requires significant manual effort to break down tasks and fine-tune prompts, tools, and APIs, limiting scalability and + +\*Contact: + +&lt;sup>†Correspondence to: Haohan Wang: + +![](_page_1_Figure_0.jpeg) + +Figure 1: The illustrative comparison between REVOLVE and first-order optimization methods. First-order methods rely solely on immediate feedback, often leading to stagnation in local optima and limiting further improvement. In contrast, REVOLVE incorporates both immediate feedback and response evolution over time, enabling continuous progress and the ability to escape stagnation. + +adaptability [\(Wei et al.,](#page-12-3) [2022;](#page-12-3) [Lyu et al.,](#page-11-4) [2023;](#page-11-4) [Zhou et al.,](#page-12-4) [2024;](#page-12-4) [Liu et al.,](#page-11-5) [2025\)](#page-11-5). This raises the need for automated, scalable optimization techniques that can enhance language agents efficiently. + +To this end, a number of recent efforts has been made on automatic optimization of language agents. For instance, DSpy [\(Khattab et al.,](#page-11-6) [2024\)](#page-11-6) uses bootstrapping and random search to optimize LLM prompts by exploring a combinatorial space of prompt components. GPTSwarm [\(Zhuge et al.,](#page-12-5) [2024\)](#page-12-5) builds on this by introducing an iterative process to manage DSpy's complexity. Other methods like Agent-Pro [\(Zhang](#page-12-6) [et al.,](#page-12-6) [2024\)](#page-12-6) and AgentOptimizer [\(Zhang et al.\)](#page-12-7) target specific modules, refining prompts, and agent policies. However, these approaches often suffer from local optimization, where improvements in isolated components do not lead to overall system performance gains—similar to early neural network practices [\(Hinton and Salakhutdinov,](#page-11-7) [2006\)](#page-11-7). + +Building on these efforts, more advanced research has introduced gradient descent-inspired techniques for automatic prompt optimization. ProTeGi [\(Pryzant et al.,](#page-11-8) [2023\)](#page-11-8) pioneered the use of textual gradients, where natural language feedback refines prompts. Agent Symbolic Learning (ASL) [\(Zhou](#page-12-4) [et al.,](#page-12-4) [2024\)](#page-12-4) extended this to optimize the entire agent system by treating prompts and tools as learnable parameters. Textgrad [\(Yuksekgonul et al.,](#page-12-8) [2024\)](#page-12-8) further applied textual gradients to instance-level optimization, refining outputs across multiple iterations. + +While effective, these methods all rely on what can be described as first-order optimization. In this context, first-order methods mean they adjust the agent's behavior based on immediate feedback from the current iteration, similar to how numerical first-order gradient descent updates parameters using only the current gradient. This limits their ability to account for how responses evolve across multiple iterations, leading to potential stagnation in suboptimal solutions. As shown in Figure [1](#page-1-0) (a), first-order methods often stagnate in local optima, resulting in repeated or minimally improved LLM responses. This challenge motivates us to explore LLM optimization techniques that consider how responses evolve over time, allowing for more adaptive and refined adjustments that can break free from local optima across iterations. + +In this paper, we introduce an optimization method called **REVOLVE** that builds upon TextGrad [\(Yuk](#page-12-8)[sekgonul et al.,](#page-12-8) [2024\)](#page-12-8) by incorporating a deeper understanding of how **R**esponses **EVOLVE** over time. The optimization process begins with a forward pass, where the system executes a series of tasks, logging inputs, outputs, and any prompt or tool usage. A language-based loss function then evaluates the quality of the generated responses, quantifying how well they align with task objectives. In the backward pass, feedback in the form of natural language critiques is used to adjust system variables. *Relove* improves this standard process by focusing on how response patterns evolve over multiple iterations, enabling the system to make more informed and effective adjustments, ultimately leading to improved performance in handling complex tasks. + +As shown in Figure [1](#page-1-0) (b), we calculate a refined gradient that accounts for changes in responses across multiple iterations, enabling the system to adjust based on both immediate feedback and longterm response patterns. This parallels the concept of second-order optimization in traditional methods, where the Hessian matrix captures how the gradient itself changes. In our case, we model the evolving + +relationship between consecutive prompts and responses, enabling the system to make more informed adjustments. By incorporating this additional layer of information, the system can avoid stagnating in suboptimal patterns, which is a common limitation of methods that rely solely on immediate feedback. + +We test the proposed method on three tasks, Prompt Optimization, Solution Optimization, and Code Optimization. These tasks require handling complex reasoning, refining solutions to scientific questions, and optimizing code under challenging constraints. Experimental results show that Relove consistently improves performance across all tasks, showing its versatility and effectiveness in overcoming the limitations of existing optimization methods. + +Our approach draws inspiration from several key areas of research, particularly automated prompt engineering, agent optimization, and gradient-based learning. Below, we highlight foundational works in these areas and situate our method within this broader context. + +**From Prompt Engineering to Agent Optimization.** Prompt engineering has become a key focus in both academia and industry, leading to several methods aimed at automating the process. Early works [\(Pryzant et al.,](#page-11-9) [2020;](#page-11-9) [Yang et al.,](#page-12-9) [2024\)](#page-12-9) explored the use of structured prompts that enable LLMs to optimize their own inputs. Other approaches [\(Prasad et al.,](#page-11-10) [2022;](#page-11-10) [Guo et al.,](#page-11-11) [2023\)](#page-11-11) use search algorithms, like genetic algorithms, to automatically refine prompts. Building on the success of automated prompt engineering, researchers have extended these concepts to broader agent optimization. Techniques like Agent-Pro [\(Zhang et al.,](#page-12-6) [2024\)](#page-12-6) and AgentOptimizer [\(Zhang et al.\)](#page-12-7) focus on optimizing individual components, such as prompts or tools. However, these methods often treat components in isolation, which can result in local improvements without significantly enhancing the overall system. Searchbased approaches, such as DSpy [\(Khattab et al.,](#page-11-6) [2024\)](#page-11-6) and GPTSwarm [\(Zhuge et al.,](#page-12-5) [2024\)](#page-12-5), take a more comprehensive view by optimizing across the combinatorial space of agent components. Despite their scope, these methods rely heavily on numerical metrics that are often inadequate for real-world tasks like software development or creative writing. Additionally, they struggle to optimize multiple components simultaneously or adapt dynamically to changes in the agent pipeline. + +**Gradient-Based Approaches for Agent Optimization.** Recent advancements have introduced gradient descent-inspired techniques to optimize prompts more effectively. ProTeGi [\(Pryzant et al.,](#page-11-8) [2023\)](#page-11-8) is among the first to use natural language feedback—referred to as textual gradients—to iteratively refine prompts. However, as a first-order optimization method, ProTeGi adjusts based only on immediate feedback from a single iteration, limiting its capacity to handle more complex, multi-step tasks. Agent Symbolic Learning (ASL) [\(Zhou et al.,](#page-12-4) [2024\)](#page-12-4) extended this concept by treating the entire agent system—including prompts, tools, and configurations—as learnable components, much like backpropagation in neural networks. This allows for a more comprehensive optimization but remains dependent on immediate feedback from each iteration. Textgrad [\(Yuksekgonul et al.,](#page-12-8) [2024\)](#page-12-8) further advanced this first-order gradient approach by optimizing LLM responses using natural language feedback. By treating feedback as a gradient, Textgrad refines responses without directly altering the model's parameters. While effective for simpler tasks, Textgrad struggles with deeper, multi-step optimizations, frequently getting stuck in suboptimal states. + +To address these limitations, momentum-based methods [\(Yuksekgonul et al.,](#page-12-8) [2024\)](#page-12-8) have been introduced. These techniques track feedback trends across iterations, adjusting step sizes when feedback becomes repetitive, which helps break stagnation. However, while feedback provides immediate guidance on how inputs should change, it doesn't directly reflect how model responses evolve over time. Feedback may remain constant as responses improve or fluctuate despite stagnation, leading to suboptimal adjustments. In complex tasks, where responses exhibit slow or subtle shifts, this disconnection can make feedback misleading or overly sensitive to temporary fluctuations, causing instability. Therefore, while momentum-based methods provide more variation, they still lack the fine-grained control needed for long-term improvement. + +**REVOLVE: Optimization Through Response Evolution.** REVOLVE enhances traditional optimization methods by focusing on how responses evolve over multiple iterations, enabling more refined adjustments throughout the process. Instead of relying solely on immediate feedback, our approach tracks the evolving relationship between consecutive prompts and their corresponding responses. This parallels second-order methods in traditional optimization, where the the Hessian matrix is used to capture changes in the gradient to guide more precise adjustments. However, rather than directly computing numerical second derivatives, we model these iterative shifts in responses to inform our + +adjustments, giving the system a broader understanding of response dynamics over time. The key advantages of our approach include: + +- **Response Evolution Awareness:** REVOLVE monitors changes across iterations, allowing for more refined and adaptive optimization, unlike first-order methods that rely only on immediate feedback. +- **Avoiding Local Optima:** By tracking iterative changes, REVOLVE prevents models from getting stuck in suboptimal solutions, effectively overcoming a common limitation of first-order methods. +- **Stabilized Optimization:** Unlike momentum methods, which risk overshooting due to large adjustments, REVOLVE applies carefully measured adjustments that ensure smoother and more consistent progress throughout the optimization process. + +# Method + +In this work, we extend the TextGrad approach [\(Yuksekgonul et al.,](#page-12-8) [2024\)](#page-12-8) by tracking the evolution of the LLM responses across iterations, allowing for more effective and precise optimization. + +Our method builds upon the general optimization pipeline used in LLM-based systems [\(Zhou et al.,](#page-12-4) [2024;](#page-12-4) [Yuksekgonul et al.,](#page-12-8) [2024\)](#page-12-8), introducing natural language feedback (textual gradients) to refine system responses over multiple iterations, instead of relying on numerical gradients. + +**Forward Pass.** In the forward pass, the AI system is modeled as a computation graph where each node represents a specific task. Inputs are processed sequentially through the nodes, with each node generating outputs based on prior results. These intermediate outputs are stored in a trajectory, which is later used in the backward pass. + +**Language Loss Computation.** After the forward pass, an evaluator LLM assesses the system's performance by generating textual feedback, which serves as the loss. This feedback reflects how well the system's outputs align with the task objectives and drives the subsequent optimization process. + +**Backward Pass.** In the backward pass, similar to numerical gradients in conventional deep learning, textual gradients are backpropagated through the nodes of the system. These gradients, in the form of natural language instructions, indicate how the system's variables—such as prompts, tools, and decisions—should be adjusted to improve the objective function. Starting from the final node, the system computes the necessary updates for these variables as it moves backward. This process mirrors backpropagation in neural networks, but the adjustments are determined by language feedback rather than numerical values. + +While the above process sets the stage for optimizing the system, the effectiveness of the optimization depends on how well this feedback is utilized. In this context, different gradient-based optimization methods come into play. + +TextGrad [\(Yuksekgonul et al.,](#page-12-8) [2024\)](#page-12-8) computes a first-order gradient based on the language loss provided by the evaluator LLM. The first-order gradient captures the difference in response quality between consecutive iterations. Mathematically, the first-order gradient is expressed as: + + +$$\nabla \mathcal{L}(r(p_t)) = \frac{\tilde{\partial} \mathcal{L}(r(p_t))}{\tilde{\partial} p_t}$$ + (1) + +where we use r(pt) to denote the response when the model is fed with input prompt pt and t to denote the iteration. Also, we use ∂˜ to denote the TextGrad-style derivative of loss function with respective to the input prompt due to its analogous nature to the actual derivative that is typically denoted as ∂. + +We seek to extend the previous method by extending Eq. [1](#page-3-0) to consider the history of previous prompts and their responses. Optimizing based on only the current response can lead to short-term improvements but might results in stagnation, especially in complex tasks where deeper issues arise over time. For example, a LLM might slightly refine responses with each iteration, yet without considering the history of prompts and responses, it risks repeating similar patterns. By factoring in the evolution of responses over multiple iterations, we aim to uncover underlying issues that cause stagnation and enable the system to break free from suboptimal cycles. + +To quantify the differences between previous responses, we need to firstly define a similarity function. We use S r(pt), r(pt−1) to denote the similarity between the responses triggered by prompts pt and pt−1. This function plays a critical role in extending the optimization process to account for the dynamics between successive iterations. + +**REVOLVE.** With this setup, to extend Eq. [1](#page-3-0) to encourage the prompt leading to more gradual and thoughtful evolution of the response over multiple iterations, our new gradient can be expressed as + + +$$REVOLVE\left(\mathcal{L}(r(p_t))\right) = \frac{\tilde{\partial}\mathcal{L}(r(p_t)) + \mathcal{S}(r(p_t), r(p_{t-1}))}{\tilde{\partial}p_t},$$ +(2) + +This new formulation guides the optimization process in a way that not only improves immediate task performance but also promotes long-term, iterative refinement. We name our new method Eq. [2](#page-4-0) REVOLVE. Practically, we rely on a LLM to evaluate the similarity function S r(pt), r(pt−1) . + +**Analogy to Second-order Derivative.** To understand the intuition behind calling our approach an analogy to second-order methods, consider how second-order derivatives (Hessians) in classical optimization capture the rate of change of the gradient. The second-order derivative provides deeper insight into the curvature of the optimization landscape, allowing for more informed adjustments that go beyond the immediate gradient. + +In our context, the similarity function S serves a parallel role by tracking how the system's responses shift from one iteration to the next. We can formalize this by using a generalized norm function (denoted ∥ · ∥) to quantify the differences between two elements (either loss functions or prompts). One way to concretely define the similarity function S r(pt), r(pt−1) is as follows: + +$$S(r(p_t), r(p_{t-1})) = \frac{\|\mathcal{L}(r(p_t)) - \mathcal{L}(r(p_{t-1}))\|}{\|p_t - p_{t-1}\|}.$$ + +This equation mirrors the classical definition of a derivative when the difference between successive prompts ∥pt − pt−1∥ is sufficiently small. Thus, by assuming ∥pt − pt−1∥ to be sufficiently small and instructing the LLM to evaluate S r(pt), r(pt−1) as above, we can have + +$$\mathcal{S}\big(r(p_t), r(p_{t-1})\big) = \frac{\tilde{\partial} \mathcal{L}\big(r(p_t)\big)}{\tilde{\partial} p_t}$$ + +As a result, our REVOLVE in Eq. [2](#page-4-0) can be rewritten as: + +$$REVOLVE\left(\mathcal{L}(r(p_t))\right) = \frac{\tilde{\partial}\mathcal{L}(r(p_t))}{\tilde{\partial}p_t} + \frac{\tilde{\partial}^2\mathcal{L}(r(p_t))}{\tilde{\partial}p_t^2},$$ +(3) + +which is a second-order derivative method. By considering this higher-order information, REVOLVE allows the system to escape from local optima and overcome the limitations of approaches that rely solely on immediate feedback. diff --git a/2412.10121/main_diagram/main_diagram.drawio b/2412.10121/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..5a5a059231fe0cacfd9399ab178262fb0ed9a3f2 --- /dev/null +++ b/2412.10121/main_diagram/main_diagram.drawio @@ -0,0 +1,181 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/2412.10121/main_diagram/main_diagram.pdf b/2412.10121/main_diagram/main_diagram.pdf new file mode 100644 index 0000000000000000000000000000000000000000..0b1ac77762e3df019dfc3dc33e02ba75d4cc9820 Binary files /dev/null and b/2412.10121/main_diagram/main_diagram.pdf differ diff --git a/2412.10121/paper_text/intro_method.md b/2412.10121/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..3137bb9e64a366c5e45b65afce1e73b6d955506e --- /dev/null +++ b/2412.10121/paper_text/intro_method.md @@ -0,0 +1,56 @@ +# Introduction + +
+ +
Impact of training data on zero-shot performance of the current state-of-the-art approach (GLiNER). Each synthetic dataset is characterized by the label overlap (yellow column) and the total number of entity mentions (purple column). While zero-shot performance (red line, macro-averaged F1 across 7 benchmarks) has significantly improved, we note a concerning increase in entity type overlaps between training and testing data.
+
+ +Zero-shot named entity recognition (NER) is the task of recognizing instances of named entities of specific types (such as [Person]{.smallcaps}, [Organization]{.smallcaps}, or [Medicine]{.smallcaps}) without any training examples. Current state-of-the-art models, such as GLiNER [@zaratiana-2023-gliner] and GoLLIE [@sainz-2024-gollie], are initially trained on datasets that contain a large set of different entity types [@aly-etal-2021-leveraging; @ma-etal-2022-label]. This allows the models to identify mentions of previously unseen entity types by leveraging their general language understanding capabilities [@golde-etal-2024-large]. Finally, these models are evaluated on zero-shot benchmarks that were excluded from the training process [@yang-katiyar-2020-simple; @das-etal-2022-container; @yang-etal-2022-see]. + +
+ +
With LLMs now capable of generating datasets that cover thousands of entity types, models trained on different datasets are subject to varying label shifts, making comparisons between them challenging. To address this, we introduce Familiarity, a metric that quantifies and accounts for label shift, enabling more accurate and fair comparisons across models.
+
+ +**Advent of large synthetic training datasets.** Recent research has developed methods that can automatically produce training datasets with over tens of thousands of distinct entity types, using available knowledge bases [@wikidata2014] or large language models [LLMs, @brown-2020-gpt]. Examples include PileNER [@zhou-2023-universalner], NuNER [@bogdanov-2024-nuner], and AskNews [@törnquist2024curatinggroundedsyntheticdata]. This represents a paradigm shift for zero-shot NER, which classically relied on hand-labeled training datasets with a much smaller set of entity types, such as Ontonotes [18 types, @hovy-etal-2006-ontonotes]. + +As [1](#figure:introduction){reference-type="ref+Label" reference="figure:introduction"} shows, the advent of large synthetic training datasets has significantly improved reported zero-shot F1 scores. However, as the figure also shows, there is a concerning increase in the overlap between entity types in synthetic datasets and the evaluation benchmarks (cf. [1](#figure:introduction){reference-type="ref+Label" reference="figure:introduction"}, yellow bars). This means that evaluated models have indeed seen many instances of highly similar (or even the same) entity types during training, raising the question of whether the reported F1 scores overestimate their true zero-shot capabilities. + +**Broader implications.** Naturally, we could strive to ensure a fair zero-shot comparison by proposing training and evaluation splits that have no overlapping entity types at all. However, ensuring no overlap is in fact not trivial since the same or highly similar entity types might have different labels (such as [Corporation]{.smallcaps} and [Organization]{.smallcaps}). But more crucially, using fixed training and evaluation splits would potentially limit process driven by advancements in generating synthetic datasets. + +We rather argue that given the advancements of LLMs and their potential to generate high-quality datasets, accepting custom synthetic training datasets is inevitable. We therefore propose to measure the transfer difficulty between the labels of a training and an evaluation dataset, referred to as *label shift* [@lipton2018detecting; @wu2021online]. + +**Contributions.** With this paper, we identify a critical issue with current zero-shot NER evaluations caused by the growing availability of large-scale synthetic training datasets. To address this issue, we propose [Familiarity]{.smallcaps}, a novel metric that quantifies the similarity between the sets of entity types in training and evaluation data, allowing us to assess the transfer difficulty of an evaluation setup (cf. [2](#figure:illustration){reference-type="ref+Label" reference="figure:illustration"}). We summarize our contributions as follows: + +1. We empirically demonstrate that label overlaps introduce undesirable biases in current zero-shot evaluation setups ([2](#sec:validation_experiment){reference-type="ref+Label" reference="sec:validation_experiment"}). + +2. We propose [Familiarity]{.smallcaps}, a metric that quantifies label shift between training data and evaluation benchmarks, providing insights into transfer difficulty ([3](#sec:metric){reference-type="ref+Label" reference="sec:metric"}). + +3. We conduct a thorough analysis of [Familiarity]{.smallcaps}, showing that it effectively mitigates the evaluation bias and can be used to generate training splits of varying difficulty levels ([4](#sec:experiments){reference-type="ref+Label" reference="sec:experiments"}). + +To enable the research community to efficiently compute [Familiarity]{.smallcaps} and incorporate it into future research, we make all code publicly available as open source[^1]. Further, we publish three benchmark scenarios on the Hugging Face hub[^2] for different levels of transfer difficulty to aid researchers in fine-grained analysis of zero-shot NER. + +As shown in [2](#figure:illustration){reference-type="ref+Label" reference="figure:illustration"}, we hypothesize that label shift between fine-tuning and evaluation datasets affects transfer performance, particularly in zero-shot NER settings. We define this transfer as the process of fine-tuning a model $\Theta$ on a dataset $\mathcal{D}$ with entity types $\mathcal{L}^{\mathcal{D}}$ and subsequently evaluating it on one or more benchmarks $\mathcal{Z}_{1, \dots, n}$, each with its own set of entity types $\mathcal{L}^{\mathcal{Z}_{1, \dots, n}}$, such that $\mathcal{Z} = \cup_{i=1}^{n} \mathcal{Z}_i$ and $\mathcal{L}^{\mathcal{Z}} = \cup_{i=1}^{n} \mathcal{L}^{\mathcal{Z}_i}$. The datasets themselves do not overlap: $\mathcal{Z} \cap \mathcal{D} = \emptyset$. + +However, the entity type sets of the training and evaluation datasets may overlap due to the broad coverage of entity types, particularly in synthetic training datasets: $\mathcal{L}^{\mathcal{Z}} \subseteq \mathcal{L}^{\mathcal{D}}.$ + +We further note that it is possible that $\mathcal{L}^{\mathcal{Z}} \cap \mathcal{L}^{\mathcal{D}} = \emptyset$. However, given that LLMs can generate fine-tuning datasets with thousands of entity types, we observe that in some cases, more than 80% of the evaluation entity types are included in the training dataset (e.g., NuNER, PileNER, and AskNews in [1](#figure:introduction){reference-type="ref+Label" reference="figure:introduction"}). This obviously distorts the genuine zero-shot nature of transfer evaluations, and we hypothesize that the performance for an entity type $\ell$ present in both the evaluation benchmark and the fine-tuning dataset ($\ell \in \mathcal{L}^{\mathcal{Z}} \cap \mathcal{L}^{\mathcal{D}}$) will be higher than for an entity type not present in the fine-tuning data ($\ell \in \mathcal{L}^{\mathcal{Z}} \setminus \mathcal{L}^{\mathcal{D}}$). + +
+ +
Transfer performance is higher on entity types that occur in both evaluation and fine-tuning datasets compared to unseen types. Further, we observe a positive, log-linear correlation between the number of entity mentions for some entity type and its final performance.
+
+ +::: {#table:overview_datasets} + ------------ ------------ ----------- ------------------- + Dataset $\#$ Sent. $\#$ Ent. $\emptyset$ Ment. + Types per Sent. + NERetrieve 3,437.6k 0.5k 2.6 + LitSet 972.6k 242.9k 0.8 + NuNER 971.8k 192.1k 4.5 + PileNER 45.9k 12.6k 20.5 + AskNews 49.4k 12.6k 20.2 + ------------ ------------ ----------- ------------------- + + : Overview of synthetic fine-tuning datasets used in our experiments with their total number of sentences, distinct number of entity types, and average number of entity mentions per sentence. +::: diff --git a/2502.08896/main_diagram/main_diagram.drawio b/2502.08896/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..b0e88081447ecafbbb5073bb60e6685947acac12 --- /dev/null +++ b/2502.08896/main_diagram/main_diagram.drawio @@ -0,0 +1,13 @@ + + + + + + + + + + + + + diff --git a/2502.08896/main_diagram/main_diagram.pdf b/2502.08896/main_diagram/main_diagram.pdf new file mode 100644 index 0000000000000000000000000000000000000000..fac25c8761338cd76779af56f7aa5c8aad32549b Binary files /dev/null and b/2502.08896/main_diagram/main_diagram.pdf differ diff --git a/2502.08896/paper_text/intro_method.md b/2502.08896/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..f10bc59b36124aa47a85c81d4ac48f0930426ddc --- /dev/null +++ b/2502.08896/paper_text/intro_method.md @@ -0,0 +1,139 @@ +# Introduction + +Persuasion techniques play a critical role in shaping societal behaviors and public opinion [@fogg2009behavior; @braca2023developing], which has led to sustained interest across a range of disciplines. Social science research has established detailed taxonomies of persuasion strategies [@shrum2012persuasion; @lukin2017argument], while datasets have been developed to cover various domains, including charitable donations [@wang2019persuasion], argument ranking in debates [@toledo2019automatic], detecting mental manipulation [@wang-etal-2024-mentalmanip; @yang2024enhanced], and understanding advertising strategies [@kumar2023persuasion]. Despite these advances, ambiguities persist in defining persuasion [@pauli2022modelling], and applying persuasion strategies across different contexts remains complex [@bai2021m2p2; @schaefer2023towards; @piskorski2023multilingual]. Additionally, the high cost of manually annotating quality data poses a significant challenge [@lai2022werewolf]. + +The advent of large language models (LLMs) has unlocked new possibilities for enhancing various forms of communication, including online political discourse [@argyle2023leveraging; @bai2023artificial], personalized advertising [@matz2024potential; @meguellati2024good], public health messaging [@lim2023artificial; @espinosa2024use], and opinion shaping on social media [@meier2024llm]. Recent research, such as that by @persugpt, has begun exploring LLM-generated persuasive dialogues. However, their approach is limited to simple, two-party dialogues where a persuader seeks to change the persuadee's viewpoint. These dialogues often lack depth, presenting brief exchanges with simplistic logic and unnatural flow, restricting their usefulness for studying persuasion in more complex settings. + +In response to these limitations, we propose a multi-agent framework for generating persuasion data. In this framework, multiple agents are assigned distinct roles, ensuring that each aspect of the dialogue generation process is handled efficiently. This structure minimizes the risk of an agent missing important details due to task abstraction or prompt complexity, a common issue in LLM prompting [@prompt-complexity-1; @prompt-complexity-2]. Additionally, auxiliary agents manage dialogue flow to ensure that the resulting exchanges are coherent, logically consistent, and incorporate diverse persuasive strategies, simulating natural human conversation. Our approach imposes no preconditions regarding speakers, language styles, domains, or persuasion strategies, allowing it to generate a wide range of dialogues. For instance, our framework can support adversarial dialogues, where both participants attempt to persuade one another while maintaining their original positions. Moreover, we employ a continuous labeling scheme to measure the degree of perspective change throughout the dialogue, avoiding the limitations of binary utterance labels. This framework also integrates ethical considerations, incorporating cultural norms and taboos from NormBank [@ziems2023normbank] to explore ethically challenging persuasive scenarios, such as dialogues involving manipulation or unethical persuasion. + +Careful analyses conducted by experts from both NLP and social sciences confirm the quality of our generated dialogues, particularly in terms of their naturalness, logical structure, and diversity of persuasion strategies. Our sentence-level persuasiveness labels align closely with human judgments (see Appendix [12](#appendix:label-validation){reference-type="ref" reference="appendix:label-validation"}). + +We further demonstrate the flexibility of our framework through experiments controlling for specific persuasion strategies and in more complex scenarios, such as multi-party conversations. Across all tested configurations, our framework consistently produced high-quality dialogues, showcasing its adaptability and generalizability. These findings indicate that our framework offers a robust platform for studying persuasion techniques, particularly in high-stakes contexts where ethical concerns, such as the spread of misinformation and propaganda, are paramount [@chen2023combating; @jones2024detecting]. + +
+ +
Overview of our data generation and annotation framework. Prior to dialogue generation, each agent is assigned specific tasks and given predefined stances to maintain throughout the conversation.
+
+ +# Method + +Our framework incorporates 6 groups of language agents as shown in Figure [1](#fig:annotation-framework){reference-type="ref" reference="fig:annotation-framework"}. In our experiments, all agents utilize a GPT-3.5 backbone, except for the utterance quality monitor and global regulation agents which are based on GPT-4 due to their need for advanced reasoning capabilities and enhanced memory retention. Note that this choice of LLMs aims to balance data generation costs with quality, and using more powerful models could further improve the effectiveness of our approach. Our preliminary experiments on model selection are outlined in Appendix [6](#appendix:model-selection){reference-type="ref" reference="appendix:model-selection"}. + +We adopt a methodology for generating multi-round conversations by cyclically using the output from one language agent as the input for another [@small-town]. This technique has been validated to produce extended, logically consistent dialogues that fulfill our project requirements. + +Our framework initializes the generative agents with a description of the task settings, the predefined tasks for each language agent, and guidelines governing the models' generations, as illustrated in Figure [8](#fig:system-persuader-persuadee){reference-type="ref" reference="fig:system-persuader-persuadee"}. The task choices for each agent are not constrained, for instance, drawing on a cultural taboo that "one should not pick flowers in a cemetery" from NormBank, we could challenge the persuader to convince the persuadee to pick flowers in a cemetery, while the persuadee is instructed to resist and, if possible, persuade the persuader to abandon such thoughts. + +The dialogues commence when we prompt a persuader agent with "Start the conversation." This setup initiates a structured yet dynamic interaction between the speakers, allowing us to closely observe and analyze their persuasive strategies. + +Due to the inherent limitations of LLMs, dialogue generation agents may occasionally produce incomplete, repetitive, or off-topic content. To address these issues, we introduce a specialized LLM agent responsible for tracking the persuasion topic and generation history to evaluate new generations. + +The initialization prompt of the utterance quality monitor agent is shown in Figure [9](#fig:system-language-utterance-monitor){reference-type="ref" reference="fig:system-language-utterance-monitor"}. During dialogue generation, this agent inspects every new utterance to check if they [ends unexpectedly]{.underline}, [repeats a previous utterance]{.underline}, or [goes off the topic of the dialogue]{.underline} in a sequence. If an utterance is red-flagged for any issue, the author agent is requested to revise the utterance based on the diagnoses. Otherwise, before proceeding to the next utterance, the utterance quality monitor agent is prompted to update its memory, storing the reviewed utterance for future judgments. + +Raw text produced by dialogue generation agents often adopts a conclusive rather than conversational tone, primarily because the agents are prompted in a question-answering format. This could lead to stylistic conflicts with surrounding utterances. Additionally, the generations frequently include tone-softening phrases like "I understand your concerns," or unnecessary affirmations such as agreeing with the other speaker's views, which dilute the strength of arguments. Over the course of the conversation, these issues can compound, leading to dialogues dominated by language softeners and lacking in persuasive content. + +To address this issue, we adopt a language refinement agent tasked with stripping out polite but superfluous phrases, thereby sharpening the dialogue's focus on substantive content. System message to this agent is shown in Figure [10](#fig:system-language-refine){reference-type="ref" reference="fig:system-language-refine"}. 2 examples are also provided to the agent to further regulate its behaviors. Subsequent operations, including continued dialogue generation and persuasiveness labeling, are predicated on the output from the language refinement agent, ensuring that the conversation maintains its relevance and effectiveness in conveying persuasive arguments. + +After generating each round of conversation, we employ a persuasiveness annotation agent to assess the extent of perspective shifts in each speaker, assigning a score ranging from 0 to 1. Figure [11](#fig:system-annotation){reference-type="ref" reference="fig:system-annotation"} illustrates the system message fed to the persuasiveness annotation agent before the generation starts. In practice, we provide the annotation agent with two scoring examples to guide its behavior and minimize scoring errors, such as incorrectly assigning a score of 1 to a conversation round with no perspective shifts (Figure [22](#fig:25){reference-type="ref" reference="fig:25"}). Note that these scores reflect the cumulative viewpoint shifts across all prior rounds of communication, facilitating the analysis of gradual persuasion rather than focusing solely on the impact of a single utterance. + +We employ a global regulation agent to ensure smooth logical flow in the generated conversations and to determine the appropriate time to conclude the dialogue. The system message to the global regulation agent is depicted in Figure [12](#fig:system-regulation){reference-type="ref" reference="fig:system-regulation"}. + +After each round of utterances is generated and annotated, we prompt this agent to verify whether any changes in each speaker's perspectives are logically influenced by the preceding utterance and whether the newly generated utterances avoid repeating previously used strategies within the same conversation. If the logical connections are insufficient or no new persuasive attempts are made, the dialogue generation agents are asked to revise their responses based on feedback from the global regulation agent. Once the revised generation passes these checks, the agent's internal memory is updated accordingly. Then the agent is prompted to assess whether the speakers have reached a mutual agreement or if no new information is likely to be introduced next, indicating that the dialogue should be concluded. Although the ideal conclusion involves the persuader and persuadee agreeing on the preset task, conversations can often devolve into repetitive and unproductive arguments (Figure [23](#fig:112){reference-type="ref" reference="fig:112"}) [@xu2022learning]. To prevent such stagnation, we allow the dialogue to conclude even if complete agreement is not reached. The global regulation agent is responsible for determining when to end the dialogue, at which point the conversation is terminated and the agent's memory is reset. + +After generating and annotating a full dialogue, we use a postprocessing agent to enhance content smoothness and naturalness. As shown in Figure [13](#fig:system-postprocessing){reference-type="ref" reference="fig:system-postprocessing"}, the agent removes redundant language, improves logical flow, and enhances language diversity. It also merges labels and reassigns them to modified utterances if the number of dialogue rounds changes. + +::: table* +::: + +To evaluate our data generation framework, we constructed a small dataset of 200 dialogues using randomly selected norms from NormBank for human validation. These norms consist of 98 taboos, 76 normal behaviors, and 26 expected behaviors. We intentionally placed greater emphasis on taboos because these behaviors often conflict with widely accepted moral standards, causing LLMs to refuse to generate persuasive dialogues (Figure [14](#fig:exp-llm-refusal-taboo){reference-type="ref" reference="fig:exp-llm-refusal-taboo"}). As such, they present a unique challenge in persuasion scenarios for both humans and LLMs. + +Our data assessment plan focuses on three key aspects, progressing from more specific to broader levels of analysis: (1) the [language fluency]{.underline} of individual utterances, (2) the [the topic, semantic, and logical coherence]{.underline} of entire conversations, and (3) the [language and strategy diversity]{.underline} of conversations generated under the same topic and context. + +A critical goal for our framework is that each generated utterance should closely resemble a human-written sentence. To validate this, we conduct (a) a quantitative annotation task to differentiate between model-generated sentences and human-rewritten sentences, followed by (b) a qualitative error analysis that combines annotator feedback with insights from an LLM on sentences that multiple annotators agreed were distinguishable. + +The differentiation task aims to assess how accurately human annotators could tell model-generated sentences apart from those rewritten by humans. Similar tasks have been discussed in @gehrmann-etal-2019-gltr, @ippolito-etal-2020-automatic and @Ma2023AIVH. For our evaluation, we obtained a stratified sample of 400 utterances from 150 random sample dialogues to ensure equal representation of utterances from both the persuader and persuadee agents, covering different rounds of persuasion to reflect the dataset distribution. + +**Manual Rewriting.** 2 native English speakers were asked to rewrite each sampled utterance to provide reference texts that model-generated utterances will be compared against. Each assistant was assigned 200 utterances. As shown in Figure [15](#fig:instr_rewrite){reference-type="ref" reference="fig:instr_rewrite"}, they were instructed to retain the original meaning while improving clarity, grammar, and natural phrasing. Additionally, they could refine any awkward or unclear phrasing without altering the intended message. + +::: table* +::: + +**Human Validation.** After manual rewriting, we created a dataset consisting of pairs of model-generated utterances and their corresponding rewritten versions. 3 fluent English-speaking annotators were then tasked with identifying the model-generated utterance in each pair. The instructions provided to the annotators are shown in Figure [16](#fig:instr_human_val){reference-type="ref" reference="fig:instr_human_val"}. + +Each utterance in the dataset was annotated by 2 annotators, and annotators were encouraged to comment on examples they found interesting. Note that if the annotators were unable to distinguish between sentences and resorted to random guessing, the expected accuracy for both annotators correctly identifying model-generated utterances would be 25%. Comparing the actual accuracy to this baseline helps determine whether the model-generated utterances appeared natural to the annotators. + +Of the 400 utterance pairs, the model-generated utterances in 117 pairs (29.25%) were correctly identified by both annotators, slightly going above the random baseline of 25%. In 49% cases (98 pairs), the annotators disagreed, and in 21.75% cases, both annotators resulted in incorrect identifications. Individual annotator accuracies were 0.546, 0.558, and 0.508. The results are close to random guessing, suggesting the challenge of distinguishing utterances generated by our framework from human-written ones. Example utterance pairs and their annotator labels are provided in Table [\[table:utterance_example\]](#table:utterance_example){reference-type="ref" reference="table:utterance_example"}. + +The quantitative findings indicate that our framework generally produces high-quality utterances nearly indistinguishable from human-written sentences. To follow up, we conducted a qualitative error analysis on the samples correctly distinguished by both annotators to identify areas for improvement. Precisely, all 117 such utterances were submitted to OpenAI's o1-preview model [@OpenAI_2024] for further analysis to understand reasons behind their distinguishability by humans. The prompt for this task is shown in Figure [17](#fig:instr_llm_val){reference-type="ref" reference="fig:instr_llm_val"}. + +Out of the 117 pairs, o1-preview correctly distinguished 72 pairs (61.2%), suggesting that even for LLMs, utterances generated by our framework are close to human writings. As suggested by o1-preview and verified by human annotators (Table [\[table:utterance_error\]](#table:utterance_error){reference-type="ref" reference="table:utterance_error"}), major causes of unnaturalness (frequencies cited in parentheses) in these 72 less human-like utterances include overly formal language or detached tone or word use (88.9%), lengthy sentences, redundancy, verbosity, and repetition (68.1%), unnatural syntax, word choice, and language style (58.3%), complex sentence structures (34.7%), use of generic words and cliché phrases (23.6%), overly perfect grammar (13.9%), and LLM-style closing phrases (12.5%). + +We further conduct dialogue-level analyses on our sample data to ensure that each generated dialogue is logically coherent and effective in persuasion. + +We first developed a systematic rubric for evaluating the overall quality of persuasive dialogues. Our evaluation is conducted on (a) the local level, which examines each argument-response pair between the speakers, and (b) the global level, which considers the conversation as a whole. Evaluations are based on existing human evaluation dimensions for open dialogue systems and emphasize three key aspects: the interaction between persuader and persuadee, the consistency of individual participants across multiple rounds, and the alignment of utterances with the topic. Detailed criteria and their references are outlined in Table [\[table:dialogue_evaluation\]](#table:dialogue_evaluation){reference-type="ref" reference="table:dialogue_evaluation"}. + +::: table* +::: + +The local (round-level) evaluation focuses on 2 conventional dimensions in dialogue systems: Coherence and Informativeness. Coherence refers to round-level logical consistency, i.e., speakers respond to each other in a manner appropriate to commonsense and the given context [@li-sun-2018-syntactically; @Young_Cambria_Chaturvedi_Zhou_Biswas_Huang_2018; @wu-etal-2019-proactive; @liang2021]. Informativeness measures the quality and progression of information, ensuring responses align with prior dialogue context while introducing new information or arguments [@zhu-etal-2019-retrieval]. + +On the global (dialogue) level, the overall dialogue should feel as if it could have been generated by human speakers [@moghe-etal-2018-towards; @lin-etal-2019-moel]. First, speakers are assessed for overall fluency. We assessed the linguistic and stylistic quality of responses, ensuring arguments are communicated clearly and easy to follow, and that the conversation flows naturally [@wu-etal-2019-proactive]. In addition, we looked at internal consistency throughout the conversation, defined as the absence of sudden, unexplained shifts in position, intention, or objective of speech [@moghe-etal-2018-towards; @ji-etal-2022-achieving]. Since the conversations are generated specific to topics, we also evaluate topic consistency, i.e., whether the conversation remains on-topic throughout [@moghe-etal-2018-towards; @ji-etal-2022-achieving]. + +Annotators are asked to rate whether a series of statements, covering the above aspects, accurately describes the conversations on a three-point scale: 3 (accurate), 2 (somewhat accurate), and 1 (not accurate). 2 annotators participated in this task, each annotating the same set of 50 dialogues not overlapping with those used for utterance-level evaluations. Before annotation, a one-hour training session with examples was conducted to ensure both annotators fully understood the criteria. The annotators achieved an average unweighted Cohen's $\kappa$ of 0.483 (ranging from 0.339 to 0.657 across items) and an average linearly weighted Cohen's $\kappa$ of 0.529 (ranging from 0.339 to 0.657 across items), indicating relatively solid inter-rater consistency for human evaluations in Natural Language Generation tasks. Detailed scores and inter-rater consistency are reported in Table [\[table:dialogue_evaluation\]](#table:dialogue_evaluation){reference-type="ref" reference="table:dialogue_evaluation"}. + +The dialogues generated by our framework are in general rated high on local-level coherence and clarity, particularly for providing logical responses within the immediate context (average score: 2.969 out of 3) and achieving perfect clarity in the arguments (average score: 3 out of 3). At the global level, the agents remain on topic for most dialogues (average score: 2.878 out of 3) and maintain good role consistency in their objectives and stances (average score: 2.765 out of 3). + +
+ +
Frequency Distribution of Persuasion Strategies in Independently Generated Dialogues. The Y-axis indicates the proportion of each strategy used within the model-generated dialogues. Each bar represents the strategy distribution of a single dialogue, organized by generation topic. Our framework adapts to various persuasion topics.
+
+ +Figure [18](#fig:dialogue_44){reference-type="ref" reference="fig:dialogue_44"} exemplifies a highly-rated dialogue where the persuader addresses the persuadee's concerns about using the attic for food storage. The persuader begins by presenting the attic as an "efficient space-saver and emergency backup" and counters concerns about unstable temperatures and pests with solutions like "airtight containers" and monitoring. Despite the persuadee's repeated objections, the persuader suggests "proper insulation" and highlights the benefits of being prepared. Eventually, the persuadee proposes using a pantry, which resolves their concerns, and the persuader agrees. In this dialogue, both parties present new arguments relevant to the other party's proposal and ultimately reach a reasonable compromise. We provide another highly-rated example in Figure [19](#fig:dialogue_101){reference-type="ref" reference="fig:dialogue_101"}. + +::: table* ++-----------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| **Topic** | **Example Utterances** | ++:==========+:======================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================+ +| | *Persuader*: [Starting a family and having children can bring immense joy and fulfillment to one's life.`[``Emotion``]`]{.mark} [The experience of raising a child, watching them grow and learn, and forming a strong bond with them is truly priceless. It also gives you the opportunity to pass on your values, beliefs, and legacy to the next generation.`[``Outcome``]`]{.mark} | +| +-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| | *Persuadee*: I understand the joys and fulfillment that come with starting a family and raising children, but I think it's not the right path for everyone. There are many other ways to find fulfillment and purpose in life, and having children is a significant responsibility that should not be taken lightly. [It's important to consider factors like personal freedom, financial stability, and the impact on the environment before making such a life-changing decision. `[``Emotion``]`]{.mark} | ++-----------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| | *Persuader*: While mandates may raise coercion concerns, it's essential to consider the potential consequences of failing to achieve herd immunity through voluntary means alone. [Striking a balance between protecting public health and respecting individual autonomy is indeed vital. This may necessitate exploring alternative strategies and ensuring that communication, transparency, and support are prioritized to build trust and address concerns effectively. `[``Moral Appeals``]`]{.mark} The ultimate goal is to create a safe and healthy environment for all community members. | +| +-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| | *Persuadee*: It's crucial to explore alternative approaches that prioritize education, communication, and support to encourage vaccination uptake without resorting to mandates. [By fostering trust and addressing concerns openly, we can work towards a common goal of safeguarding public health while respecting personal choice and autonomy.`[``Emotion``]`]{.mark} Collaborative efforts and understanding from all sides are key to effectively navigating complex health challenges. | ++-----------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| | *Persuader*: The joy and fulfillment that comes with starting a family is immeasurable. [Raising children, watching them grow, and sharing life's special moments with them is a unique experience. It's an opportunity to create a loving environment, to shape and nurture future generations.`[``Scarcity, Outcome``]`]{.mark} The bond you forge with your children is priceless. I'd love to hear your thoughts on this. | +| +-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| | *Persuadee*: Text: Certainly, but I believe everyone has the right to choose their own path. [For me, fulfillment comes from focusing on my career, achieving personal goals, and making a difference in the world in my own way. `[``Emotion``]`]{.mark} The state of the world and the potential challenges future generations may face concern me, hence my choice to contribute differently. | ++-----------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| | *Persuader*: While it's true that values can be instilled and positive principles passed on through various avenues, the unique bond and influence of a supportive family unit play a vital role in shaping individuals and society. [Families, in their diverse forms, offer a foundation of love, stability, and shared experiences that can have a profound impact on personal development and well-being.`[``Emotion``]`]{.mark} | +| +-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +| | *Persuadee*: Text: While families indeed offer a foundation of love, stability, and shared experiences, it's important to remember that not everyone may have access to such supportive family structures or may find similar sources of love and stability elsewhere. [Many individuals create strong bonds and find support in chosen families, close friendships, or community networks.`[``Popularity``]`]{.mark} | ++-----------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ +::: + +Despite high overall performance, the dialogues received lower scores regarding introducing new information (2.337 out of 3) and maintaining naturalness (2.561 out of 3). Based on annotator feedback, we identified the following common issues that explain these lower scores: + +**Argument repetition.** A most common error is argument repetition, where speakers restate the same points over multiple rounds of conversation with only slight variations in phrasing. As Table [\[table:dialogue_error\]](#table:dialogue_error){reference-type="ref" reference="table:dialogue_error"}\[Argument Repetition\] shows, the persuader repeatedly emphasizes that refraining from picking flowers will help every visitors' enjoyment, while the persuadee reiterates the importance of striking a balance between nature appreciation and nature preservation. + +**Formalized Language.** Another common issue identified is the use of overly formal language and arguments. While both speakers articulate their arguments clearly, the language is respectful and often appears more polished and structured than what would be expected in natural, everyday interactions. In comparison, human interactions tend to be more casual and spontaneous. + +As exemplified in Table [\[table:dialogue_error\]](#table:dialogue_error){reference-type="ref" reference="table:dialogue_error"}\[Formalized Language\], the persuader's word choice such as "detriment" and "savor the food", and their description of eating with their hands as "relishing the moment" and "cherished tradition" are relatively formal descriptions given the context. The clear progression from one argument to the next also resembles a structured exchange, in contrast to more dynamic interactions with immediate reactions. + +**Decay of informativeness Over Rounds.** There is a general tendency for conversation informativeness to decrease over rounds. Both speakers introduce new information or arguments more frequently at the beginning of a conversation while later they tend to repeat or reinforce each other's arguments without adding substantive new content, especially when an agreement is reached. Table [\[table:dialogue_error2\]](#table:dialogue_error2){reference-type="ref" reference="table:dialogue_error2"} illustrates this point by comparing the earlier and later rounds of the same dialogue. + +One advantage of our framework is its ability to generate diverse persuasion dialogues across various topics and contexts by adapting its persuasion strategies to suit each context. Ideally, the model should also be able to vary its strategies within the same context across different replicates. + +
+ +
Heatmap displaying the cosine similarity between strategy distributions across different dialogues. Each group of 5 dialogues belongs to the same topic, with the grid indicating the different topics.
+
+ +To evaluate diversity across and within the same context, we identified 9 persuasive strategies based on existing literature (see Table [\[table:strategy_diversity\]](#table:strategy_diversity){reference-type="ref" reference="table:strategy_diversity"} for a full list of techniques and references) and designed a detailed human annotation task. The persuasion strategies are categorized into 5 groups, as outlined by @anand_2011. External validity involves appeals to external authority or expertise, or using popular experiences and arguments to build trust. Outcomes refers to highlighting potential consequences, such as benefits, risks, or engaging the persuadee through threats or promises. Generalizations involve framing an uptake as positive or negative, often incorporating a moral aspect. Interpersonal strategies focus on prompting individuals to connect, compete, or comply with others. Other tactics include logical and emotional appeals. + +For this task, we provided the framework with 5 topics covering controversial issues and personal decisions: mandatory vaccination, climate change regulation, increasing social media regulation, life in the countryside, and building a family. 5 dialogues are generated for each topic, resulting in 25 dialogues with 446 utterances in total. Human annotators then read each dialogue and identified all the strategies used by persuaders and persuadees. + +From this fine-grained annotation, we counted the frequency of different strategies and calculated the proportion of each strategy within each dialogue. The distribution of strategy usage is shown in Figure[2](#fig:strategy_distribution){reference-type="ref" reference="fig:strategy_distribution"}. Overall, the models used logical appeals and outcome descriptions more frequently than other strategies, which aligns with the goal of persuasion. However, there were significant variations in strategy usage across different contexts, indicating that the framework effectively adapts to each persuasion topic. For example, more emotional appeals were used when discussing personal matters, e.g., in a "building a family" dialogue, the framework emphasized the unique joys and fulfillment that come with having children, highlighting the personal growth it can provide. On the other hand, moral appeals are more prominent in policy discussions. For example, when addressing vaccination mandates, the framework stressed the importance of balancing public health with personal choice, fostering trust and collaboration to navigate complex health challenges. This reflects real-life persuasive strategies across different topics. + +Within each topic, the strategies used by the agents were not unchanged as well. For example, when discussing building family, 3 out of 5 dialogues used popularity appeals, and 3 out of 5 involved scarcity. Some examples of this are provided in [\[table:utterance_example\]](#table:utterance_example){reference-type="ref+label" reference="table:utterance_example"}. Additionally, there were notable differences in the distribution of moral appeals within the topics of vaccination mandates as well as social media regulation. + +Moreover, to compare the distributions of strategies within and between topics, we first represent each dialogue as a distribution of the strategies used. We then compute the cosine similarity between these distributions. The heatmap in Figure [3](#fig:heatmap){reference-type="ref" reference="fig:heatmap"} illustrates the similarity between pairs of dialogues, highlighting the distribution of persuasion strategies across five different topics. While higher similarity values along the diagonal indicate greater overlap in strategy usage within the same topic, variations in strategy selection still exist, demonstrating flexibility within topics. This suggests that our framework not only generates dialogues with diverse strategies across different topics but also maintains strategic variation within each topic, ensuring adaptability in dialogue generation. diff --git a/2510.08044/main_diagram/main_diagram.drawio b/2510.08044/main_diagram/main_diagram.drawio new file mode 100644 index 0000000000000000000000000000000000000000..5fd191a03d73db27016f708123f322f7108b2b18 --- /dev/null +++ b/2510.08044/main_diagram/main_diagram.drawio @@ -0,0 +1,88 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/2510.08044/paper_text/intro_method.md b/2510.08044/paper_text/intro_method.md new file mode 100644 index 0000000000000000000000000000000000000000..e17a2834402989f2592a085692b70f26550c37cf --- /dev/null +++ b/2510.08044/paper_text/intro_method.md @@ -0,0 +1,111 @@ +# Introduction + +Large Language Models (LLMs) have shown exceptional versatility [1, 2, 3], as evidenced by their ability to handle a wide array of tasks. These tasks include answering complex questions [4], solving mathematical problems [5], generating computer code [6], and performing sophisticated reasoning during inference [7]. When robots try to execute tasks instructed by language descriptions, they can leverage LLM capabilities to interpret task instructions in natural language [8, 9], apply the common sense reasoning of LLMs to understand their environment [10], and formulate high-level action plans based on the robot's abilities and available resources. + +However, a significant challenge with current LLMs is their propensity to hallucinate [11]—i.e., to confidently generate outputs that appear plausible but are actually incorrect and practically infeasible. [12]. This misplaced confidence in erroneous outputs presents a considerable obstacle to LLM-based planning in robotics. Additionally, natural language instructions in real-world environments often contain inherent or unintentional ambiguity from human instructors [13]. And following a flawed plan with excessive confidence may lead to undesirable or even unsafe actions. Consequently, offering an uncertainty estimation for the planning model prior to execution is desirable to reflect the level of confidence in their plans, allowing for halting execution or seeking human assistance in cases of high uncertainty. + +∗Correspondence to Chenjia Bai + +Previous studies on uncertainty in reinforcement learning (RL) typically distinguish two components of model uncertainty: epistemic uncertainty and intrinsic uncertainty [14]. However, in research on robotics LLM planners, these uncertainty components have not been analyzed at a fine-grained level. This lack of comprehensive analysis can lead to inaccurate assessments of model capabilities, such as misattributing uncertainty caused by environmental randomness to the model itself, thereby underestimating its true abilities [15]. Furthermore, when environmental changes occur, it is challenging to isolate and update intrinsic uncertainty, hindering rapid adaptation to new conditions. + +The empirical evidence from cognitive research reveals that clear task descriptions significantly enhance decision accuracy by providing a precise framework for cognitive processing and reducing the cognitive load associated with ambiguity [16]. Conversely, vague objectives tend to introduce substantial cognitive dissonance, thereby increasing hesitation and the likelihood of errors [17]. Similarly, the familiarity of tasks plays a crucial role in execution efficiency. Familiar tasks, which have been ingrained through repeated exposure and practice, allow for streamlined cognitive and motor processes, thereby optimizing performance [18]. In contrast, unfamiliar tasks necessitate additional cognitive resources for schema construction and adaptation, thus introducing greater uncertainty and potential inefficiencies. These cognitive insights indicate that epistemic uncertainty can be decomposed into components related to task description clarity and task familiarity, enabling more precise uncertainty modeling for LLM task planning. + +In this paper, we separately estimate planning uncertainty and introduce a new method for uncertainty estimation. First, we examine epistemic uncertainty in planning tasks, further breaking it down into two factors: task clarity and task familiarity. Second, we model intrinsic uncertainty as the expected success rate of a given plan. A lower expected success rate indicates a higher level of intrinsic uncertainty as it implies that despite following the plan, there still exist factors within the environment that may lead to failure. To evaluate these uncertainties, we use multi-layer perceptrons regression heads driven by LLM features to estimate task clarity and expected success rate, while a Random Network Distillation (RND) network is employed to assess task familiarity. In addition, we propose a slower yet more precise task clarity evaluation method based on LLM-query. By combining these components, we provide a comprehensive uncertainty assessment for LLM-based planning, enhancing the accuracy of uncertainty estimation. + +We compare the proposed algorithm with state-of-the-art uncertainty estimation techniques in robot planning, as well as widely-used uncertainty estimation methods in LLMs. The experiments include mobile manipulator in kitchen tasks and tabletop rearrangement tasks. The results demonstrate that the uncertainty estimates produced by the proposed method were the most accurate and comprehensive, showing the strongest correlation with the actual execution result. + +Our main contributions are summarized as follows: First, we propose a novel uncertainty evaluation method CURE that separately assesses epistemic uncertainty and intrinsic uncertainty to better align model confidence with task success. Second, we present an effective method for assessing task similarity in task planning, leveraging RND to enhance the foundation for uncertainty estimation. Third, our experiments with a mobile manipulator in kitchen and tabletop rearrangement scenarios demonstrate that the proposed algorithm significantly enhances the accuracy of uncertain estimation, outperforming existing methods in embodied planning and LLM uncertain estimation. + +# Method + +In this section, we introduce the core concepts underlying our proposed CURE framework, including the problem formulation and an overview of RND. + +Problem statement. We consider a manipulation problem in which a high-level, free-form human instruction I is given (e.g., Give me something to drink). The observation O includes an instruction pertaining to the current environment. An LLM planner decomposes the task into an object-centric executable task plan A. The central problem investigated in this work is how to generate an uncertainty estimate U that quantifies the planner's confidence in the task planning process. For the convenience in subsequent processing, we input I, O, and A into the LLama3.2-8B model to obtain the hidden activations of the last token from the last layer, which serve as the task encoding vector T. + +Let S = {s1, s2, . . . , sn} represent the set of observed success rates, and U = {u1, u2, . . . , un} denote the corresponding set of planning uncertainty estimates, where n is the number of samples. + +![](_page_2_Figure_0.jpeg) + +Figure 1: Overview of the proposed uncertainty-aware LLM planning framework. Given a natural language instruction (e.g., Give me sth to drink) and environmental context (e.g., Coke, Sprite, Apple), the LLM planner generates a high-level plan (e.g., I will give user Coke). Our framework estimates the overall planning uncertainty using CURE module, which decomposes uncertainty into epistemic and intrinsic components. Epistemic uncertainty encompasses task similarity and task clarity, while intrinsic uncertainty is represented by the expected success rate of the generated plan. The final uncertainty score then guides the decision to proceed, halt, or request clarification, thereby enhancing planning reliability in uncertain or ambiguous scenarios. + +The objective is to maximize the statistical dependence between S and U, as measured by Spearman's rank correlation coefficient $\rho$ . Formally, we aim to solve: + +$$\max \rho(S, U) \tag{1}$$ + +where $\rho(S, U)$ is defined as: + +$$\rho(S, U) = 1 - \frac{6\sum_{i=1}^{n} d_i^2}{n(n^2 - 1)}, \quad d_i = \text{rank}(s_i) - \text{rank}(u_i).$$ + (2) + +Here, $rank(s_i)$ and $rank(u_i)$ denote the ranks of $s_i$ and $u_i$ within their respective datasets. + +Random Network Distillation. Random Network Distillation (RND) is an approach initially utilized to measure the novelty of a given state, aiming to encourage exploration for rarely-visited states in large state space in RL. The core idea behind RND is to use a randomly initialized neural network, referred to as the *target network*, which is kept fixed during training. A separate *predictor network* is then trained to predict the output of the target network, typically via a regression loss. The target network's weights are not updated throughout the training process, and its outputs remain fixed. This fixed target network is essentially a source of "random" or "unknown" knowledge, and the predictor network attempts to minimize the difference between its predictions and the outputs of the target network. The error in the predictor network's predictions provides a measure of how well the predictor is able to capture the structure of the environment. High prediction errors typically indicate regions in the environment that are novel or unfamiliar to the agent, and low prediction errors signify regions that are more familiar, as the predictor network has learned to approximate the target network's behavior well. Formally, let $f_{\theta}$ denote the target network with parameters $\theta$ , and let $g_{\phi}$ represent the predictor network with parameters $\phi$ . The objective is to minimize the following loss: + +$$\mathcal{L}(\phi) = \mathbb{E}_{x \sim \mathcal{X}} \left[ \|g_{\phi}(x) - f_{\theta}(x)\|^2 \right], \tag{3}$$ + +where $\mathcal{X}$ represents the input space (typically the state or observation space of the task). The error between $g_{\phi}(x)$ and $f_{\theta}(x)$ reflects how well the predictor network approximates the fixed target network. In environments where the target network is highly unpredictable, the prediction error will be large, indicating the currently visited states are not familiar to the historical states. Utilizing this characteristic, we can assess the familiarity of tasks. + +Sections 3.1 and 3.2 provide a detailed description of the CURE architecture and our approach to estimating task familiarity. Next, Section 3.3 delves into the evaluation of task clarity and the prediction of the success rate. + +We propose an overview of our method, CURE, which operates independently of any specific LLM planner. Additionally, this approach is crafted to be plug-and-play, meaning it requires no alterations to the foundational structures of the planners. After task planning is completed, we separately compute epistemic uncertainty and intrinsic uncertainty associated with the planning process. Epistemic uncertainty is further divided into task familiarity and task clarity. First, we employ RND to estimate task familiarity, yielding Asim. Next, task clarity is assessed using two proposed methods: (i) A query-based approach leveraging the LLM, which is relatively slower. (ii) A multi-layer neural network inference approach, which is computationally efficient. Both methods yield a measure of task clarity, denoted as Aamb. Finally, intrinsic uncertainty is computed using a multi-layer neural network to infer the expected success rate p. The final uncertainty U is determined using the following formula: + +$$U = 1 - \alpha_1 \cdot (1 - \alpha_2 \cdot A_{amb}) \cdot p + \alpha_3 \cdot A_{sim}, \tag{4}$$ + +where α1, α2, α3 are tunable parameters that balance the contributions of task clarity, expected task success rate, and task similarity to overall uncertainty. + +Firstly, we define the two key components in the RND framework: the target network and the predictor network. The outputs of the target network ftarget and the predictor network fpred are the embedding vectors ztarget and zpred, respectively. Specifically: + +$$\mathbf{z}_{\text{target}} = f_{\text{target}}(\mathbf{T}),$$ + +$$\mathbf{z}_{\text{pred}} = f_{\text{pred}}(\mathbf{T}).$$ +(5) + +where T denotes the feature vector of the task description, as detailed in Section 2. + +During the training phase of the task, we train the RND network by inputting the vectors of known tasks, employing the Mean Squared Error (MSE) as the loss function to optimize the output discrepancy between the target network and the predictor network: + +$$\mathcal{L}_{MSE} = \frac{1}{|\mathcal{D}|} \sum_{(\mathbf{T}, \{\mathbf{S}_i\}) \in \mathcal{D}} \|\mathbf{z}_{target} - \mathbf{z}_{pred}\|_2^2, (6)$$ + +![](_page_3_Picture_12.jpeg) + +Figure 2: The process of task familiarity assessment + +where D is the dataset of known tasks. + +During the inference phase, the new task description vector T is input into the trained RND network. By computing the Euclidean distance between the outputs of the target network and the predictor network, we obtain the contextual similarity metric Asim: + +$$A_{\text{sim}} = \|\mathbf{z}_{\text{target}} - \mathbf{z}_{\text{pred}}\|_{2}.\tag{7}$$ + +This similarity metric reflects the degree of similarity between the current task and the known tasks. A larger value of Asim indicates a greater difference between the current task and known tasks, suggesting that the task is relatively unfamiliar, thereby appropriately increasing the uncertainty in the output. Conversely, a higher similarity indicates that the current task is more familiar, allowing for a reduction in the uncertainty assessment. + +**Slow Evaluation of Task Clarity Using LLM (Ambiguity)** We employed a vanilla approach to assess task clarity by designing a dedicated prompt for ambiguity evaluation. This prompt requires the model to determine whether the given task description provides sufficient information to infer the intended object and location of the user's intent. The specific prompt can be seen in Section C: + +Through this prompt, we prompt the model to determine whether the task description is sufficiently clear to infer the user's intended object and location. If the model deems the task description inadequate for inference (choose multi item or target location), we consider the task to exhibit semantic ambiguity, thereby increasing the uncertainty of the output. Conversely, if the task description is sufficiently clear (choose one item and one target location), the model's response indicates that the task instructions are adequate, thereby reducing uncertainty. + +The final assessment result is obtained as follows: + +$$A_{\text{amb}} = \begin{cases} 0 & \text{Sufficient (Information Complete)} \\ 1 & \text{Insufficient (Ambiguity Present)} \end{cases}$$ + (8) + +Fast Evaluation of Task Clarity and Expected Success Rate Using Uncertainty Assessment Network (UAN) First, we construct a training dataset using either code generation or LLM-generated data. This dataset comprises the task encoding vector T, the presence of task ambiguity $A_{\rm amb} \in \{0,1\}$ , and whether the planned subtasks were successfully executed $y \in \{0,1\}$ . + +We design a multilayer neural network $f_{\text{UAN}}$ , which takes as input the feature vector $\mathbf{T}$ and outputs two scalar values: + +- 1. Task clarity score $\hat{A}_{amb} \in [0,1]$ +- 2. Expected task success rate $\hat{p} \in [0, 1]$ + +The network training process employs a crossentropy loss function: + +$$\mathcal{L}_{\text{total}} = \mathcal{L}_{\text{amb}} + \mathbb{I}(A_{\text{amb}} = 0) \cdot \mathcal{L}_{\text{success}}$$ + (9) + +![](_page_4_Picture_11.jpeg) + +Figure 3: **The process of UAN**. During the training process, Llama consistently remained frozen. For tasks with clear objectives, both task clarity and expected success rate were trained. In tasks with ambiguous goals, only task clarity was trained. + +where $\mathbb{I}(\cdot)$ is an indicator function, and the loss terms are defined as follows: + +$$\mathcal{L}_{amb} = -\left(A_{amb}\log(\hat{A}_{amb}) + (1 - A_{amb})\log(1 - \hat{A}_{amb})\right)$$ +(10) + +$$\mathcal{L}_{\text{success}} = -\left(y\log(\hat{p}) + (1-y)\log(1-\hat{p})\right) \tag{11}$$