_id stringlengths 4 10 | text stringlengths 0 18.4k | title stringlengths 0 8.56k |
|---|---|---|
d258833682 | In standard adversarial training, models are optimized to fit one-hot labels within allowable adversarial perturbation budgets. However, the ignorance of underlying distribution shifts brought by perturbations causes the problem of robust overfitting. To address this issue and enhance adversarial robustness, we analyze the characteristics of robust models and identify that robust models tend to produce smoother and well-calibrated outputs. Based on the observation, we propose a simple yet effective method, Annealing Self-Distillation Rectification (ADR), which generates soft labels as a better guidance mechanism that accurately reflects the distribution shift under attack during adversarial training. By utilizing ADR, we can obtain rectified distributions that significantly improve model robustness without the need for pre-trained models or extensive extra computation. Moreover, our method facilitates seamless plug-and-play integration with other adversarial training techniques by replacing the hard labels in their objectives. We demonstrate the efficacy of ADR through extensive experiments and strong performances across datasets.Preprint. Under review. | Annealing Self-Distillation Rectification Improves Adversarial Training |
d3511502 | Recent research has shown that one can train a neural network with binary weights and activations at train time by augmenting the weights with a high-precision continuous latent variable that accumulates small changes from stochastic gradient descent. However, there is a dearth of theoretical analysis to explain why we can effectively capture the features in our data with binary weights and activations. Our main result is that the neural networks with binary weights and activations trained using the method of Courbariaux, work because of the highdimensional geometry of binary vectors. In particular, the ideal continuous vectors that extract out features in the intermediate representations of these BNNs are wellapproximated by binary vectors in the sense that dot products are approximately preserved. Compared to previous research that demonstrated the viability of such BNNs, our work explains why these BNNs work in terms of the HD geometry. Our theory serves as a foundation for understanding not only BNNs but a variety of methods that seek to compress traditional neural networks. Furthermore, a better understanding of multilayer binary neural networks serves as a starting point for generalizing BNNs to other neural network architectures such as recurrent neural networks.Recent work by and has shown that one can efficiently train neural networks with binary weights and activations that have similar performance to their continuous counterparts. They demonstrate that such BNNs execute 7 times faster using a dedicated GPU kernel at test time. Furthermore, they argue that such BNNs require at least a factor of 32 fewer memory accesses at test time that should result in an even larger energy savings. There are two key ideas in their papers(Fig. 1). First, they associate a continuous weight, w c , with each binary weight, w b , that accumulates small changes from stochastic gradient descent. Second, they replace the non-differentiable binarize function (θ(x) = 1 if x > 0 and −1 otherwise) with a continuous one during backpropagation. These modifications allow them to train neural networks that have | The High-Dimensional Geometry of Binary Neural Networks |
d262013578 | We explore the impact of parameter sparsity on the scaling behavior of Transformers trained on massive datasets (i.e., "foundation models"), in both vision and language domains.In this setting, we identify the first scaling law describing the relationship between weight sparsity, number of non-zero parameters, and amount of training data, which we validate empirically across model and data scales; on ViT/JFT-4B and T5/C4.These results allow us to characterize the "optimal sparsity", the sparsity level which yields the best performance for a given effective model size and training budget.For a fixed number of non-zero parameters, we identify that the optimal sparsity increases with the amount of data used for training.We also extend our study to different sparsity structures (such as the hardware-friendly n:m pattern) and strategies (such as starting from a pretrained dense model).Our findings shed light on the power and limitations of weight sparsity across various parameter and computational settings, offering both theoretical understanding and practical implications for leveraging sparsity towards computational efficiency improvements. | SCALING LAWS FOR SPARSELY-CONNECTED FOUNDATION MODELS |
d247779010 | Machine learning models that achieve high overall accuracy often make systematic errors on important subsets (or slices) of data. Identifying underperforming slices is particularly challenging when working with high-dimensional inputs (e.g. images, audio), where important slices are often unlabeled. In order to address this issue, recent studies have proposed automated slice discovery methods (SDMs), which leverage learned model representations to mine input data for slices on which a model performs poorly. To be useful to a practitioner, these methods must identify slices that are both underperforming and coherent (i.e. united by a human-understandable concept). However, no quantitative evaluation framework currently exists for rigorously assessing SDMs with respect to these criteria. Additionally, prior qualitative evaluations have shown that SDMs often identify slices that are incoherent. In this work, we address these challenges by first designing a principled evaluation framework that enables a quantitative comparison of SDMs across 1,235 slice discovery settings in three input domains (natural images, medical images, and time-series data). Then, motivated by the recent development of powerful cross-modal representation learning approaches, we present Domino, an SDM that leverages cross-modal embeddings and a novel error-aware mixture model to discover and describe coherent slices. We find that Domino accurately identifies 36% of the 1,235 slices in our framework -a 12 percentage point improvement over prior methods. Further, Domino is the first SDM that can provide natural language descriptions of identified slices, correctly generating the exact name of the slice in 35% of settings. | DOMINO: DISCOVERING SYSTEMATIC ERRORS WITH CROSS-MODAL EMBEDDINGS |
d222272028 | Continual (sequential) training and multitask (simultaneous) training are often attempting to solve the same overall objective: to find a solution that performs well on all considered tasks. The main difference is in the training regimes, where continual learning can only have access to one task at a time, which for neural networks typically leads to catastrophic forgetting. That is, the solution found for a subsequent task does not perform well on the previous ones anymore. However, the relationship between the different minima that the two training regimes arrive at is not well understood. What sets them apart? Is there a local structure that could explain the difference in performance achieved by the two different schemes? Motivated by recent work showing that different minima of the same task are typically connected by very simple curves of low error, we investigate whether multitask and continual solutions are similarly connected. We empirically find that indeed such connectivity can be reliably achieved and, more interestingly, it can be done by a linear path, conditioned on having the same initialization for both. We thoroughly analyze this observation and discuss its significance for the continual learning process. Furthermore, we exploit this finding to propose an effective algorithm that constrains the sequentially learned minima to behave as the multitask solution. We show that our method outperforms several state of the art continual learning algorithms on various vision benchmarks 1 . * Equal contribution 1 The code is available at: https://github.com/imirzadeh/MC-SGD arXiv:2010.04495v1 [cs.LG] | LINEAR MODE CONNECTIVITY IN MULTITASK AND CONTINUAL LEARNING |
d252408526 | Modern machine learning research relies on relatively few carefully curated datasets. Even in these datasets, and typically in 'untidy' or raw data, practitioners are faced with significant issues of data quality and diversity which can be prohibitively labor intensive to address. Existing methods for dealing with these challenges tend to make strong assumptions about the particular issues at play, and often require a priori knowledge or metadata such as domain labels. Our work is orthogonal to these methods: we instead focus on providing a unified and efficient framework for Metadata Archaeology -uncovering and inferring metadata of examples in a dataset. We curate different subsets of data that might exist in a dataset (e.g. mislabeled, atypical, or out-ofdistribution examples) using simple transformations, and leverage differences in learning dynamics between these probe suites to infer metadata of interest. Our method is on par with far more sophisticated mitigation methods across different tasks: identifying and correcting mislabeled examples, classifying minority-group samples, prioritizing points relevant for training and enabling scalable human auditing of relevant examples. | Metadata Archaeology: Unearthing Data Subsets by Leveraging Training Dynamics |
d32737310 | Softmax loss is widely used in deep neural networks for multi-class classification, where each class is represented by a weight vector, a sample is represented as a feature vector, and the feature vector has the largest projection on the weight vector of the correct category when the model correctly classifies a sample. To ensure generalization, weight decay that shrinks the weight norm is often used as regularizer. Different from traditional learning algorithms where features are fixed and only weights are tunable, features are also tunable as representation learning in deep learning. Thus, we propose feature incay to also regularize representation learning, which favors feature vectors with large norm when the samples can be correctly classified. With the feature incay, feature vectors are further pushed away from the origin along the direction of their corresponding weight vectors, which achieves better inter-class separability. In addition, the proposed feature incay encourages intra-class compactness along the directions of weight vectors by increasing the small feature norm faster than the large ones. Empirical results on MNIST, CIFAR10 and CIFAR100 demonstrate feature incay can improve the generalization ability. | Feature Incay for Representation Regularization |
d247222761 | Drawing inspiration from gradient-based meta-learning methods with infinitely small gradient steps, we introduce Continuous-Time Meta-Learning (COMLN), a meta-learning algorithm where adaptation follows the dynamics of a gradient vector field. Specifically, representations of the inputs are meta-learned such that a taskspecific linear classifier is obtained as a solution of an ordinary differential equation (ODE). Treating the learning process as an ODE offers the notable advantage that the length of the trajectory is now continuous, as opposed to a fixed and discrete number of gradient steps. As a consequence, we can optimize the amount of adaptation necessary to solve a new task using stochastic gradient descent, in addition to learning the initial conditions as is standard practice in gradient-based meta-learning. Importantly, in order to compute the exact meta-gradients required for the outer-loop updates, we devise an efficient algorithm based on forward mode differentiation, whose memory requirements do not scale with the length of the learning trajectory, thus allowing longer adaptation in constant memory. We provide analytical guarantees for the stability of COMLN, we show empirically its efficiency in terms of runtime and memory usage, and we illustrate its effectiveness on a range of few-shot image classification problems. . Modular Meta-Learning with Shrinkage. Neural Information Processing Systems, 2020b.John R Dormand and Peter J Prince. A family of embedded Runge-Kutta formulae. Journal of computational and applied mathematics, 1980.Leonhard Euler. De integratione aequationum differentialium per approximationem. Opera Omnia, 1913.William F Feehery, John E Tolsma, and Paul I Barton. Efficient sensitivity analysis of large-scale differential-algebraic systems. | CONTINUOUS-TIME META-LEARNING WITH FORWARD MODE DIFFERENTIATION |
d218889280 | Under mild regularity conditions, gradient-based methods converge globally to a critical point in the single-loss setting. This is known to break down for vanilla gradient descent when moving to multi-loss optimization, but can we hope to build some algorithm with global guarantees? We negatively resolve this open problem by proving that any reasonable algorithm will exhibit limit cycles or diverge to infinite losses in some differentiable game, even in two-player games with zero-sum interactions. A reasonable algorithm is simply one which avoids strict maxima, an exceedingly weak assumption since converging to maxima would be the opposite of minimization. This impossibility theorem holds even if we impose existence of a strict minimum and no other critical points. The proof is constructive, enabling us to display explicit limit cycles for existing gradient-based methods. Nonetheless, it remains an open question whether cycles arise in high-dimensional games of interest to ML practitioners, such as GANs or multi-agent RL.Preprint. Under review. | On the Impossibility of Global Convergence in Multi-Loss Optimization |
d249209577 | Strategic classification studies learning in settings where users can modify their features to obtain favorable predictions. Most current works focus on simple classifiers that trigger independent user responses. Here we examine the implications of learning with more elaborate models that break the independence assumption. Motivated by the idea that applications of strategic classification are often social in nature, we focus on graph neural networks, which make use of social relations between users to improve predictions. Using a graph for learning introduces inter-user dependencies in prediction; our key point is that strategic users can exploit these to promote their own goals. As we show through analysis and simulation, this can work either against the system-or for it. Based on this, we propose a differentiable framework for strategically-robust learning of graph-based classifiers. Experiments on several real networked datasets demonstrate the utility of our approach. * Equal contribution, alphabetical order | STRATEGIC CLASSIFICATION WITH GRAPH NEURAL NETWORKS |
d259145304 | We investigate learning the equilibria in non-stationary multi-agent systems and address the challenges that differentiate multi-agent learning from single-agent learning. Specifically, we focus on games with bandit feedback, where testing an equilibrium can result in substantial regret even when the gap to be tested is small, and the existence of multiple optimal solutions (equilibria) in stationary games poses extra challenges. To overcome these obstacles, we propose a versatile black-box approach applicable to a broad spectrum of problems, such as general-sum games, potential games, and Markov games, when equipped with appropriate learning and testing oracles for stationary environments. Our algorithms can achieve O ∆ 1/4 T 3/4 regret when the degree of nonstationarity, as measured by total variation ∆, is known, and O ∆ 1/5 T 4/5 regret when ∆ is unknown, where T is the number of rounds. Meanwhile, our algorithm inherits the favorable dependence on number of agents from the oracles. As a side contribution that may be independent of interest, we show how to test for various types of equilibria by a black-box reduction to single-agent learning, which includes Nash equilibria, correlated equilibria, and coarse correlated equilibria. arXiv:2306.07465v1 [cs.LG] 12 Jun 2023 allows us to leverage existing algorithms for various stationary games, while also facilitating seamless adaptation to future algorithms that may offer improved guarantees.Main Contributions and Novelties1. Identifying challenges in non-stationary games with bandit feedback. First, we point out that the bandit feedback is incompatible with online-learning based algorithms. Then, we show that the bandit feedback and non-uniqueness of equilibria greatly complicates the application of test-based algorithms. Additionally, we point out that it is non-trivial to generalize an algorithm for non-stationary Markov games to a parameter-free version. 2. Generic black-box approach for non-stationary games. Our approach is a black-box reduction that can transform any base algorithm designed for (near-)stationary games into an algorithm capable of learning in a non-stationary environment. This approach not only inherits potential benign properties of the base algorithm, such as breaking the curse of multiagents and decentralization, but also directly adapts to future algorithmic advancements. 3. Restart-based algorithm when non-stationarity budget is known. Consider the case where we know a bound on the degree of non-stationarity, often measured by switching number or total variation (which from here on, we refer to as the "nonstationarity budget"). In this case, we design a simple restart-based algorithm achieving sublinear dynamic equilibrium regret of O(L 1/4 T 3/4 ) or O(∆ 1/4 T 3/4 ), where L is the switching number and ∆ is the total variation non-stationarity budget. In words, this result implies that all the players follow a near-equilibrium strategy in most episodes. 4. Multi-scale testing algorithm when non-stationarity budget is unknown. We also propose a multi-scale testing algorithm to optimize the regret when the non-stationarity budget is unknown, which can adaptively avoid the strategy deviating from equilibrium for too many rounds. The algorithm can achieve the same O(L 1/4 T 3/4 ) regret for unknown switching number L while a marginally higher O(∆ 1/4 T 13/16 ) regret for unknown total variation budget ∆. -armed-bandit problem with non-stationary rewards. Advances in neural information processing systems, 27, 2014. . Settling the complexity of computing two-player nash equilibria. . Reinforcement learning for non-stationary markov decision processes: The blessing of (more) Learning in congestion games with bandit feedback. arXiv preprint arXiv:2206.01880, 2022.Qiwen Cui, Kaiqing Zhang, and Simon S Du. Breaking the curse of multiagents in a large state space: Rl in markov games with independent linear function approximation. arXiv preprint arXiv:2302.03673, 2023.Constantinos Daskalakis, Noah Golowich, and Kaiqing Zhang. The complexity of markov equilibrium in stochastic games. arXiv preprint arXiv:2204.03991, 2022. | A Black-box Approach for Non-stationary Multi-agent Reinforcement Learning |
d247292293 | There is increasing evidence suggesting neural networks' sensitivity to distribution shifts, so that research on out-of-distribution (OOD) generalization comes into the spotlight. Nonetheless, current endeavors mostly focus on Euclidean data, and its formulation for graph-structured data is not clear and remains under-explored, given two-fold fundamental challenges: 1) the inter-connection among nodes in one graph, which induces non-IID generation of data points even under the same environment, and 2) the structural information in the input graph, which is also informative for prediction. In this paper, we formulate the OOD problem on graphs and develop a new invariant learning approach, Explore-to-Extrapolate Risk Minimization (EERM), that facilitates graph neural networks to leverage invariance principles for prediction. EERM resorts to multiple context explorers (specified as graph structure editers in our case) that are adversarially trained to maximize the variance of risks from multiple virtual environments. Such a design enables the model to extrapolate from a single observed environment which is the common case for node-level prediction. We prove the validity of our method by theoretically showing its guarantee of a valid OOD solution and further demonstrate its power on various real-world datasets for handling distribution shifts from artificial spurious features, cross-domain transfers and dynamic graph evolution 1 . | HANDLING DISTRIBUTION SHIFTS ON GRAPHS: AN INVARIANCE PERSPECTIVE |
d258832670 | Text-driven diffusion models have unlocked unprecedented abilities in image generation, whereas their video counterpart still lags behind due to the excessive training cost of temporal modeling. Besides the training burden, the generated videos also suffer from appearance inconsistency and structural flickers, especially in long video synthesis. To address these challenges, we design a training-free framework called ControlVideo to enable natural and efficient text-to-video generation. ControlVideo, adapted from ControlNet, leverages coarsely structural consistency from input motion sequences, and introduces three modules to improve video generation. Firstly, to ensure appearance coherence between frames, ControlVideo adds fully cross-frame interaction in self-attention modules. Secondly, to mitigate the flicker effect, it introduces an interleaved-frame smoother that employs frame interpolation on alternated frames. Finally, to produce long videos efficiently, it utilizes a hierarchical sampler that separately synthesizes each short clip with holistic coherency. Empowered with these modules, ControlVideo outperforms the state-of-the-arts on extensive motion-prompt pairs quantitatively and qualitatively. Notably, thanks to the efficient designs, it generates both short and long videos within several minutes using one NVIDIA 2080Ti. Code is available at https://github.com/YBYBZhang/ControlVideo.Recent studies[15,40]have explored leveraging the structure controllability of ControlNet[43]or DDIM inversion [35] for video generation. Rather than synthesizing all frames independently,[15,40]enhance appearance coherence by replacing original self-attention with the sparser crossframe attention. Nevertheless, their video quality is still far behind photo-realistic videos in terms of: (i) inconsistent appearance between some frames (seeFig. 4 (a)), (ii) visible artifacts in large motion videos (seeFig. 4(b)), and (iii) structural flickers during inter-frame transitions. For (i) andPreprint. Under review. | ControlVideo: Training-free Controllable Text-to-Video Generation |
d245334722 | Currently, it is hard to reap the benefits of deep learning for Bayesian methods, which allow the explicit specification of prior knowledge and accurately capture model uncertainty. We present Prior-Data Fitted Networks (PFNs). PFNs leverage large-scale machine learning techniques to approximate a large set of posteriors. The only requirement for PFNs to work is the ability to sample from a prior distribution over supervised learning tasks (or functions). Our method restates the objective of posterior approximation as a supervised classification problem with a set-valued input: it repeatedly draws a task (or function) from the prior, draws a set of data points and their labels from it, masks one of the labels and learns to make probabilistic predictions for it based on the set-valued input of the rest of the data points. Presented with a set of samples from a new supervised learning task as input, PFNs make probabilistic predictions for arbitrary other data points in a single forward propagation, having learned to approximate Bayesian inference. We demonstrate that PFNs can near-perfectly mimic Gaussian processes and also enable efficient Bayesian inference for intractable problems, with over 200-fold speedups in multiple setups compared to current methods. We obtain strong results in very diverse areas such as Gaussian process regression, Bayesian neural networks, classification for small tabular data sets, and few-shot image classification, demonstrating the generality of PFNs. Code and trained PFNs are released at https://github. com/automl/TransformersCanDoBayesianInference. | TRANSFORMERS CAN DO BAYESIAN INFERENCE |
d209324424 | Lack of reliability is a well-known issue for reinforcement learning (RL) algorithms. This problem has gained increasing attention in recent years, and efforts to improve it have grown substantially. To aid RL researchers and production users with the evaluation and improvement of reliability, we propose a set of metrics that quantitatively measure different aspects of reliability. In this work, we focus on variability and risk, both during training and after learning (on a fixed policy). We designed these metrics to be general-purpose, and we also designed complementary statistical tests to enable rigorous comparisons on these metrics. In this paper, we first describe the desired properties of the metrics and their design, the aspects of reliability that they measure, and their applicability to different scenarios. We then describe the statistical tests and make additional practical recommendations for reporting results. The metrics and accompanying statistical tools have been made available as an open-source library. 1 We apply our metrics to a set of common RL algorithms and environments, compare them, and analyze the results. * Work done as part of the Google AI Residency 1 https://github.com/google-research/rl-reliability-metrics arXiv:1912.05663v2 [stat.ML] | MEASURING THE RELIABILITY OF REINFORCEMENT LEARNING ALGORITHMS |
d260334759 | Despite the advancements of open-source large language models (LLMs), e.g., LLaMA, they remain significantly limited in tool-use capabilities, i.e., using external tools (APIs) to fulfill human instructions.The reason is that current instruction tuning largely focuses on basic language tasks but ignores the tool-use domain.This is in contrast to the excellent tool-use capabilities of state-of-the-art (SOTA) closed-source LLMs, e.g., ChatGPT.To bridge this gap, we introduce ToolLLM, a general tool-use framework encompassing data construction, model training, and evaluation.We first present ToolBench, an instruction-tuning dataset for tool use, which is constructed automatically using ChatGPT.Specifically, the construction can be divided into three stages: (i) API collection: we collect 16, 464 real-world RESTful APIs spanning 49 categories from RapidAPI Hub; (ii) instruction generation: we prompt ChatGPT to generate diverse instructions involving these APIs, covering both single-tool and multi-tool scenarios; (iii) solution path annotation: we use ChatGPT to search for a valid solution path (chain of API calls) for each instruction.To enhance the reasoning capabilities of LLMs, we develop a novel depth-first search-based decision tree algorithm.It enables LLMs to evaluate multiple reasoning traces and expand the search space.Moreover, to evaluate the tool-use capabilities of LLMs, we develop an automatic evaluator: ToolEval.Based on ToolBench, we fine-tune LLaMA to obtain an LLM ToolLLaMA, and equip it with a neural API retriever to recommend appropriate APIs for each instruction.Experiments show that ToolLLaMA demonstrates a remarkable ability to execute complex instructions and generalize to unseen APIs, and exhibits comparable performance to ChatGPT.Our ToolLLaMA also demonstrates strong zero-shot generalization ability in an out-of-distribution tool-use dataset: APIBench.The codes, trained models, and demo are publicly available at https://github.com/OpenBMB/ToolBench. | TOOLLLM: FACILITATING LARGE LANGUAGE MODELS TO MASTER 16000+ REAL-WORLD APIS |
d202565538 | While most approaches to the problem of Inverse Reinforcement Learning (IRL) focus on estimating a reward function that best explains an expert agent's policy or demonstrated behavior on a control task, it is often the case that such behavior is more succinctly described by a simple reward combined with a set of hard constraints. In this setting, the agent is attempting to maximize cumulative rewards subject to these given constraints on their behavior. We reformulate the problem of IRL on Markov Decision Processes (MDPs) such that, given a nominal model of the environment and a nominal reward function, we seek to estimate state, action, and feature constraints in the environment that motivate an agent's behavior. Our approach is based on the Maximum Entropy IRL framework, which allows us to reason about the likelihood of an expert agent's demonstrations given our knowledge of an MDP. Using our method, we can infer which constraints can be added to the MDP to most increase the likelihood of observing these demonstrations. We present an algorithm which iteratively infers the Maximum Likelihood Constraint to best explain observed behavior, and we evaluate its efficacy using both simulated behavior and recorded data of humans navigating around an obstacle. | Maximum Likelihood Constraint Inference for Inverse Reinforcement Learning |
d67856178 | Modern neural networks are over-parametrized. In particular, each rectified linear hidden unit can be modified by a multiplicative factor by adjusting input and output weights, without changing the rest of the network. Inspired by the Sinkhorn-Knopp algorithm, we introduce a fast iterative method for minimizing the 2 norm of the weights, equivalently the weight decay regularizer. It provably converges to a unique solution. Interleaving our algorithm with SGD during training improves the test accuracy. For small batches, our approach offers an alternative to batchand group-normalization on CIFAR-10 and ImageNet with a ResNet-18. | EQUI-NORMALIZATION OF NEURAL NETWORKS |
d204960684 | We introduce the Convolutional Conditional Neural Process (CONVCNP), a new member of the Neural Process family that models translation equivariance in the data. Translation equivariance is an important inductive bias for many learning problems including time series modelling, spatial data, and images. The model embeds data sets into an infinite-dimensional function space as opposed to a finitedimensional vector space. To formalize this notion, we extend the theory of neural representations of sets to include functional representations, and demonstrate that any translation-equivariant embedding can be represented using a convolutional deep set. We evaluate CONVCNPs in several settings, demonstrating that they achieve state-of-the-art performance compared to existing NPs. We demonstrate that building in translation equivariance enables zero-shot generalization to challenging, out-of-domain tasks.way: (i) CNNs require data to live "on the grid" (e.g. image pixels form a regularly spaced grid), while many of the above domains have data that live "off the grid" (e.g. time series data may be observed irregularly at any time t ∈ R). (ii) NPs operate on partially observed context sets whereas CNNs typically do not. (iii) NPs rely on embedding sets into a finite-dimensional vector space for which the notion of equivariance with respect to input translations is not natural, as we detail in Section 3. In this work, we introduce the CONVCNP, a new member of the NP family that accounts for translation equivariance. 1 This is achieved by extending the theory of learning on sets to include functional representations, which in turn can be used to express any translation-equivariant NP model. Our key contributions can be summarized as follows. (i) We provide a representation theorem for translation-equivariant functions on sets, extending a key result of Zaheer et al. (2017) to functional embeddings, including sets of varying size.(ii) We extend the NP family of models to include translation equivariance.(iii) We evaluate the CONVCNP and demonstrate that it exhibits excellent performance on several synthetic and real-world benchmarks.BACKGROUND AND FORMAL PROBLEM STATEMENTIn this section we introduce the notation and precisely define the problem this paper addresses.Notation. In the following, let X = R d and Y ⊆ R d (with Y compact) be the spaces of inputs and outputs respectively. To ease notation, we often assume scalar outputs Y ⊆ R. Define Z M = (X × Y) M as the collection of M input-output pairs, Z ≤M = M m=1 Z M as the collection of at most M pairs, and Z = | CONVOLUTIONAL CONDITIONAL NEURAL PROCESSES |
d208268589 | Consider an imitation learning problem that the imitator and the expert have different dynamics models. Most of the current imitation learning methods fail because they focus on imitating actions. We propose a novel state alignment based imitation learning method to train the imitator to follow the state sequences in expert demonstrations as much as possible. The state alignment comes from both local and global perspectives and we combine them into a reinforcement learning framework by a regularized policy update objective. We show the superiority of our method on standard imitation learning settings and imitation learning settings where the expert and imitator have different dynamics models. | STATE ALIGNMENT-BASED IMITATION LEARNING |
d252280667 | Variational Autoencoder (VAE)-based generative models offer flexible representation learning by incorporating meta-priors, general premises considered beneficial for downstream tasks. However, the incorporated meta-priors often involve ad-hoc model deviations from the original likelihood architecture, causing undesirable changes in their training. In this paper, we propose a novel representation learning method, Gromov-Wasserstein Autoencoders (GWAE), which directly matches the latent and data distributions using the variational autoencoding scheme. Instead of likelihood-based objectives, GWAE models minimize the Gromov-Wasserstein (GW) metric between the trainable prior and given data distributions. The GW metric measures the distance structure-oriented discrepancy between distributions even with different dimensionalities, which provides a direct measure between the latent and data spaces. By restricting the prior family, we can introduce meta-priors into the latent space without changing their objective. The empirical comparisons with VAE-based models show that GWAE models work in two prominent meta-priors, disentanglement and clustering, with their GW objective unchanged. arXiv:2209.07007v2 [cs.LG] 24 Feb 2023 • We conduct empirical evaluations on the capability of GWAE in prominent meta-priors: disentanglement and clustering. Several experiments on image datasets CelebA (Liu et al., 2015), MNIST (LeCun et al., 1998), and 3D Shapes (Burgess & Kim, 2018), show that GWAE models outperform the VAE-based representation learning methods whereas their GW objective is not changed over different meta-priors. 2 RELATED WORK VAE-based Representation Learning. VAE (Kingma & Welling, 2014) is a prominent deep generative model for representation learning. Following its theoretical consistency and explicit handling of latent variables, many state-of-the-art representation learning methods are proposed based on VAE | GROMOV-WASSERSTEIN AUTOENCODERS |
d245131182 | Automatic speech recognition systems have created exciting possibilities for applications, however they also enable opportunities for systematic eavesdropping. We propose a method to camouflage a person's voice over-the-air from these systems without inconveniencing the conversation between people in the room. Standard adversarial attacks are not effective in real-time streaming situations because the characteristics of the signal will have changed by the time the attack is executed. We introduce predictive attacks, which achieve real-time performance by forecasting the attack that will be the most effective in the future. Under real-time constraints, our method jams the established speech recognition system Deep-Speech 3.9x more than baselines as measured through word error rate, and 6.6x more as measured through character error rate. We furthermore demonstrate our approach is practically effective in realistic environments over physical distances.Published as a conference paper at ICLR 2022 is spoken, and general to mean applicable to the majority of vocabulary in a language. Existing prior work has successfully tackled at least one of these three requirements, but none all three. While some work is real-time(Chen et al., 2020;, these disruptions can be filtered out as they are constrained to specific frequency ranges. Universal attacks(Lu et al., 2021)can be similarly subtracted.Gong et al. (2019)achieved both real-time and robust obstructions, but are limited to a predefined set of ten words.Streaming audio is a particularly demanding domain to disrupt because the calculation needs to be performed in real-time. By the time a sound is computed, time will have passed and the streaming signal will have changed, making standard generative methods obsolete. The sampling rate of audio is at least 16 kHz, meaning the corruption for a given input must be estimated and played over a speaker within milliseconds, which is currently infeasible. Additionally, when attacks are played over-the-air, the attack needs to be loud enough to disrupt any rogue microphone that could be far away. The attack sound needs to carry the same distance as the voice.We introduce predictive attacks, which are able to disrupt any word that automatic speech recognition models are trained to transcribe. Our approach achieves real-time performance by forecasting an attack on the future of the signal, conditioned on two seconds of input speech. Our attack is optimized to have a volume similar to normal background noise, allowing people in a room to converse naturally and without monitoring from an automatic speech recognition system.Forecasting with deep neural networks has already been successfully used in other domains to achieve real-time performance, for instance in packet loss concealment(Pascual et al., 2021). In this paper, we demonstrate how and why this approach lends itself particularly well to developing general, robust and real-time attacks for automatic speech recognition models. Our experiments show that predictive attacks are able to largely disrupt the established DeepSpeech (Amodei et al., 2016) recognition system which was trained on the LibriSpeech dataset(Panayotov et al., 2015). On the standard, large-scale dataset LibriSpeech, our approach causes at least a three fold increase in word error rate over baselines, and at least a six fold increase in character error rate.Our method is practical and straightforward to implement on commodity hardware. We additionally demonstrate the method works inside real-world rooms with natural ambient noise and complex scene geometries. We call our method Neural Voice Camouflage. | REAL-TIME NEURAL VOICE CAMOUFLAGE |
d258480276 | Driven by large-data pre-training, Segment Anything Model (SAM) has been demonstrated as a powerful promptable framework, revolutionizing the segmentation field.Despite the generality, customizing SAM for specific visual concepts without man-powered prompting is under-explored, e.g., automatically segmenting your pet dog in numerous images.In this paper, we introduce a training-free Personalization approach for SAM, termed PerSAM.Given only one-shot data, i.e., a single image with a reference mask, we first obtain a positive-negative location prior for the target concept in new images.Then, aided by target visual semantics, we empower SAM for personalized object segmentation via two proposed techniques: target-guided attention and target-semantic prompting.In this way, we can effectively customize the general-purpose SAM for private use without any training.To further alleviate the ambiguity of segmentation scales, we present an efficient one-shot fine-tuning variant, PerSAM-F.Freezing the entire SAM, we introduce a scale-aware fine-tuning to aggregate multi-scale masks, which only tunes 2 parameters within 10 seconds for improved performance.To demonstrate our efficacy, we construct a new dataset, PerSeg, for the evaluation of personalized object segmentation, and also test our methods on various one-shot image and video segmentation benchmarks.Besides, we propose to leverage PerSAM to improve DreamBooth for personalized text-to-image synthesis.By mitigating the disturbance of training-set backgrounds, our approach showcases better target appearance generation and higher fidelity to the input text prompt.Code is released | PERSONALIZE SEGMENT ANYTHING MODEL WITH ONE SHOT |
d3484550 | Hierarchical reinforcement learning methods offer a powerful means of planning flexible behavior in complicated domains. However, learning an appropriate hierarchical decomposition of a domain into subtasks remains a substantial challenge. We present a novel algorithm for subtask discovery, based on the recently introduced multitask linearly-solvable Markov decision process (MLMDP) framework. The MLMDP can perform never-before-seen tasks by representing them as a linear combination of a previously learned basis set of tasks. In this setting, the subtask discovery problem can naturally be posed as finding an optimal low-rank approximation of the set of tasks the agent will face in a domain. We use non-negative matrix factorization to discover this minimal basis set of tasks, and show that the technique learns intuitive decompositions in a variety of domains. Our method has several qualitatively desirable features: it is not limited to learning subtasks with single goal states, instead learning distributed patterns of preferred states; it learns qualitatively different hierarchical decompositions in the same domain depending on the ensemble of tasks the agent will face; and it may be straightforwardly iterated to obtain deeper hierarchical decompositions. | Hierarchical Subtask Discovery With Non-Negative Matrix Factorization |
d2721941 | Deep generative models have been successfully used to learn representations for high-dimensional discrete spaces by representing discrete objects as sequences and employing powerful sequence-based deep models. Unfortunately, these sequencebased models often produce invalid sequences: sequences which do not represent any underlying discrete structure; invalid sequences hinder the utility of such models. As a step towards solving this problem, we propose to learn a deep recurrent validator model, which can estimate whether a partial sequence can function as the beginning of a full, valid sequence. This validator provides insight as to how individual sequence elements influence the validity of the overall sequence, and can be used to constrain sequence based models to generate valid sequences -and thus faithfully model discrete objects. Our approach is inspired by reinforcement learning, where an oracle which can evaluate validity of complete sequences provides a sparse reward signal. We demonstrate its effectiveness as a generative model of Python 3 source code for mathematical expressions, and in improving the ability of a variational autoencoder trained on SMILES strings to decode valid molecular structures. | LEARNING A GENERATIVE MODEL FOR VALIDITY IN COMPLEX DISCRETE STRUCTURES |
d203641746 | Recent theoretical work has established connections between over-parametrized neural networks and linearized models governed by the Neural Tangent Kernels (NTKs). NTK theory leads to concrete convergence and generalization results, yet the empirical performance of neural networks are observed to exceed their linearized models, suggesting insufficiency of this theory.Towards closing this gap, we investigate the training of over-parametrized neural networks that are beyond the NTK regime yet still governed by the Taylor expansion of the network. We bring forward the idea of randomizing the neural networks, which allows them to escape their NTK and couple with quadratic models. We show that the optimization landscape of randomized two-layer networks are nice and amenable to escaping-saddle algorithms. We prove concrete generalization and expressivity results on these randomized networks, which leads to sample complexity bounds (of learning certain simple functions) that match the NTK and can in addition be better by a dimension factor when mild distributional assumptions are present. We demonstrate that our randomization technique can be generalized systematically beyond the quadratic case, by using it to find networks that are coupled with higher-order terms in their Taylor series. * Salesforce Research. yu.bai@salesforce.com † Princeton University. jasonlee@princeton.edu design. arXiv preprint arXiv:1711.00501, 2017. . Characterizing implicit bias in terms of optimization geometry. arXiv preprint arXiv:1802.08246, 2018a.Suriya Gunasekar, Jason D Lee, Daniel Soudry, and Nati Srebro. Implicit bias of gradient descent on linear convolutional networks. In Advances in Neural Information Processing Systems, pages 9461-9471, 2018b.Benjamin D Haeffele and René Vidal. Global optimality in tensor factorization, deep learning, and beyond. arXiv preprint arXiv:1506.07540, 2015.Moritz Hardt and Tengyu Ma. Identity matters in deep learning. arXiv preprint arXiv:1611.04231, 2016. . Kernel and deep regimes in overparametrized models. arXiv preprint arXiv:1906.05827, 2019.Gilad Yehudai and Ohad Shamir. On the power and limitations of random features for understanding neural networks. arXiv preprint arXiv:1904.00687, 2019. gradient descent optimizes over-parameterized deep ReLU networks. arXiv preprint arXiv:1811.08888, 2018. | Beyond Linearization: On Quadratic and Higher-Order Approximation of Wide Neural Networks |
d11277821 | The Wasserstein distance received a lot of attention recently in the community of machine learning, especially for its principled way of comparing distributions. It has found numerous applications in several hard problems, such as domain adaptation, dimensionality reduction or generative models. However, its use is still limited by a heavy computational cost. Our goal is to alleviate this problem by providing an approximation mechanism that allows to break its inherent complexity. It relies on the search of an embedding where the Euclidean distance mimics the Wasserstein distance. We show that such an embedding can be found with a siamese architecture associated with a decoder network that allows to move from the embedding space back to the original input space. Once this embedding has been found, computing optimization problems in the Wasserstein space (e.g. barycenters, principal directions or even archetypes) can be conducted extremely fast. Numerical experiments supporting this idea are conducted on image datasets, and show the wide potential benefits of our method. * All three authors contributed equally. | Learning Wasserstein Embeddings |
d261100669 | In this paper, we present a novel defense against backdoor attacks on deep neural networks (DNNs), wherein adversaries covertly implant malicious behaviors (backdoors) into DNNs.Our defense falls within the category of post-development defenses that operate independently of how the model was generated.Our proposed defense is built upon an intriguing concept: given a backdoored model, we reverse engineer it to directly extract its backdoor functionality to a backdoor expert model.To accomplish this, we finetune the backdoored model over a small set of intentionally mislabeled clean samples, such that it unlearns the normal functionality while still preserving the backdoor functionality, and thus resulting in a model (dubbed a backdoor expert model) that can only recognize backdoor inputs.Based on the extracted backdoor expert model, we show the feasibility of devising robust backdoor input detectors that filter out the backdoor inputs during model inference.Further augmented by an ensemble strategy with a finetuned auxiliary model, our defense, BaDExpert (Backdoor Input Detection with Backdoor Expert), effectively mitigates 17 SOTA backdoor attacks while minimally impacting clean utility.The effectiveness of BaDExpert has been verified on multiple datasets (CIFAR10, GTSRB, and ImageNet) across multiple model architectures (ResNet, VGG, | BADEXPERT: EXTRACTING BACKDOOR FUNCTIONALITY FOR ACCURATE BACKDOOR INPUT DETECTION |
d238198747 | We prove that Fp sketch, a well-celebrated streaming algorithm for frequency moments estimation, is differentially private as is when p ∈ (0, 1]. Fp sketch uses only polylogarithmic space, exponentially better than existing DP baselines and only worse than the optimal non-private baseline by a logarithmic factor. The evaluation shows that Fp sketch can achieve reasonable accuracy with differential privacy guarantee. | Differentially Private Fractional Frequency Moments Estimation with Polylogarithmic Space |
d252355382 | The task of building general agents that perform well over a wide range of tasks has been an important goal in reinforcement learning since its inception. The problem has been subject of research of a large body of work, with performance frequently measured by observing scores over the wide range of environments contained in the Atari 57 benchmark. Agent57 was the first agent to surpass the human benchmark on all 57 games, but this came at the cost of poor data-efficiency, requiring nearly 80 billion frames of experience to achieve. Taking Agent57 as a starting point, we employ a diverse set of strategies to achieve a 200-fold reduction of experience needed to outperform the human baseline. We investigate a range of instabilities and bottlenecks we encountered while reducing the data regime, and propose effective solutions to build a more robust and efficient agent. We also demonstrate competitive performance with high-performing methods such as Muesli and MuZero. The four key components to our approach are (1) an approximate trust region method which enables stable bootstrapping from the online network, (2) a normalisation scheme for the loss and priorities which improves robustness when learning a set of value functions with a wide range of scales, (3) an improved architecture employing techniques from NFNets in order to leverage deeper networks without the need for normalization layers, and (4) a policy distillation method which serves to smooth out the instantaneous greedy policy over time. | Human-level Atari 200x faster |
d222177165 | Prior AI breakthroughs in complex games have focused on either the purely adversarial or purely cooperative settings. In contrast, Diplomacy is a game of shifting alliances that involves both cooperation and competition. For this reason, Diplomacy has proven to be a formidable research challenge. In this paper we describe an agent for the no-press variant of Diplomacy that combines supervised learning on human data with one-step lookahead search via regret minimization. Regret minimization techniques have been behind previous AI successes in adversarial games, most notably poker, but have not previously been shown to be successful in large-scale games involving cooperation. We show that our agent greatly exceeds the performance of past no-press Diplomacy bots, is unexploitable by expert humans, and ranks in the top 2% of human players when playing anonymous games on a popular Diplomacy website. * Equal Contribution.arXiv:2010.02923v2 [cs.AI] 3 May 2021Published as a conference paper at ICLR 2021 importantly, that our agent ranks in the top 2% of human players when playing anonymous games on a popular Diplomacy website.Published as a conference paper at ICLR 2021 Sergiu Hart and Andreu Mas-Colell. A simple adaptive procedure leading to correlated equilibrium.Econometrica, 68(5):1127-1150, 2000.Johannes Heinrich and David Silver. Deep reinforcement learning from self-play in imperfectinformation games. arXiv preprint arXiv:1603.01121, 2016.Ralf Herbrich, Tom Minka, and Thore Graepel. Trueskill™: a bayesian skill rating system. In Advances in neural information processing systems, pp. 569-576, 2007.Ronald A Howard. Dynamic programming and markov processes. 1960.Junling Hu and Michael P Wellman. Nash q-learning for general-sum stochastic games. Journal of machine learning research, 4(Nov):sampling for regret minimization in extensive games. In Advances in neural information processing systems, pp. | HUMAN-LEVEL PERFORMANCE IN NO-PRESS DIPLOMACY VIA EQUILIBRIUM SEARCH |
d249152222 | We study offline reinforcement learning (RL) in partially observable Markov decision processes. In particular, we aim to learn an optimal policy from a dataset collected by a behavior policy which possibly depends on the latent state. Such a dataset is confounded in the sense that the latent state simultaneously affects the action and the observation, which is prohibitive for existing offline RL algorithms. To this end, we propose the Proxy variable Pessimistic Policy Optimization (P3O) algorithm, which addresses the confounding bias and the distributional shift between the optimal and behavior policies in the context of general function approximation. At the core of P3O is a coupled sequence of pessimistic confidence regions constructed via proximal causal inference, which is formulated as minimax estimation. Under a partial coverage assumption on the confounded dataset, we prove that P3O achieves a n −1/2 -suboptimality, where n is the number of trajectories in the dataset. To our best knowledge, P3O is the first provably efficient offline RL algorithm for POMDPs with a confounded dataset. | Pessimism in the Face of Confounders: Provably Efficient Offline Reinforcement Learning in Partially Observable Markov Decision Processes |
d247011290 | When transferring a pretrained model to a downstream task, two popular methods are full fine-tuning (updating all the model parameters) and linear probing (updating only the last linear layer-the "head"). It is well known that fine-tuning leads to better accuracy in-distribution (ID). However, in this paper, we find that fine-tuning can achieve worse accuracy than linear probing out-of-distribution (OOD) when the pretrained features are good and the distribution shift is large. On 10 distribution shift datasets (Breeds-Living17, Breeds-Entity30, DomainNet, CIFAR → STL, CIFAR10.1, FMoW, ImageNetV2, ImageNet-R, ImageNet-A, ImageNet-Sketch), fine-tuning obtains on average 2% higher accuracy ID but 7% lower accuracy OOD than linear probing. We show theoretically that this tradeoff between ID and OOD accuracy arises even in a simple setting: fine-tuning overparameterized two-layer linear networks. We prove that the OOD error of fine-tuning is high when we initialize with a fixed or random head-this is because while fine-tuning learns the head, the lower layers of the neural network change simultaneously and distort the pretrained features. Our analysis suggests that the easy two-step strategy of linear probing then full fine-tuning (LP-FT), sometimes used as a fine-tuning heuristic, combines the benefits of both fine-tuning and linear probing. Empirically, LP-FT outperforms both fine-tuning and linear probing on the above datasets (1% better ID, 10% better OOD than full fine-tuning). . Smart: Robust and efficient fine-tuning for pre-trained natural language models through principled regularized optimization. | Fine-Tuning can Distort Pretrained Features and Underperform Out-of-Distribution |
d13046179 | We consider the two related problems of detecting if an example is misclassified or out-of-distribution. We present a simple baseline that utilizes probabilities from softmax distributions. Correctly classified examples tend to have greater maximum softmax probabilities than erroneously classified and out-of-distribution examples, allowing for their detection. We assess performance by defining several tasks in computer vision, natural language processing, and automatic speech recognition, showing the effectiveness of this baseline across all. We then show the baseline can sometimes be surpassed, demonstrating the room for future research on these underexplored detection tasks.Published as a conference paper at ICLR 2017 one method which outperforms the baseline on some (but not all) tasks. This new method evaluates the quality of a neural network's input reconstruction to determine if an example is abnormal. | A BASELINE FOR DETECTING MISCLASSIFIED AND OUT-OF-DISTRIBUTION EXAMPLES IN NEURAL NETWORKS |
d254877510 | Fine-tuning pre-trained language models has become the prevalent paradigm for building downstream NLP models.Oftentimes fine-tuned models are readily available but their training data is not, due to data privacy or intellectual property concerns.This creates a barrier to fusing knowledge across individual models to yield a better single model.In this paper, we study the problem of merging individual models built on different training data sets to obtain a single model that performs well both across all data set domains and can generalize on out-ofdomain data.We propose a dataless knowledge fusion method that merges models in their parameter space, guided by weights that minimize prediction differences between the merged model and the individual models.Over a battery of evaluation settings, we show that the proposed method significantly outperforms baselines such as Fisher-weighted averaging or model ensembling.Further, we find that our method is a promising alternative to multi-task learning that can preserve or sometimes improve over the individual models without access to the training data.Finally, model merging is more efficient than training a multi-task model, thus making it applicable to a wider set of scenarios. 1 | DATALESS KNOWLEDGE FUSION BY MERGING WEIGHTS OF LANGUAGE MODELS |
d13298214 | Learning to navigate in complex environments with dynamic elements is an important milestone in developing AI agents. In this work we formulate the navigation question as a reinforcement learning problem and show that data efficiency and task performance can be dramatically improved by relying on additional auxiliary tasks leveraging multimodal sensory inputs. In particular we consider jointly learning the goal-driven reinforcement learning problem with auxiliary depth prediction and loop closure classification tasks. This approach can learn to navigate from raw sensory input in complicated 3D mazes, approaching human-level performance even under conditions where the goal location changes frequently. We provide detailed analysis of the agent behaviour 1 , its ability to localise, and its network activity dynamics, showing that the agent implicitly learns key navigation abilities. | LEARNING TO NAVIGATE IN COMPLEX ENVIRONMENTS |
d1257772 | Most existing machine learning classifiers are highly vulnerable to adversarial examples. An adversarial example is a sample of input data which has been modified very slightly in a way that is intended to cause a machine learning classifier to misclassify it. In many cases, these modifications can be so subtle that a human observer does not even notice the modification at all, yet the classifier still makes a mistake. Adversarial examples pose security concerns because they could be used to perform an attack on machine learning systems, even if the adversary has no access to the underlying model. Up to now, all previous work have assumed a threat model in which the adversary can feed data directly into the machine learning classifier. This is not always the case for systems operating in the physical world, for example those which are using signals from cameras and other sensors as an input. This paper shows that even in such physical world scenarios, machine learning systems are vulnerable to adversarial examples. We demonstrate this by feeding adversarial images obtained from cell-phone camera to an ImageNet Inception classifier and measuring the classification accuracy of the system. We find that a large fraction of adversarial examples are classified incorrectly even when perceived through the camera. | ADVERSARIAL EXAMPLES IN THE PHYSICAL WORLD |
d247778377 | Multiobjective combinatorial optimization (MOCO) problems can be found in many real-world applications. However, exactly solving these problems would be very challenging, particularly when they are NP-hard. Many handcrafted heuristic methods have been proposed to tackle different MOCO problems over the past decades. In this work, we generalize the idea of neural combinatorial optimization, and develop a learning-based approach to approximate the whole Pareto set for a given MOCO problem without further search procedure. We propose a single preference-conditioned model to directly generate approximate Pareto solutions for any trade-off preference, and design an efficient multiobjective reinforcement learning algorithm to train this model. Our proposed method can be treated as a learning-based extension for the widely-used decomposition-based multiobjective evolutionary algorithm (MOEA/D). It uses a single model to accommodate all the possible preferences, whereas other methods use a finite number of solutions to approximate the Pareto set. Experimental results show that our proposed method significantly outperforms some other methods on the multiobjective traveling salesman problem, multiobjective vehicle routing problem, and multiobjective knapsack problem in terms of solution quality, speed, and model efficiency. | PARETO SET LEARNING FOR NEURAL MULTI-OBJECTIVE COMBINATORIAL OPTIMIZATION |
d259274820 | Representation learning has been evolving from traditional supervised training to Contrastive Learning (CL) and Masked Image Modeling (MIM). Previous works have demonstrated their pros and cons in specific scenarios, i.e., CL and supervised pre-training excel at capturing longer-range global patterns and enabling better feature discrimination, while MIM can introduce more local and diverse attention across all transformer layers. In this paper, we explore how to obtain a model that combines their strengths. We start by examining previous feature distillation and mask feature reconstruction methods and identify their limitations. We find that their increasing diversity mainly derives from the asymmetric designs, but these designs may in turn compromise the discrimination ability. In order to better obtain both discrimination and diversity, we propose a simple but effective Hybrid Distillation strategy, which utilizes both the supervised/CL teacher and the MIM teacher to jointly guide the student model. Hybrid Distill imitates the token relations of the MIM teacher to alleviate attention collapse, as well as distills the feature maps of the supervised/CL teacher to enable discrimination. Furthermore, a progressive redundant token masking strategy is also utilized to reduce the distilling costs and avoid falling into local optima. Experiment results prove that Hybrid Distill can achieve superior performance on different benchmarks.IntroductionPre-training followed by fine-tuning has been a common paradigm for computer vision tasks since the advent of deep learning. In the past decade, supervised image classification [16, 10, 24] over the widely used ImageNet [32] has dominated the pretraining mode. Recently, self-supervised learning has emerged as a promising alternative, particularly with two approaches: Contrastive Learning (CL) and Masked Image Modeling (MIM). The former one, typical representatives are MoCo[14]and SimCLR [4], learns invariant representation for positive views, which are usually defined as different augmentations of the same image. Furthermore, CLIP [30] extends CL to a multi-modal manner, which utilizes the corresponding text description of the given image as positive pairs. While the latter, including MAE [13] and SimMIM[44], aims to reconstruct the masked image patches and has become mainstream due to its efficiency brought by mask operations.The different pre-training paradigms of CL and MIM facilitate a series of studies[43,27,38] that aim at understanding their respective properties. These studies point out that CL pre-training behaves more similar to supervised pre-training, i.e., it provides models with longer-range global patterns targeting object shape, particularly in the last few layers[27], and enables feature representation with better discrimination. However, as shown inFig. 1(a), CL pre-training causes self-attention in the last few layers to collapse into homogeneity, with attention distances located within a very small distance range. In contrast, MIM pre-training can bring more diverse attention and evenly distributed Preprint. Under review. | Hybrid Distillation: Connecting Masked Autoencoders with Contrastive Learners |
d265037895 | Generative Adversarial Networks (GANs) are one of the most popular tools for learning complex high dimensional distributions. However, generalization properties of GANs have not been well understood. In this paper, we analyze the generalization of GANs in practical settings. We show that discriminators trained on discrete datasets with the original GAN loss have poor generalization capability and do not approximate the theoretically optimal discriminator. We propose a zero-centered gradient penalty for improving the generalization of the discriminator by pushing it toward the optimal discriminator. The penalty guarantees the generalization and convergence of GANs. Experiments on synthetic and large scale datasets verify our theoretical analysis.Published as a conference paper at ICLR 2019 2. We show that the original GAN objective encourages gradient exploding in the discriminator. Gradient exploding in the discriminator can lead to mode collapse in the generator.3. We propose a zero-centered gradient penalty (0-GP) for improving the generalization capability of the discriminator. We show that non-zero centered GP and the zero-centered GP proposed inMescheder et al. (2018)cannot make the discriminator generalize. Our 0-GP helps GANs to converge to generalizable equilibria. Theoretical results are verified on real world datasets.4. We show that 0-GP helps the discriminator to distribute its capacity more equally between regions of the space, effectively preventing mode collapse. Experiments on synthetic and real world datasets verify that 0-GP can prevent mode collapse. GANs with 0-GP is much more robust to changes in hyper parameters, optimizers, and network architectures than the original GAN and GANs with other gradient penalties. | IMPROVING GENERALIZATION AND STABILITY OF GENERATIVE ADVERSARIAL NETWORKS |
d56895453 | Learning when to communicate and doing that effectively is essential in multi-agent tasks. Recent works show that continuous communication allows efficient training with back-propagation in multiagent scenarios, but have been restricted to fullycooperative tasks. In this paper, we present Individualized Controlled Continuous Communication Model (IC3Net) which has better training efficiency than simple continuous communication model, and can be applied to semi-cooperative and competitive settings along with the cooperative settings. IC3Net controls continuous communication with a gating mechanism and uses individualized rewards for each agent to gain better performance and scalability while fixing credit assignment issues. Using variety of tasks including StarCraft BroodWars TM explore and combat scenarios, we show that our network yields improved performance and convergence rates than the baselines as the scale increases. Our results convey that IC3Net agents learn when to communicate based on the scenario and profitability.Published as a conference paper at ICLR 2019Teaching agents how to communicate makes it is unnecessary to hand code the communication protocol with expert knowledge(Sukhbaatar et al., 2016)(Kottur et al., 2017. While the content of communication is important, it is also important to know when to communicate either to increase scalability and performance or to increase competitive edge. For example, a prey needs to learn when to communicate to avoid communicating its location with predators.Sukhbaatar et al. (2016)showed that agents communicating through a continuous vector are easier to train and have a higher information throughput than communication based on discrete symbols. Their continuous communication is differentiable, so it can be trained efficiently with back-propagation. However, their model assumes full-cooperation between agents and uses average global rewards. This restricts the model from being used in mixed or competitive scenarios as full-cooperation involves sharing hidden states to everyone; exposing everything and leading to poor performance by all agents as shown by our results. Furthermore, the average global reward for all agents makes the credit assignment problem even harder and difficult to scale as agents don't know their individual contributions in mixed or competitive scenarios where they want themselves to succeed before others.To solve above mentioned issues, we make the following contributions: | LEARNING WHEN TO COMMUNICATE AT SCALE IN MULTIAGENT COOPERATIVE AND COMPETITIVE TASKS |
d252917667 | A powerful framework for studying graphs is to consider them as geometric graphs: nodes are randomly sampled from an underlying metric space, and any pair of nodes is connected if their distance is less than a specified neighborhood radius. Currently, the literature mostly focuses on uniform sampling and constant neighborhood radius. However, real-world graphs are likely to be better represented by a model in which the sampling density and the neighborhood radius can both vary over the latent space. For instance, in a social network communities can be modeled as densely sampled areas, and hubs as nodes with larger neighborhood radius. In this work, we first perform a rigorous mathematical analysis of this (more general) class of models, including derivations of the resulting graph shift operators. The key insight is that graph shift operators should be corrected in order to avoid potential distortions introduced by the non-uniform sampling. Then, we develop methods to estimate the unknown sampling density in a self-supervised fashion. Finally, we present exemplary applications in which the learnt density is used to 1) correct the graph shift operator and improve performance on a variety of tasks, 2) improve pooling, and 3) extract knowledge from networks. Our experimental findings support our theory and provide strong evidence for our model. | Unveiling the Sampling Density in Non-Uniform Geometric Graphs |
d257985378 | Pruning neural networks before training has received increasing interest due to its potential to reduce training time and memory. One popular method is to prune the connections based on a certain metric, but it is not entirely clear what metric is the best choice. Recent advances in neural tangent kernel (NTK) theory suggest that the training dynamics of large enough neural networks is closely related to the spectrum of the NTK. Motivated by this finding, we propose to prune the connections that have the least influence on the spectrum of the NTK. This method can help maintain the NTK spectrum, which may help align the training dynamics to that of its dense counterpart. However, one possible issue is that the fixedweight-NTK corresponding to a given initial point can be very different from the NTK corresponding to later iterates during the training phase. We further propose to sample multiple realizations of random weights to estimate the NTK spectrum. Note that our approach is weight-agnostic, which is different from most existing methods that are weight-dependent. In addition, we use random inputs to compute the fixed-weight-NTK, making our method data-agnostic as well. We name our foresight pruning algorithm Neural Tangent Kernel Spectrum-Aware Pruning (NTK-SAP). Empirically, our method achieves better performance than all baselines on multiple datasets. Our code is available at https://github. com/YiteWang/NTK-SAP. Zhang et al. (2021) and the surveys Sun et al. (2020b); Sun (2020). Among them, one line of research uses neural tangent kernel (NTK) (Jacot et al., 2018) to describe the gradient descent dynamics of DNNs when the network size is large enough. More specifically, for large enough DNNs, the NTK is asymptotically constant during training, and the convergence behavior can be characterized by the spectrum of the NTK. This theory indicates that the spectrum of the NTK might be a reasonable metric for the whole training dynamics instead of just a few initial iterations. It is then natural to consider the following conceptual pruning method: prune the connections that have the least impact on the NTK spectrum.There are a few questions on implementing this conceptual pruning method. First, what metric to compute? Computing the whole eigenspectrum of the NTK is too timeconsuming. Following the practice in numerical linear algebra and deep learning (Lee et al., 2019a; Xiao et al., 2020), we use the nuclear norm (sum of eigenvalues) as a scalar indicator of the spectrum.Second, what "NTK" matrix to pick? We call the NTK matrix defined for the given architecture with a random initialization as a "fixed-weight-NTK", and use "analytic NTK" (Jacot et al., 2018) to refer to the asymptotic limit of the fixed-weight-NTK as the network width goes to infinity. The analytic NTK is the one studied in NTK theory (Jacot et al., 2018), and we think its spectrum may serve as a performance indicator of a certain architecture throughout the whole training process. 1 However, computing its nuclear norm is still too time-consuming (either using the analytic form given in Jacot et al. (2018) or handling an ultra-wide network). The nuclear norm of a fixed-weight-NTK is easy to compute, but the fixed-weight-NTK may be quite different from the analytic NTK. To resolve this issue, we notice a less-mentioned fact: the analytic NTK is also the limit of the expectation (over random weights) of fixed-weight-NTK 2 , and thus it can be approximated by the expectation of fixedweight-NTK for a given width. The expectation of fixed-weight-NTK shall be a better approximation of analytic NTK than a single fixed-weight-NTK. Of course, to estimate the expectation, we can use a few "samples" of weight configurations and compute the average of a few fixed-weight-NTKs.One more possible issue arises: would computing, say, 100 fixed-weight-NTKs take 100 times more computation cost? We use one more computation trick to keep the computation cost low: for each mini-batch of input, we use a fresh sample of weight configuration to compute one fixed-weight-NTK (or, more precisely, its nuclear norm). This will not increase the computation cost compared to computing the fixed-weight-NTK for one weight configuration with 100 mini-batches. We call this "new-input-new-weight" (NINW) trick.We name the proposed foresight pruning algorithm Neural Tangent Kernel Spectrum-Aware Pruning (NTK-SAP). We show that NTK-SAP is competitive on multiple datasets, including CIFAR-10, CIFAR-100, Tiny-ImageNet, and ImageNet. In summary, our contributions are:• We propose a theory-motivated foresight pruning method named NTK-SAP, which prunes networks based on the spectrum of NTK. • We introduce a multi-sampling formulation which uses different weight configurations to better capture the expected behavior of pruned neural networks. A "new-input-new-weight" (NINW) trick is leveraged to reduce the computational cost, and may be of independent interest. • Empirically, we show that NTK-SAP, as a data-agnostic foresight pruning method, achieves state-of-the-art performance in multiple settings. | NTK-SAP: IMPROVING NEURAL NETWORK PRUNING BY ALIGNING TRAINING DYNAMICS |
d263909549 | Self-supervised learning has unlocked the potential of scaling up pretraining to billions of images, since annotation is unnecessary.But are we making the best use of data?How more economical can we be?In this work, we attempt to answer this question by making two contributions.First, we investigate first-person videos and introduce a "Walking Tours" dataset.These videos are high-resolution, hourslong, captured in a single uninterrupted take, depicting a large number of objects and actions with natural scene transitions.They are unlabeled and uncurated, thus realistic for self-supervision and comparable with human learning.Second, we introduce a novel self-supervised image pretraining method tailored for learning from continuous videos.Existing methods typically adapt imagebased pretraining approaches to incorporate more frames.Instead, we advocate a "tracking to learn to recognize" approach.Our method called DORA, leads to attention maps that Discover and tRAck objects over time in an end-to-end manner, using transformer cross-attention.We derive multiple views from the tracks and use them in a classical self-supervised distillation loss.Using our novel approach, a single Walking Tours video remarkably becomes a strong competitor to ImageNet for several image and video downstream tasks. | IS IMAGENET WORTH 1 VIDEO? LEARNING STRONG IMAGE ENCODERS FROM 1 LONG UNLABELLED VIDEO |
d252780361 | We present 3DiM, a diffusion model for 3D novel view synthesis, which is able to translate a single input view into consistent and sharp completions across many views. The core component of 3DiM is a pose-conditional image-to-image diffusion model, which takes a source view and its pose as inputs, and generates a novel view for a target pose as output. 3DiM can generate multiple views that are 3D consistent using a novel technique called stochastic conditioning. The output views are generated autoregressively, and during the generation of each novel view, one selects a random conditioning view from the set of available views at each denoising step. We demonstrate that stochastic conditioning significantly improves the 3D consistency of a naïve sampler for an image-to-image diffusion model, which involves conditioning on a single fixed view. We compare 3DiM to prior work on the SRN ShapeNet dataset, demonstrating that 3DiM's generated completions from a single view achieve much higher fidelity, while being approximately 3D consistent. We also introduce a new evaluation methodology, 3D consistency scoring, to measure the 3D consistency of a generated object by training a neural field on the model's output views. 3DiM is geometry free, does not rely on hyper-networks or test-time optimization for novel view synthesis, and allows a single model to easily scale to a large number of scenes.Input view 3DiM outputs conditioned on different posesFigure 1: Given a single input image on the left, 3DiM performs novel view synthesis and generates the four views on the right. We trained a single ∼471M parameter 3DiM on all of ShapeNet (without classconditioning) and sample frames with 256 steps (512 score function evaluations with classifier-free guidance). | NOVEL VIEW SYNTHESIS WITH DIFFUSION MODELS |
d259212224 | The accurate modeling of dynamics in interactive environments is critical for successful long-range prediction. Such a capability could advance Reinforcement Learning (RL) and Planning algorithms, but achieving it is challenging. Inaccuracies in model estimates can compound, resulting in increased errors over long horizons. We approach this problem from the lens of Koopman theory, where the nonlinear dynamics of the environment can be linearized in a high-dimensional latent space. This allows us to efficiently parallelize the sequential problem of long-range prediction using convolution while accounting for the agent's action at every time step. Our approach also enables stability analysis and better control over gradients through time. Taken together, these advantages result in significant improvement over the existing approaches, both in the efficiency and the accuracy of modeling dynamics over extended horizons. We also show that this model can be easily incorporated into dynamics modeling for model-based planning and model-free RL and report promising experimental results. * Correspondence: arnab.mondal@mila.quebec 2 The term observables can be misleading as it refers to the latent space in the machine learning jargon.Preprint. Under review. | Efficient Dynamics Modeling in Interactive Environments with Koopman Theory |
d253255225 | Offline reinforcement learning (RL) struggles in environments with rich and noisy inputs, where the agent only has access to a fixed dataset without environment interactions. Past works have proposed common workarounds based on the pre-training of state representations, followed by policy training. In this work, we introduce a simple, yet effective approach for learning state representations. Our method, Behavior Prior Representation (BPR), learns state representations with an easy-to-integrate objective based on behavior cloning of the dataset: we first learn a state representation by mimicking actions from the dataset, and then train a policy on top of the fixed representation, using any off-theshelf Offline RL algorithm. Theoretically, we prove that BPR carries out performance guarantees when integrated into algorithms that have either policy improvement guarantees (conservative algorithms) or produce lower bounds of the policy values (pessimistic algorithms). Empirically, we show that BPR combined with existing state-of-the-art Offline RL algorithms leads to significant improvements across several offline control benchmarks. The code is available at https://github.com/bit1029public/offline_bpr. . End to end learning for self-driving cars. CoRR, abs/1604.07316, 2016.David Brandfonbrener, Rémi Tachet des Combes, and Romain Laroche. Deep-SPIBB: Scaling up safe policy improvement for offline reinforcement learning. | BEHAVIOR PRIOR REPRESENTATION LEARNING FOR OFFLINE REINFORCEMENT LEARNING |
d252596234 | A zero-shot RL agent is an agent that can solve any RL task in a given environment, instantly with no additional planning or learning, after an initial reward-free learning phase. This marks a shift from the reward-centric RL paradigm towards "controllable" agents that can follow arbitrary instructions in an environment. Current RL agents can solve families of related tasks at best, or require planning anew for each task. Strategies for approximate zero-shot RL have been suggested using successor features (SFs) [BBQ + 18] or forward-backward (FB) representations [TO21], but testing has been limited.After clarifying the relationships between these schemes, we introduce improved losses and new SF models, and test the viability of zero-shot RL schemes systematically on tasks from the Unsupervised RL benchmark [LYL + 21]. To disentangle universal representation learning from exploration, we work in an offline setting and repeat the tests on several existing replay buffers.SFs appear to suffer from the choice of the elementary state features. SFs with Laplacian eigenfunctions do well, while SFs based on auto-encoders, inverse curiosity, transition models, low-rank transition matrix, contrastive learning, or diversity (APS), perform unconsistently. In contrast, FB representations jointly learn the elementary and successor features from a single, principled criterion. They perform best and consistently across the board, reaching 85% of supervised RL performance with a good replay buffer, in a zero-shot manner. | Does Zero-Shot Reinforcement Learning Exist? |
d245005802 | For robots operating in the real world, it is desirable to learn reusable behaviours that can effectively be transferred and adapted to numerous tasks and scenarios. We propose an approach to learn abstract motor skills from data using a hierarchical mixture latent variable model. In contrast to existing work, our method exploits a three-level hierarchy of both discrete and continuous latent variables, to capture a set of high-level behaviours while allowing for variance in how they are executed. We demonstrate in manipulation domains that the method can effectively cluster offline data into distinct, executable behaviours, while retaining the flexibility of a continuous latent variable model. The resulting skills can be transferred and fine-tuned on new tasks, unseen objects, and from state to vision-based policies, yielding better sample efficiency and asymptotic performance compared to existing skill-and imitation-based methods. We further analyse how and when the skills are most beneficial: they encourage directed exploration to cover large regions of the state space relevant to the task, making them most effective in challenging sparse-reward settings.Adrià Colomé and Carme Torras. Dimensionality reduction in learning gaussian mixture models of movement primitives for contextualized action selection and adaptation. IEEE Robotics and Automation Letters, 3(4):3922-3929, 2018.Murtaza Dalal, Deepak Pathak, and Ruslan Salakhutdinov. Accelerating robotic reinforcement learning via parameterized action primitives.cient unsupervised temporal segmentation of motion data. . Learning movement primitive libraries through probabilistic segmentation. . Learning to select and generalize striking movements in robot table tennis. | LEARNING TRANSFERABLE MOTOR SKILLS WITH HIERARCHICAL LATENT MIXTURE POLICIES |
d239016655 | A key challenge in neural architecture search (NAS) is quickly inferring the predictive performance of a broad spectrum of neural networks to discover statistically accurate and computationally efficient ones. We refer to this task as model performance inference (MPI). The current practice for efficient MPI is gradient-based methods that leverage the gradients of a network at initialization to infer its performance. However, existing gradient-based methods rely only on heuristic metrics and lack the necessary theoretical foundations to consolidate their designs. We propose GradSign, an accurate, simple, and flexible metric for model performance inference with theoretical insights. A key idea behind GradSign is a quantity Ψ to analyze the sample-wise optimization landscape of different networks. Theoretically, we show that Ψ is an upper bound for both the training and true population losses of a neural network under reasonable assumptions. However, it is computationally prohibitive to directly calculate Ψ for modern neural networks. To address this challenge, we design GradSign, an accurate and simple approximation of Ψ using the gradients of a network evaluated at a random initialization state. Evaluation on seven NAS benchmarks across three training datasets shows that GradSign generalizes well to real-world neural networks and consistently outperforms state-of-the-art gradient-based methods for MPI evaluated by Spearman's ρ and Kendall's Tau. Additionally, we have integrated GradSign into four existing NAS algorithms and show that the GradSign-assisted NAS algorithms outperform their vanilla counterparts by improving the accuracies of best-discovered networks by up to 0.3%, 1.1%, and 1.0% on three real-world tasks. Code is available at Published as a conference paper at ICLR 2022 dataset, gradient-based methods are computationally more efficient since they only require evaluating a mini-batch of gradients at initialization. However, existing gradient-based methods rely only on heuristic metrics and lack the necessary theoretical insights to consolidate their designs.In this paper, we propose GradSign, a simple yet accurate metric for MPI with theoretical foundations. GradSign is inspired by analyzing the sample-wise optimization landscape of a network. GradSign takes as inputs a mini-batch of sample-wise gradients evaluated at a random initialization point and outputs a statistical evidence of a network that highly correlates to its well-trained predictive performance measured by accuracy on the entire dataset.Prior theoretical results (Allen-Zhu et al., 2019) show that the optimization landscape of a randomly initialized network is nearly convex and semi-smooth for a sufficiently large neighborhood. To realize its potential for MPI, we generalize these results to sample-wise optimization landscapes and propose a quantity Ψ to measure the density of sample-wise local optima in the convex areas around a random initialization point. Additionally, we prove that both the training loss and generalization error of a network are proportionally upper bounded by Ψ 2 under reasonable assumptions.Based on our theoretical results, we design GradSign, an accurate and simple approximation of Ψ. Empirically, we show that GradSign can also generalize to realistic setups that may violate our assumptions. In addition, GradSign is efficient to compute and easy to implement as it uses only the sample-wise gradient information of a network at a random initialization point.Extensive evaluation of GradSign on seven NAS benchmarks (i.e., NAS-Bench-101, NAS-Bench-201, and five design spaces of NDS) across three datasets (i.e., CIFAR-10, CIFAR-100, and ImageNet16-120) shows that GradSign consistently outperforms existing gradient-based methods in all circumstances. Furthermore, we have integrated GradSign into existing NAS algorithms and show that the GradSign-assisted variants of these NAS algorithms lead to more accurate architectures.Contributions. This paper makes the following contributions:• We provide a new perspective to view the overall optimization landscape of a network as a combination of sample-wise optimization landscapes. Based on this insight, we introduce a new quantity Ψ that provides an upper bound on both the training loss and generalization error of a network under reasonable assumptions.Jonathan Frankle and Michael Carbin. The lottery ticket hypothesis: Finding sparse, trainable neural networks. arXiv preprint arXiv:1803.03635, 2018.Rong Ge, Jason D Lee, and Tengyu Ma. Learning one-hidden-layer neural networks with landscape design. arXiv preprint arXiv:1711.00501, 2017. | GRADSIGN: MODEL PERFORMANCE INFERENCE WITH THEORETICAL INSIGHTS |
d262954661 | We investigate the internal behavior of Transformer-based Large Language Models (LLMs) when they generate factually incorrect text.We propose modeling factual queries as Constraint Satisfaction Problems and use this framework to investigate how the model interacts internally with factual constraints.Specifically, we discover a strong positive relation between the model's attention to constraint tokens and the factual accuracy of its responses.In our curated suite of 11 datasets with over 40,000 prompts, we study the task of predicting factual errors with the Llama-2 family across all scales (7B, 13B, 70B).We propose SAT Probe, a method probing self-attention patterns, that can predict constraint satisfaction and factual errors, and allows early error identification.The approach and findings demonstrate how using the mechanistic understanding of factuality in LLMs can enhance reliability. 1 | ATTENTION SATISFIES: A CONSTRAINT-SATISFACTION LENS ON FACTUAL ERRORS OF LANGUAGE MODELS |
d3476101 | The vast majority of natural sensory data is temporally redundant. Video frames or audio samples which are sampled at nearby points in time tend to have similar values. Typically, deep learning algorithms take no advantage of this redundancy to reduce computation. This can be an obscene waste of energy. We present a variant on backpropagation for neural networks in which computation scales with the rate of change of the data -not the rate at which we process the data. We do this by having neurons communicate a combination of their state, and their temporal change in state. Intriguingly, this simple communication rule give rise to units that resemble biologically-inspired leaky integrate-and-fire neurons, and to a weight-update rule that is equivalent to a form of Spike-Timing Dependent Plasticity (STDP), a synaptic learning rule observed in the brain. We demonstrate that on MNIST and a temporal variant of MNIST, our algorithm performs about as well as a Multilayer Perceptron trained with backpropagation, despite only communicating discrete values between layers.IntroductionSuppose we are trying to track objects in a scene. A typical system used today would consist of sending camera-frames into a convolutional network which predicts bounding boxes. Such a system may be trained by going over many hours of video with manually annotated bounding boxes, and learning to predict their locations. This system has to execute a forward pass of a convolutional network at each iteration. If we double the frame rate, we double the amount of computation, even if the contents of the video are mostly static. Intuitively, it does not feel that this should be necessary. Given the similarity between neighbouring frames of video, could we not reuse some of the computation from the last frame to update the bounding box inferences for the current frame? Is it really necessary to recompute the entire network on each frame?Many robotic systems consist of many sensors operating at wildly different frame rates. Some "neuromorphic" sensors, such as the Dynamic Vision Sensor Lichtsteiner et al. [2008] have done away with the concept of frames altogether and instead send asynchronous "events" whenever the value of a pixel changes beyond some threshold. It's not obvious, using current methods in deep learning, how we can efficiently integrate asynchronous sensory signals into a unified, trainable, latent representation, without recomputing the function of the network every time a new signal arrives.There has been a lot of work on increasing the computational efficiency of neural networks by quantizing neural weights or activations (see Section 4), but comparatively little work on exploiting redundancies in the data to reduce the amount of computation.O'Connor and Welling [2016b], set out to exploit the temporal redundancy in video, by having neurons only send their quantized changes in activation to downstream neurons, and having the downstream neurons integrate these changes. This approach works for efficiently approximating the function of the network, but fails for training, because when the weights are changing with time, this approach (take the temporal 31st difference, multiply by weights, temporally integrate) fails to reconstruct the correct activation for the next layer. In other words, t τ =0 (x τ − x τ −1 ) · w τ = x t · w t unless w t = w 0 ∀t.Figure 2describes the problem visually. In this paper, we correct for this by instead encoding a mixture of two components of the layers activation x t : the proportional component k p x t , and the derivative component k d (x t − x t−1 ). When we invert this encoding scheme, we get get a decoding scheme which corresponds to taking an exponentially decaying temporal average of past inputs. | Temporally Efficient Deep Learning with Spikes |
d232290577 | Searching for novel molecules with desired chemical properties is crucial in drug discovery. Existing work focuses on developing neural models to generate either molecular sequences or chemical graphs. However, it remains a big challenge to find novel and diverse compounds satisfying several properties. In this paper, we propose MARS, a method for multi-objective drug molecule discovery. MARS is based on the idea of generating the chemical candidates by iteratively editing fragments of molecular graphs. To search for high-quality candidates, it employs Markov chain Monte Carlo sampling (MCMC) on molecules with an annealing scheme and an adaptive proposal. To further improve sample efficiency, MARS uses a graph neural network (GNN) to represent and select candidate edits, where the GNN is trained on-the-fly with samples from MCMC. Experiments show that MARS achieves state-of-the-art performance in various multi-objective settings where molecular bio-activity, drug-likeness, and synthesizability are considered. Remarkably, in the most challenging setting where all four objectives are simultaneously optimized, our approach outperforms previous methods significantly in comprehensive evaluations. The code is available at | MARS: MARKOV MOLECULAR SAMPLING FOR MULTI-OBJECTIVE DRUG DISCOVERY |
d43964415 | Mini-batch stochastic gradient descent (SGD) is the state of the art in large scale parallel machine learning, but its scalability is limited by a communication bottleneck. Recent work proposed local SGD, i.e. running SGD independently in parallel on different workers and averaging only once in a while. This scheme shows promising results in practice, but eluded thorough theoretical analysis.We prove concise convergence rates for local SGD on convex problems and show that it converges at the same rate as mini-batch SGD in terms of number of evaluated gradients, that is, the scheme achieves linear speed-up in the number of workers and mini-batch size. Moreover, the number of communication rounds can be reduced up to a factor of T 1/2 -where T denotes the number of total steps-compared to mini-batch SGD. | Local SGD Converges Fast and Communicates Little |
d251765117 | Textual content is often the output of a collaborative writing process: We start with an initial draft, ask for suggestions, and repeatedly make changes. Agnostic of this process, today's language models are trained to generate only the final result. As a consequence, they lack several abilities crucial for collaborative writing: They are unable to update existing texts, difficult to control and incapable of verbally planning or explaining their actions. To address these shortcomings, we introduce PEER, a collaborative language model that is trained to imitate the entire writing process itself: PEER can write drafts, add suggestions, propose edits and provide explanations for its actions. Crucially, we train multiple instances of PEER able to infill various parts of the writing process, enabling the use of selftraining techniques for increasing the quality, amount and diversity of training data. This unlocks PEER's full potential by making it applicable in domains for which no edit histories are available and improving its ability to follow instructions, to write useful comments, and to explain its actions. We show that PEER achieves strong performance across various domains and editing tasks. | PEER: A Collaborative Language Model |
d235254358 | Graph Attention Networks (GATs) are one of the most popular GNN architectures and are considered as the state-of-the-art architecture for representation learning with graphs. In GAT, every node attends to its neighbors given its own representation as the query. However, in this paper we show that GAT computes a very limited kind of attention: the ranking of the attention scores is unconditioned on the query node. We formally define this restricted kind of attention as static attention and distinguish it from a strictly more expressive dynamic attention. Because GATs use a static attention mechanism, there are simple graph problems that GAT cannot express: in a controlled problem, we show that static attention hinders GAT from even fitting the training data. To remove this limitation, we introduce a simple fix by modifying the order of operations and propose GATv2: a dynamic graph attention variant that is strictly more expressive than GAT. We perform an extensive evaluation and show that GATv2 outperforms GAT across 12 OGB and other benchmarks while we match their parametric costs. Our code is available at https://github.com/tech-srl/how_attentive_are_ gats. | HOW ATTENTIVE ARE GRAPH ATTENTION NETWORKS? |
d221470196 | Despite significant progress in challenging problems across various domains, applying state-of-the-art deep reinforcement learning (RL) algorithms remains challenging due to their sensitivity to the choice of hyperparameters. This sensitivity can partly be attributed to the non-stationarity of the RL problem, potentially requiring different hyperparameter settings at different stages of the learning process. Additionally, in the RL setting, hyperparameter optimization (HPO) requires a large number of environment interactions, hindering the transfer of the successes in RL to real-world applications. In this work, we tackle the issues of sample-efficient and dynamic HPO in RL. We propose a population-based automated RL (AutoRL) framework to meta-optimize arbitrary off-policy RL algorithms. In this framework, we optimize the hyperparameters, including architecture hyperparameters while simultaneously training the agent. By sharing the collected experience across the population, we substantially increase the sample efficiency of the meta-optimization. We demonstrate the capabilities of our sample-efficient AutoRL approach in a case study with the popular TD3 algorithm in the MuJoCo benchmark suite, where we reduce the number of environment interactions needed for meta-optimization by up to an order of magnitude compared to population-based training. arXiv Preprint J. Snoek, H. Larochelle, and R. Adams. Practical Bayesian optimization of machine learning . A hypercube-based encoding for evolving large-scale neural networks. Artificial life, 15(2):185-212, 2009. Clune. Deep neuroevolution: Genetic algorithms are a competitive alternative for training deep neural networks for reinforcement learning. arXiv preprint arXiv:1712.06567, 2017. E. Todorov, T. Erez, and Y. Tassa. Mujoco: A physics engine for model-based control. | SAMPLE-EFFICIENT AUTOMATED DEEP REINFORCEMENT LEARNING |
d248986748 | Offline reinforcement learning algorithms still lack trust in practice due to the risk that the learned policy performs worse than the original policy that generated the dataset or behaves in an unexpected way that is unfamiliar to the user. At the same time, offline RL algorithms are not able to tune their most important hyperparameter -the proximity of the learned policy to the original policy. We propose an algorithm that allows the user to tune this hyperparameter at runtime, thereby addressing both of the above mentioned issues simultaneously. This allows users to start with the original behavior and grant successively greater deviation, as well as stopping at any time when the policy deteriorates or the behavior is too far from the familiar one. | User-Interactive Offline Reinforcement Learning |
d247450626 | The problem of molecular generation has received significant attention recently. Existing methods are typically based on deep neural networks and require training on large datasets with tens of thousands of samples. In practice, however, the size of class-specific chemical datasets is usually limited (e.g., dozens of samples) due to labor-intensive experimentation and data collection. This presents a considerable challenge for the deep learning generative models to comprehensively describe the molecular design space. Another major challenge is to generate only physically synthesizable molecules. This is a non-trivial task for neural network-based generative models since the relevant chemical knowledge can only be extracted and generalized from the limited training data. In this work, we propose a data-efficient generative model that can be learned from datasets with orders of magnitude smaller sizes than common benchmarks. At the heart of this method is a learnable graph grammar that generates molecules from a sequence of production rules. Without any human assistance, these production rules are automatically constructed from training data. Furthermore, additional chemical knowledge can be incorporated in the model by further grammar optimization. Our learned graph grammar yields state-of-the-art results on generating high-quality molecules for three monomer datasets that contain only ∼20 samples each. Our approach also achieves remarkable performance in a challenging polymer generation task with only 117 training samples and is competitive against existing methods using 81k data points. Code is available at | DATA-EFFICIENT GRAPH GRAMMAR LEARNING FOR MOLECULAR GENERATION |
d264289264 | Innovations like protein diffusion have enabled significant progress in de novo protein design, which is a vital topic in life science.These methods typically depend on protein structure encoders to model residue backbone frames, where atoms do not exist.Most prior encoders rely on atom-wise features, such as angles and distances between atoms, which are not available in this context.Thus far, only several simple encoders, such as IPA(Jumper et al., 2021), have been proposed for this scenario, exposing the frame modeling as a bottleneck.In this work, we proffer the Vector Field Network (VFN), which enables network layers to perform learnable vector computations between coordinates of frame-anchored virtual atoms, thus achieving a higher capability for modeling frames.The vector computation operates in a manner similar to a linear layer, with each input channel receiving 3D virtual atom coordinates instead of scalar values.The multiple feature vectors output by the vector computation are then used to update the residue representations and virtual atom coordinates via attention aggregation.Remarkably, VFN also excels in modeling both frames and atoms, as the real atoms can be treated as the virtual atoms for modeling, positioning VFN as a potential universal encoder.In protein diffusion (frame modeling), VFN exhibits an impressive performance advantage over IPA, excelling in terms of both designability (67.04% vs. 53.58%)and diversity (66.54% vs. 51.98%).In inverse folding (frame and atom modeling), VFN outperforms the previous SoTA model, PiFold (54.7% vs. 51.66%), on sequence recovery rate.We also propose a method of equipping VFN with the ESM model(Lin et al., 2023), which significantly surpasses the previous ESM-based SoTA (62.67% vs. 55.65%),LM-Design (Zheng et al., 2023), by a substantial margin.* WM, ZS and MZ contributed equally.Work was done when WM was visiting Zhejiang University. | DE NOVO PROTEIN DESIGN USING GEOMETRIC VECTOR FIELD NETWORKS |
d244527086 | Object-centric representations are a promising path toward more systematic generalization by providing flexible abstractions upon which compositional world models can be built. Recent work on simple 2D and 3D datasets has shown that models with object-centric inductive biases can learn to segment and represent meaningful objects from the statistical structure of the data alone without the need for any supervision. However, such fully-unsupervised methods still fail to scale to diverse realistic data, despite the use of increasingly complex inductive biases such as priors for the size of objects or the 3D geometry of the scene. In this paper, we instead take a weakly-supervised approach and focus on how 1) using the temporal dynamics of video data in the form of optical flow and 2) conditioning the model on simple object location cues can be used to enable segmenting and tracking objects in significantly more realistic synthetic data. We introduce a sequential extension to Slot Attention which we train to predict optical flow for realistic looking synthetic scenes and show that conditioning the initial state of this model on a small set of hints, such as center of mass of objects in the first frame, is sufficient to significantly improve instance segmentation. These benefits generalize beyond the training distribution to novel objects, novel backgrounds, and to longer video sequences. We also find that such initial-state-conditioning can be used during inference as a flexible interface to query the model for specific objects or parts of objects, which could pave the way for a range of weakly-supervised approaches and allow more effective interaction with trained models. | CONDITIONAL OBJECT-CENTRIC LEARNING FROM VIDEO |
d258762142 | Subject-driven text-to-image generation aims to generate customized images of the given subject based on the text descriptions, which has drawn increasing attention recently. Existing methods mainly resort to finetuning a pretrained generative model, where the identity-relevant information and the identity-irrelevant information are entangled in the latent embedding space. However, the highly entangled latent embedding may lead to the failure of subject-driven text-to-image generation as follows: (i) the identity-irrelevant information hidden in the entangled embedding may dominate the generation process, resulting in the generated images heavily dependent on the irrelevant information while ignoring the given text descriptions; (ii) the identity-relevant information carried in the entangled embedding can not be appropriately preserved, resulting in identity change of the subject in the generated images. To tackle the problems, we propose Disen-Booth, an identity-preserving disentangled tuning framework for subject-driven text-to-image generation in this paper. Specifically, DisenBooth finetunes the pretrained diffusion model in the denoising process. Different from previous works that utilize an entangled embedding to denoise each image, DisenBooth instead utilizes disentangled embeddings to respectively preserve the subject identity and capture the identity-irrelevant information. We further design the novel weak denoising and contrastive embedding auxiliary tuning objectives to achieve the disentanglement. Extensive experiments show that our proposed DisenBooth framework outperforms baseline models for subject-driven text-to-image generation with the identity-preserved embedding. Additionally, by combining the identity-preserved embedding and identity-irrelevant embedding, DisenBooth demonstrates more generation flexibility and controllability 1 . | DisenBooth: Identity-Preserving Disentangled Tuning for Subject-Driven Text-to-Image Generation |
d46898034 | To optimize a neural network one often thinks of optimizing its parameters, but it is ultimately a matter of optimizing the function that maps inputs to outputs. Since a change in the parameters might serve as a poor proxy for the change in the function, it is of some concern that primacy is given to parameters but that the correspondence has not been tested. Here, we show that it is simple and computationally feasible to calculate distances between functions in a L 2 Hilbert space. We examine how typical networks behave in this space, and compare how parameter 2 distances compare to function L 2 distances between various points of an optimization trajectory. We find that the two distances are nontrivially related. In particular, the L 2 / 2 ratio decreases throughout optimization, reaching a steady value around when test error plateaus. We then investigate how the L 2 distance could be applied directly to optimization. We first propose that in multitask learning, one can avoid catastrophic forgetting by directly limiting how much the input/output function changes between tasks. Secondly, we propose a new learning rule that constrains the distance a network can travel through L 2 -space in any one update. This allows new examples to be learned in a way that minimally interferes with what has previously been learned. These applications demonstrate how one can measure and regularize function distances directly, without relying on parameters or local approximations like loss curvature. | MEASURING AND REGULARIZING NETWORKS IN FUNCTION SPACE |
d203593355 | Training neural networks to be certifiably robust is a powerful defense against adversarial attacks. However, while promising, state-of-the-art results with certified training are far from satisfactory. Currently, it is very difficult to train a neural network that is both accurate and certified on realistic datasets and specifications (e.g., robustness). Given this difficulty, a pressing existential question is: given a dataset and a specification, is there a network that is both certified and accurate with respect to these? While the evidence suggests "no", we prove that for realistic datasets and specifications, such a network does exist and its certification can be established by propagating lower and upper bounds of each neuron through the network (interval analysis) -the most relaxed yet computationally efficient convex relaxation. Our result can be seen as a Universal Approximation Theorem for interval-certified ReLU networks. To the best of our knowledge, this is the first work to prove the existence of accurate, interval-certified networks. | UNIVERSAL APPROXIMATION WITH CERTIFIED NETWORKS |
d108296236 | Euclidean embeddings of data are fundamentally limited in their ability to capture latent semantic structures, which need not conform to Euclidean spatial assumptions. Here we consider an alternative, which embeds data as discrete probability distributions in a Wasserstein space, endowed with an optimal transport metric. Wasserstein spaces are much larger and more flexible than Euclidean spaces, in that they can successfully embed a wider variety of metric structures. We exploit this flexibility by learning an embedding that captures semantic information in the Wasserstein distance between embedded distributions. We examine empirically the representational capacity of our learned Wasserstein embeddings, showing that they can embed a wide variety of metric structures with smaller distortion than an equivalent Euclidean embedding. We also investigate an application to word embedding, demonstrating a unique advantage of Wasserstein embeddings: We can visualize the high-dimensional embedding directly, since it is a probability distribution on a low-dimensional space. This obviates the need for dimensionality reduction techniques like t-SNE for visualization.arXiv:1905.03329v1 [cs.LG] 8 May 2019Published as a conference paper at ICLR 2019 such, they make attractive targets for embeddings in machine learning, where this flexibility might capture complex relationships between objects when other embeddings fail to do so. | LEARNING EMBEDDINGS INTO ENTROPIC WASSERSTEIN SPACES |
d246995268 | Simulation-based inference (SBI) refers to statistical inference on stochastic models for which we can generate samples, but not compute likelihoods. Like SBI algorithms, generative adversarial networks (GANs) do not require explicit likelihoods. We study the relationship between SBI and GANs, and introduce GATSBI, an adversarial approach to SBI. GATSBI reformulates the variational objective in an adversarial setting to learn implicit posterior distributions. Inference with GATSBI is amortised across observations, works in high-dimensional posterior spaces and supports implicit priors. We evaluate GATSBI on two SBI benchmark problems and on two high-dimensional simulators. On a model for wave propagation on the surface of a shallow water body, we show that GATSBI can return well-calibrated posterior estimates even in high dimensions. On a model of camera optics, it infers a high-dimensional posterior given an implicit prior, and performs better than a state-of-the-art SBI approach. We also show how GATSBI can be extended to perform sequential posterior estimation to focus on individual observations. Overall, GATSBI opens up opportunities for leveraging advances in GANs to perform Bayesian inference on high-dimensional simulation-based models. | GATSBI: GENERATIVE ADVERSARIAL TRAINING FOR SIMULATION-BASED INFERENCE |
d258059971 | We present a method that enables synthesizing novel views and novel poses of arbitrary human performers from sparse multi-view images. A key ingredient of our method is a hybrid appearance blending module that combines the advantages of the implicit body NeRF representation and image-based rendering. Existing generalizable human NeRF methods that are conditioned on the body model have shown robustness against the geometric variation of arbitrary human performers. Yet they often exhibit blurry results when generalized onto unseen identities. Meanwhile, image-based rendering shows high-quality results when sufficient observations are available, whereas it suffers artifacts in sparse-view settings. We propose Neural Image-based Avatars (NIA) that exploits the best of those two methods: to maintain robustness under new articulations and self-occlusions while directly leveraging the available (sparse) source view colors to preserve appearance details of new subject identities. Our hybrid design outperforms recent methods on both in-domain identity generalization as well as challenging cross-dataset generalization settings. Also, in terms of the pose generalization, our method outperforms even the per-subject optimized animatable NeRF methods. The video results are available at https://youngjoongunc.github.io/nia. | NEURAL IMAGE-BASED AVATARS: GENERALIZABLE RADIANCE FIELDS FOR HUMAN AVATAR MODELING |
d264435719 | This paper tackles two related questions at the heart of machine learning; how can we predict if a minimum will generalize to the test set, and why does stochastic gradient descent find minima that generalize well? Our work is inspired by Zhang et al. (2016), who showed deep networks can easily memorize randomly labeled training data, despite generalizing well when shown real labels of the same inputs. We show here that the same phenomenon occurs in small linear models. These observations are explained by evaluating the Bayesian evidence, which penalizes sharp minima but is invariant to model parameterization. We also explore the "generalization gap" between small and large batch training, identifying an optimum batch size which maximizes the test set accuracy. Interpreting stochastic gradient descent as a stochastic differential equation, we identify a "noise scale" g = ( N B − 1) ≈ N/B, where is the learning rate, N training set size and B batch size. Consequently the optimum batch size is proportional to the learning rate and the training set size, B opt ∝ N . We verify these predictions empirically. * Work done as a member of the Google Brain Residency Program (g.co/brainresidency) | A BAYESIAN PERSPECTIVE ON GENERALIZATION AND STOCHASTIC GRADIENT DESCENT |
d245704504 | Guided image synthesis enables everyday users to create and edit photo-realistic images with minimum effort. The key challenge is balancing faithfulness to the user inputs (e.g., hand-drawn colored strokes) and realism of the synthesized images. Existing GAN-based methods attempt to achieve such balance using either conditional GANs or GAN inversions, which are challenging and often require additional training data or loss functions for individual applications. To address these issues, we introduce a new image synthesis and editing method, Stochastic Differential Editing (SDEdit), based on a diffusion model generative prior, which synthesizes realistic images by iteratively denoising through a stochastic differential equation (SDE). Given an input image with user guide in a form of manipulating RGB pixels, SDEdit first adds noise to the input, then subsequently denoises the resulting image through the SDE prior to increase its realism. SDEdit does not require task-specific training or inversions and can naturally achieve the balance between realism and faithfulness. SDEdit outperforms state-of-the-art GAN-based methods by up to 98.09% on realism and 91.72% on overall satisfaction scores, according to a human perception study, on multiple tasks, including stroke-based image synthesis and editing as well as image compositing. | SDEDIT: GUIDED IMAGE SYNTHESIS AND EDITING WITH STOCHASTIC DIFFERENTIAL EQUATIONS |
d219792208 | We solve the problem of 6-DoF localisation and 3D dense reconstruction in spatial environments as approximate Bayesian inference in a deep generative approach which combines learned with engineered models. This principled treatment of uncertainty and probabilistic inference overcomes the shortcoming of current state-of-the-art solutions to rely on heavily engineered, heterogeneous pipelines. Variational inference enables us to use neural networks for system identification, while a differentiable raycaster is used for the emission model. This ensures that our model is amenable to end-to-end gradient-based optimisation. We evaluate our approach on realistic unmanned aerial vehicle flight data, nearing the performance of a state-of-the-art visual inertial odometry system. The applicability of the learned model to downstream tasks such as generative prediction and planning is investigated. | Variational State-Space Models for Localisation and Dense 3D Mapping in 6 DoF |
d252185491 | Cross-silo Federated learning (FL) has become a promising tool in machine learning applications for healthcare. It allows hospitals/institutions to train models with sufficient data while the data is kept private. To make sure the FL model is robust when facing heterogeneous data among FL clients, most efforts focus on personalizing models for clients. However, the latent relationships between clients' data are ignored. In this work, we focus on a special non-iid FL problem, called Domain-mixed FL, where each client's data distribution is assumed to be a mixture of several predefined domains. Recognizing the diversity of domains and the similarity within domains, we propose a novel method, FedDAR, which learns a domain shared representation and domain-wise personalized prediction heads in a decoupled manner. For simplified linear regression settings, we have theoretically proved that FedDAR enjoys a linear convergence rate. For general settings, we have performed intensive empirical studies on both synthetic and real-world medical datasets which demonstrate its superiority over prior FL methods. * Equal contribution. Preprint. Under review. arXiv:2209.04007v1 [cs.LG] 8 Sep 2022FedAvg [25,49,26, 31]. As an ideal model that can perform well on all clients may not exist, it requires FL algorithms to personalize the model for different data distributions.Prior theoretical work [36] shows that it is impossible to improve performances on all clients without making assumptions about the client's data distributions. Past works on personalized FL methods [36,49,16,35,11] make their own assumptions and tailor their methods to those assumptions. In this paper, we propose a new and more realistic assumption where each client's data distribution is a mixture of several predefined domains. We call our problem setting Domain-mixed FL. It is inspired by the fact that the diversity of the medical data can be attributed to some known concept of domains, such as different demographic/ethnic groups of patients[52,46,44], different manufacturers or protocols/workflows of image scanners [37,5], and so on. Despite of the domain shifts between data domains, same domain at different clients are usually considered to have the same distribution. The data heterogeneity between FL clients actually comes from the distinct mixtures of diverse domains at clients. Furthermore, it is necessary to address the ubiquitous issue of domain shifts in healthcare data. For instance, different ethic groups could have significant differences in disease patterns and treatment responses[52,46,44]. In addition, one ethic group could be a majority in one location/client, but a minority in another location/client; the mixture or composition of ethnicity could be different in local study cohorts. To reduce the potential bias in the FL model, we need put ethniciy related domain-wise personalization into our algorithm design. Real world applications like this motivate us to personalize model for each domain instead of client.FedEM[36] and FedMinMax[45] makes similar assumption on data distribution as ours. However, FedEM assumes the domains are unknown and tries to learn a linear combination of several shared component models with personalized mixture weights through an EM-like algorithm. FedMinMax doesn't acknowledge the domain shift between domains and still aims to learn one shared model across domains by adapting minmax optimization to FL setting .Our Contributions. We formulate the proposed problem setting, Domain-mixed FL. Through our analysis, we find prior FL methods, both generic FL methods like FedAvg [38], and personalized FL methods like FedRep [6], are sub-optimal under our setting. To address this issue, we propose a new algorithm, Federated Domain-Aware Representation Learning (FedDAR). FedDAR learns a shared model for all the clients but embedded with domain-wise personalized modules. The model contains two parts: an shared encoder across all domains and a multi-headed predictor whose heads are associated with domains. For an input from one specific domain, the model extracts representation via the shared encoder and then use the corresponding head to make the prediction. FedDAR decouples the learning of the encoder and heads by alternating between the updates of the encoder and the heads. It allows the clients to run many local updates on the heads without overfitting on domains with limited data samples. This also leads to faster convergence and better performed model. FedDAR also adapts different aggregation strategies for the two parts. We use a weighted average operation to aggregate the local updates for the encoder. With additional sample re-weighting, the overall training objective is equally weighted for each domain to encourage the fairness among domains. While for the heads, we propose a novel second-order aggregation algorithm to improve the optimality of aggregated heads.We theoretically show our method enjoys nice properties like linear convergence and small sample complexity in a linear case. Through extensive experiments on both synthetic and real-world datasets, we demonstrate that FedDAR significantly improves performance over the state-of-the-art personalized FL methods. To the best of our knowledge, our paper is among the first efforts in domain-wise personalized federated learning that achieve such superior performance.Related workBesides the literature we have discussed above, other works on personalization and fairness in federated learning are also closely related to our work.Personalized Federated Learning. Personalized federated learning has been studied from a variety of perspectives: i) local fine-tuning[58,61]ii) meta-learning[2,13,25, 28] iii) local/global model interpolation[11,7,35]. iv) clustered FL that partition clients into clusters and learn optimal model for each cluster[49,35,16]. v) Multi-Task Learning(MTL) [55, 51, 64] [19, 18, 53, 22, 30] vi) local representations or heads for clients[1,34,6]. vii) personalized model through hypernetwork or super | FedDAR: Federated Domain-Aware Representation Learning |
d220265858 | Neural network scaling has been critical for improving the model quality in many real-world machine learning applications with vast amounts of training data and compute. Although this trend of scaling is affirmed to be a sure-fire approach for better model quality, there are challenges on the path such as the computation cost, ease of programming, and efficient implementation on parallel devices. GShard is a module composed of a set of lightweight annotation APIs and an extension to the XLA compiler. It provides an elegant way to express a wide range of parallel computation patterns with minimal changes to the existing model code. GShard enabled us to scale up multilingual neural machine translation Transformer model with Sparsely-Gated Mixture-of-Experts beyond 600 billion parameters using automatic sharding. We demonstrate that such a giant model can efficienctly be trained on 2048 TPU v3 accelerators in 4 days to achieve far superior quality for translation from 100 languages to English compared to the prior art.Preprint. Under review. | GShard: Scaling Giant Models with Conditional Computation and Automatic Sharding |
d263152829 | Large language models (LLMs) can "lie", which we define as outputting false statements despite "knowing" the truth in a demonstrable sense.LLMs might "lie", for example, when instructed to output misinformation.Here, we develop a simple lie detector that requires neither access to the LLM's activations (black-box) nor ground-truth knowledge of the fact in question.The detector works by asking a predefined set of unrelated follow-up questions after a suspected lie, and feeding the LLM's yes/no answers into a logistic regression classifier.Despite its simplicity, this lie detector is highly accurate and surprisingly general.When trained on examples from a single setting-prompting GPT-3.5 to lie about factual questionsthe detector generalises out-of-distribution to (1) other LLM architectures, (2) LLMs fine-tuned to lie, (3) sycophantic lies, and (4) lies emerging in real-life scenarios such as sales.These results indicate that LLMs have distinctive lierelated behavioural patterns, consistent across architectures and contexts, which could enable general-purpose lie detection. | HOW TO CATCH AN AI LIAR: LIE DETECTION IN BLACK-BOX LLMS BY ASKING UNRELATED QUESTIONS |
d3047732 | Most contemporary multi-task learning methods assume linear models. This setting is considered shallow in the era of deep learning. In this paper, we present a new deep multi-task representation learning framework that learns cross-task sharing structure at every layer in a deep network. Our approach is based on generalising the matrix factorisation techniques explicitly or implicitly used by many conventional MTL algorithms to tensor factorisation, to realise automatic learning of end-to-end knowledge sharing in deep networks. This is in contrast to existing deep learning approaches that need a user-defined multi-task sharing strategy. Our approach applies to both homogeneous and heterogeneous MTL, as well as multi-domain learning (MDL). Experiments demonstrate the efficacy of our deep multi-task representation learning in terms of both higher accuracy and fewer design choices. | Deep Multi-task Representation Learning: A Tensor Factorisation Approach |
d233474778 | 3D convolution is powerful for video classification but often computationally expensive, recent studies mainly focus on decomposing it on spatial-temporal and/or channel dimensions. Unfortunately, most approaches fail to achieve a preferable balance between convolutional efficiency and feature-interaction sufficiency. For this reason, we propose a concise and novel Channel Tensorization Network (CT-Net), by treating the channel dimension of input feature as a multiplication of K sub-dimensions. On one hand, it naturally factorizes convolution in a multiple dimension way, leading to a light computation burden. On the other hand, it can effectively enhance feature interaction from different channels, and progressively enlarge the 3D receptive field of such interaction to boost classification accuracy. Furthermore, we equip our CT-Module with a Tensor Excitation (TE) mechanism. It can learn to exploit spatial, temporal and channel attention in a high-dimensional manner, to improve the cooperative power of all the feature dimensions in our CT-Module. Finally, we flexibly adapt ResNet as our CT-Net. Extensive experiments are conducted on several challenging video benchmarks, e.g., Kinetics-400, Something-Something V1 and V2. Our CT-Net outperforms a number of recent SOTA approaches, in terms of accuracy and/or efficiency. The codes and models will be available on https://github.com | CT-NET: CHANNEL TENSORIZATION NETWORK FOR VIDEO CLASSIFICATION |
d247778993 | Text-based games (TBG) have emerged as promising environments for driving research in grounded language understanding and studying problems like generalization and sample efficiency. Several deep reinforcement learning (RL) methods with varying architectures and learning schemes have been proposed for TBGs. However, these methods fail to generalize efficiently, especially under distributional shifts. In a departure from deep RL approaches, in this paper, we propose a general method inspired by case-based reasoning to train agents and generalize out of the training distribution. The case-based reasoner collects instances of positive experiences from the agent's interaction with the world in the past and later reuses the collected experiences to act efficiently. The method can be applied in conjunction with any existing on-policy neural agent in the literature for TBGs. Our experiments show that the proposed approach consistently improves existing methods, obtains good out-of-distribution generalization, and achieves new state-of-the-art results on widely used environments. | CASE-BASED REASONING FOR BETTER GENERALIZATION IN TEXTUAL REINFORCEMENT LEARNING |
d208526932 | We study the phenomenon that some modules of deep neural networks (DNNs) are more critical than others. Meaning that rewinding their parameter values back to initialization, while keeping other modules fixed at the trained parameters, results in a large drop in the network's performance. Our analysis reveals interesting properties of the loss landscape which leads us to propose a complexity measure, called module criticality, based on the shape of the valleys that connect the initial and final values of the module parameters. We formulate how generalization relates to the module criticality, and show that this measure is able to explain the superior generalization performance of some architectures over others, whereas, earlier measures fail to do so. | The intriguing role of module criticality in the generalization of deep networks |
d246822526 | Deployment efficiency is an important criterion for many real-world applications of reinforcement learning (RL). Despite the community's increasing interest, there lacks a formal theoretical formulation for the problem. In this paper, we propose such a formulation for deployment-efficient RL (DE-RL) from an "optimization with constraints" perspective: we are interested in exploring an MDP and obtaining a near-optimal policy within minimal deployment complexity, whereas in each deployment the policy can sample a large batch of data. Using finite-horizon linear MDPs as a concrete structural model, we reveal the fundamental limit in achieving deployment efficiency by establishing information-theoretic lower bounds, and provide algorithms that achieve the optimal deployment efficiency. Moreover, our formulation for DE-RL is flexible and can serve as a building block for other practically relevant settings; we give "Safe DE-RL" and "Sample-Efficient DE-RL" as two examples, which may be worth future investigation.arXiv:2202.06450v3 [cs.LG] 31 Aug 2022Published as a conference paper at ICLR 2022To provide a more quantitative understanding, we instantiate our DE-RL framework in finite-horizon linear MDPs 1 (Jin et al., 2019) and develop the essential theory. The main questions we address are: Q1: What is the optimum of the deployment efficiency in our DE-RL setting? Q2: Can we achieve the optimal deployment efficiency in our DE-RL setting?PRELIMINARIESNotation Throughout our paper, for n P Z`, we will denote rns " t1, 2, ..., nu. r¨s denotes the ceiling function. Unless otherwise specified, for vector x P R d and matrix X P R dˆd , }x} denotes the vector l 2 -norm of x and }X} denotes the largest singular value of X. We will use standard big-oh notations Op¨q, Ωp¨q, Θp¨q, and notations such as r Op¨q to suppress logarithmic factors.EPISODIC REINFORCEMENT LEARNINGWe consider an episodic Markov Decision Process denoted by M pS, A, H, P, rq, where S is the state space, A is the finite action space, H is the horizon length, and P " tP h u H h"1 and r " tr h u H h"1 denote the transition and the reward functions. At the beginning of each episode, the environment will sample an initial state s 1 from the initial state distribution d 1 . Then, for each time step h P rHs, the agent selects an action a h P A, interacts with the environment, receives a reward r h ps h , a h q, and transitions to the next state s h`1 . The episode will terminate once s H`1 is reached.A (Markov) policy π h p¨q at step h is a function mapping from S Ñ ∆pAq, where ∆pAq denotes the probability simplex over the action space. With a slight abuse of notation, when π h p¨q is a deterministic policy, we will assume π h p¨q : S Ñ A. A full (Markov) policy π " tπ 1 , π 2 , ..., π H u specifies such a mapping for each time step. We use V π h psq and Q π h ps, aq to denote the value function Abdellah Bennane et al. Adaptive educational software by applying reinforcement learning. sub-sampling for reinforcement learning with general function approximation. ArXiv, abs/2106.07203, 2021. | TOWARDS DEPLOYMENT-EFFICIENT REINFORCEMENT LEARNING: LOWER BOUND AND OPTIMALITY |
d202888876 | Value-based methods constitute a fundamental methodology in planning and deep reinforcement learning (RL). In this paper, we propose to exploit the underlying structures of the state-action value function, i.e., Q function, for both planning and deep RL. In particular, if the underlying system dynamics lead to some global structures of the Q function, one should be capable of inferring the function better by leveraging such structures. Specifically, we investigate the lowrank structure, which widely exists for big data matrices. We verify empirically the existence of low-rank Q functions in the context of control and deep RL tasks (Atari games). As our key contribution, by leveraging Matrix Estimation (ME) techniques, we propose a general framework to exploit the underlying low-rank structure in Q functions, leading to a more efficient planning procedure for classical control, and additionally, a simple scheme that can be applied to any value-based RL techniques to consistently achieve better performance on "low-rank" tasks. Extensive experiments on control tasks and Atari games confirm the efficacy of our approach. * Equal contribution 1 arXiv:1909.12255v1 [cs.LG] 26 Sep 2019Emmanuel Abbe and Colin Sandon. Community detection in general stochastic block models:Fundamental limits and efficient algorithms for recovery. . Automated synthesis of low-rank control systems from sc-ltl specifications using tensor-train decompositions. | HARNESSING STRUCTURES FOR VALUE-BASED PLANNING AND REINFORCEMENT LEARNING |
d14711954 | We describe a framework for multitask deep reinforcement learning guided by policy sketches. Sketches annotate tasks with sequences of named subtasks, providing information about high-level structural relationships among tasks but not how to implement them-specifically not providing the detailed guidance used by much previous work on learning policy abstractions for RL (e.g. intermediate rewards, subtask completion signals, or intrinsic motivations). To learn from sketches, we present a model that associates every subtask with a modular subpolicy, and jointly maximizes reward over full task-specific policies by tying parameters across shared subpolicies. Optimization is accomplished via a decoupled actor-critic training objective that facilitates learning common behaviors from multiple dissimilar reward functions. We evaluate the effectiveness of our approach in three environments featuring both discrete and continuous control, and with sparse rewards that can be obtained only after completing a number of high-level subgoals. Experiments show that using our approach to learn policies guided by sketches gives better performance than existing techniques for learning task-specific or shared policies, while naturally inducing a library of interpretable primitive behaviors that can be recombined to rapidly adapt to new tasks. | Modular Multitask Reinforcement Learning with Policy Sketches |
d90258365 | As 3D scanning solutions become increasingly popular, several deep learning setups have been developed geared towards that task of scan completion, i.e., plausibly filling in regions there were missed in the raw scans. These methods, however, largely rely on supervision in the form of paired training data, i.e., partial scans with corresponding desired completed scans. While these methods have been successfully demonstrated on synthetic data, the approaches cannot be directly used on real scans in absence of suitable paired training data. We develop a first approach that works directly on input point clouds, does not require paired training data, and hence can directly be applied to real scans for scan completion. We evaluate the approach qualitatively on several real-world datasets (ScanNet, Matterport, KITTI), quantitatively on 3D-EPN shape completion benchmark dataset, and demonstrate realistic completions under varying levels of incompleteness. ScanNet raw scan Matterport raw scan KITTI raw scan 3D EPN test scan arXiv:1904.00069v1 [cs.CV] | Unpaired Point Cloud Completion on Real Scans using Adversarial Training |
d247849778 | The performance of existing text style transfer models is severely limited by the non-parallel datasets on which the models are trained. In non-parallel datasets, no direct mapping exists between sentences of the source and target style; the style transfer models thus only receive weak supervision of the target sentences during training, which often leads the model to discard too much style-independent information, or utterly fail to transfer the style. In this work, we propose LaMer, a novel text style transfer framework based on large-scale language models. LaMer first mines the roughly parallel expressions in the non-parallel datasets with scene graphs, and then employs MLE training, followed by imitation learning refinement, to leverage the intrinsic parallelism within the data. On two benchmark tasks (sentiment & formality transfer) and a newly proposed challenging task (political stance transfer), our model achieves qualitative advances in transfer accuracy, content preservation, and fluency. Further empirical and human evaluations demonstrate that our model not only makes training more efficient, but also generates more readable and diverse expressions than previous models. | NON-PARALLEL TEXT STYLE TRANSFER WITH SELF-PARALLEL SUPERVISION |
d232257645 | Voice style transfer, also called voice conversion, seeks to modify one speaker's voice to generate speech as if it came from another (target) speaker.Previous works have made progress on voice conversion with parallel training data and pre-known speakers.However, zero-shot voice style transfer, which learns from non-parallel data and generates voices for previously unseen speakers, remains a challenging problem.We propose a novel zero-shot voice transfer method via disentangled representation learning.The proposed method first encodes speakerrelated style and voice content of each input voice into separated low-dimensional embedding spaces, and then transfers to a new voice by combining the source content embedding and target style embedding through a decoder.With informationtheoretic guidance, the style and content embedding spaces are representative and (ideally) independent of each other.On real-world VCTK datasets, our method outperforms other baselines and obtains state-of-the-art results in terms of transfer accuracy and voice naturalness for voice style transfer experiments under both many-to-many and zero-shot setups. | IMPROVING ZERO-SHOT VOICE STYLE TRANSFER VIA DISENTANGLED REPRESENTATION LEARNING |
d12130431 | This paper presents incremental network quantization (INQ), a novel method, targeting to efficiently convert any pre-trained full-precision convolutional neural network (CNN) model into a low-precision version whose weights are constrained to be either powers of two or zero. Unlike existing methods which are struggled in noticeable accuracy loss, our INQ has the potential to resolve this issue, as benefiting from two innovations. On one hand, we introduce three interdependent operations, namely weight partition, group-wise quantization and re-training. A wellproven measure is employed to divide the weights in each layer of a pre-trained CNN model into two disjoint groups. The weights in the first group are responsible to form a low-precision base, thus they are quantized by a variable-length encoding method. The weights in the other group are responsible to compensate for the accuracy loss from the quantization, thus they are the ones to be re-trained. On the other hand, these three operations are repeated on the latest re-trained group in an iterative manner until all the weights are converted into low-precision ones, acting as an incremental network quantization and accuracy enhancement procedure. Extensive experiments on the ImageNet classification task using almost all known deep CNN architectures including AlexNet, VGG-16, GoogleNet and ResNets well testify the efficacy of the proposed method. Specifically, at 5-bit quantization (a variable-length encoding: 1 bit for representing zero value, and the remaining 4 bits represent at most 16 different values for the powers of two) 1 , our models have improved accuracy than the 32-bit floating-point references. Taking ResNet-18 as an example, we further show that our quantized models with 4-bit, 3-bit and 2-bit ternary weights have improved or very similar accuracy against its 32-bit floating-point baseline. Besides, impressive results with the combination of network pruning and INQ are also reported. We believe that our method sheds new insights on how to make deep CNNs to be applicable on mobile or embedded devices. The code is available at https://github.com | INCREMENTAL NETWORK QUANTIZATION: TOWARDS LOSSLESS CNNS WITH LOW-PRECISION WEIGHTS |
d3462549 | Deep autoregressive models have shown state-of-the-art performance in density estimation for natural images on large-scale datasets such as ImageNet. However, such models require many thousands of gradient-based weight updates and unique image examples for training. Ideally, the models would rapidly learn visual concepts from only a handful of examples, similar to the manner in which humans learns across many vision tasks. In this paper, we show how 1) neural attention and 2) meta learning techniques can be used in combination with autoregressive models to enable effective few-shot density estimation. Our proposed modifications to PixelCNN result in state-of-the art few-shot density estimation on the Omniglot dataset. Furthermore, we visualize the learned attention policy and find that it learns intuitive algorithms for simple tasks such as image mirroring on ImageNet and handwriting on Omniglot without supervision. Finally, we extend the model to natural images and demonstrate few-shot image generation on the Stanford Online Products dataset. | FEW-SHOT AUTOREGRESSIVE DENSITY ESTIMATION: TOWARDS LEARNING TO LEARN DISTRIBUTIONS |
d204824061 | State of the art sequence-to-sequence models for large scale tasks perform a fixed number of computations for each input sequence regardless of whether it is easy or hard to process. In this paper, we train Transformer models which can make output predictions at different stages of the network and we investigate different ways to predict how much computation is required for a particular sequence. Unlike dynamic computation in Universal Transformers, which applies the same set of layers iteratively, we apply different layers at every step to adjust both the amount of computation as well as the model capacity. On IWSLT German-English translation our approach matches the accuracy of a well tuned baseline Transformer while using less than a quarter of the decoder layers. * Work done during an internship at Facebook AI Research. | DEPTH-ADAPTIVE TRANSFORMER |
d238419267 | Many practical problems need the output of a machine learning model to satisfy a set of constraints, K. There are, however, no known guarantees that classical neural networks can exactly encode constraints while simultaneously achieving universality. We provide a quantitative constrained universal approximation theorem which guarantees that for any convex or non-convex compact set K and any continuous function f : R n → K, there is a probabilistic transformerF whose randomized outputs all lie in K and whose expected output uniformly approximates f . Our second main result is a "deep neural version" of Berge (1963)'s Maximum Theorem. The result guarantees that given an objective function L, a constraint set K, and a family of soft constraint sets, there is a probabilistic transformerF that approximately minimizes L and whose outputs belong to K; moreover,F approximately satisfies the soft constraints. Our results imply the first universal approximation theorem for classical transformers with exact convex constraint satisfaction, and a chart-free universal approximation theorem for Riemannian manifold-valued functions subject to geodesically-convex constraints. | UNIVERSAL APPROXIMATION UNDER CONSTRAINTS IS POSSIBLE WITH TRANSFORMERS |
d88517649 | VariationalAutoencoders (VAEs) provide a theoretically-backed framework for deep generative models. However, they often produce "blurry" images, which is linked to their training objective. Sampling in the most popular implementation, the Gaussian VAE, can be interpreted as simply injecting noise to the input of a deterministic decoder. In practice, this simply enforces a smooth latent space structure. We challenge the adoption of the full VAE framework on this specific point in favor of a simpler, deterministic one. Specifically, we investigate how substituting stochasticity with other explicit and implicit regularization schemes can lead to a meaningful latent space without having to force it to conform to an arbitrarily chosen prior. To retrieve a generative mechanism for sampling new data points, we propose to employ an efficient ex-post density estimation step that can be readily adopted both for the proposed deterministic autoencoders as well as to improve sample quality of existing VAEs. We show in a rigorous empirical study that regularized deterministic autoencoding achieves state-of-the-art sample quality on the common MNIST, CIFAR-10 and CelebA datasets. | From Variational to Deterministic Autoencoders |
d258741298 | By design, large language models (LLMs) are static general-purpose models, expensive to retrain or update frequently.As they are increasingly adopted for knowledge-intensive tasks, it becomes evident that these design choices lead to failures to generate factual, relevant, and up-to-date knowledge.To this end, we propose KNOWLEDGE CARD, a modular framework to plug in new factual and relevant knowledge into general-purpose LLMs.We first introduce knowledge cards-specialized language models trained on corpora from specific domains and sources.Knowledge cards serve as parametric repositories that are selected at inference time to generate background knowledge for the base LLM.We then propose three content selectors to dynamically select and retain information in documents generated by knowledge cards, specifically controlling for relevance, brevity, and factuality of outputs.Finally, we propose two complementary integration approaches to augment the base LLM with the (relevant, factual) knowledge curated from the specialized LMs.Through extensive experiments, we demonstrate that KNOWLEDGE CARD achieves state-of-the-art performance on six benchmark datasets.Ultimately, KNOWLEDGE CARD framework enables dynamic synthesis and updates of knowledge from diverse domains.Its modularity will ensure that relevant knowledge can be continuously updated through the collective efforts of the research community. | KNOWLEDGE CARD: FILLING LLMS' KNOWLEDGE GAPS WITH PLUG-IN SPECIALIZED LANGUAGE MODELS |
d252118863 | Masked autoencoders have become popular training paradigms for self-supervised visual representation learning. These models randomly mask a portion of the input and reconstruct the masked portion according to assigned target representations. In this paper, we show that a careful choice of the target representation is unnecessary for learning good visual representation since different targets tend to derive similarly behaved models. Driven by this observation, we propose a multi-stage masked distillation pipeline and use a randomly initialized model as the teacher, enabling us to effectively train high-capacity models without any effort to carefully design the target representation. On various downstream tasks of classification, transfer learning, object detection, and semantic segmentation, the proposed method to perform masked knowledge distillation with bootstrapped teachers (dBOT) outperforms previous self-supervised methods by nontrivial margins. We hope our findings, as well as the proposed method, could motivate people to rethink the roles of target representations in pre-training masked autoencoders. The code and pre-trained models are publicly available at | Exploring Target Representations for Masked Autoencoders |
d259165410 | We revisit the problem of sampling from a target distribution that has a smooth strongly log-concave density everywhere in R p . In this context, if no additional density information is available, the randomized midpoint discretization for the kinetic Langevin diffusion is known to be the most scalable method in high dimensions with large condition numbers. Our main result is a nonasymptotic and easy to compute upper bound on the W 2 -error of this method. To provide a more thorough explanation of our method for establishing the computable upper bound, we conduct an analysis of the midpoint discretization for the vanilla Langevin process. This analysis helps to clarify the underlying principles and provides valuable insights that we use to establish an improved upper bound for the kinetic Langevin process with the midpoint discretization. Furthermore, by applying these techniques we establish new guarantees for the kinetic Langevin process with Euler discretization, which have a better dependence on the condition number than existing upper bounds.Preprint. Under review. | Langevin Monte Carlo for strongly log-concave distributions: Randomized midpoint revisited |
d238408085 | We introduce space-time graph neural network (ST-GNN), a novel GNN architecture, tailored to jointly process the underlying space-time topology of time-varying network data. The cornerstone of our proposed architecture is the composition of time and graph convolutional filters followed by pointwise nonlinear activation functions. We introduce a generic definition of convolution operators that mimic the diffusion process of signals over its underlying support. On top of this definition, we propose space-time graph convolutions that are built upon a composition of time and graph shift operators. We prove that ST-GNNs with multivariate integral Lipschitz filters are stable to small perturbations in the underlying graphs as well as small perturbations in the time domain caused by time warping. Our analysis shows that small variations in the network topology and time evolution of a system does not significantly affect the performance of ST-GNNs. Numerical experiments with decentralized control systems showcase the effectiveness and stability of the proposed ST-GNNs. | SPACE-TIME GRAPH NEURAL NETWORKS |
d264306022 | We present a novel variational framework for performing inference in (neural) stochastic differential equations (SDEs) driven by Markov-approximate fractional Brownian motion (fBM).SDEs offer a versatile tool for modeling real-world continuous-time dynamic systems with inherent noise and randomness.Combining SDEs with the powerful inference capabilities of variational methods, enables the learning of representative function distributions through stochastic gradient descent.However, conventional SDEs typically assume the underlying noise to follow a Brownian motion (BM), which hinders their ability to capture long-term dependencies.In contrast, fractional Brownian motion (fBM) extends BM to encompass non-Markovian dynamics, but existing methods for inferring fBM parameters are either computationally demanding or statistically inefficient.In this paper, building upon the Markov approximation of fBM, we derive the evidence lower bound essential for efficient variational inference of posterior path measures, drawing from the well-established field of stochastic analysis.Additionally, we provide a closed-form expression to determine optimal approximation coefficients.Furthermore, we propose the use of neural networks to learn the drift, diffusion and control terms within our variational posterior, leading to the variational training of neural-SDEs.In this framework, we also optimize the Hurst index, governing the nature of our fractional noise.Beyond validation on synthetic data, we contribute a novel architecture for variational latent video prediction,-an approach that, to the best of our knowledge, enables the first variational neural-SDE application to video perception. | VARIATIONAL INFERENCE FOR SDES DRIVEN BY FRACTIONAL NOISE |
d204907203 | We study the problem of model extraction in natural language processing, in which an adversary with only query access to a victim model attempts to reconstruct a local copy of that model. Assuming that both the adversary and victim model fine-tune a large pretrained language model such as BERT (Devlin et al., 2019), we show that the adversary does not need any real training data to successfully mount the attack. In fact, the attacker need not even use grammatical or semantically meaningful queries: we show that random sequences of words coupled with task-specific heuristics form effective queries for model extraction on a diverse set of NLP tasks including natural language inference and question answering. Our work thus highlights an exploit only made feasible by the shift towards transfer learning methods within the NLP community: for a query budget of a few hundred dollars, an attacker can extract a model that performs only slightly worse than the victim model. Finally, we study two defense strategies against model extraction-membership classification and API watermarking-which while successful against naive adversaries, are ineffective against more sophisticated ones. | THIEVES ON SESAME STREET! MODEL EXTRACTION OF BERT-BASED APIS |
d238408158 | A conventional approach to entity linking is to first find mentions in a given document and then infer their underlying entities in the knowledge base. A well-known limitation of this approach is that it requires finding mentions without knowing their entities, which is unnatural and difficult. We present a new model that does not suffer from this limitation called EntQA, which stands for Entity linking as Question Answering. EntQA first proposes candidate entities with a fast retrieval module, and then scrutinizes the document to find mentions of each candidate with a powerful reader module. Our approach combines progress in entity linking with that in open-domain question answering and capitalizes on pretrained models for dense entity retrieval and reading comprehension. Unlike in previous works, we do not rely on a mention-candidates dictionary or large-scale weak supervision. EntQA achieves strong results on the GERBIL benchmarking platform. | ENTQA: ENTITY LINKING AS QUESTION ANSWERING |
d254096162 | Interpretability is a pressing issue for machine learning. Common approaches to interpretable machine learning constrain interactions between features of the input, rendering the effects of those features on a model's output comprehensible but at the expense of model complexity. We approach interpretability from a new angle: constrain the information about the features without restricting the complexity of the model. Borrowing from information theory, we use the Distributed Information Bottleneck to find optimal compressions of each feature that maximally preserve information about the output. The learned information allocation, by feature and by feature value, provides rich opportunities for interpretation, particularly in problems with many features and complex feature interactions. The central object of analysis is not a single trained model, but rather a spectrum of models serving as approximations that leverage variable amounts of information about the inputs. Information is allocated to features by their relevance to the output, thereby solving the problem of feature selection by constructing a learned continuum of feature inclusion-to-exclusion. The optimal compression of each feature-at every stage of approximation-allows fine-grained inspection of the distinctions among feature values that are most impactful for prediction. We develop a framework for extracting insight from the spectrum of approximate models and demonstrate its utility on a range of tabular datasets. | INTERPRETABILITY WITH FULL COMPLEXITY BY CONSTRAINING FEATURE INFORMATION |
d257482484 | Deformable object manipulation (DOM) is a long-standing challenge in robotics and has attracted significant interest recently. This paper presents DaXBench, a differentiable simulation framework for DOM. While existing work often focuses on a specific type of deformable objects, DaXBench supports fluid, rope, cloth . . . ; it provides a general-purpose benchmark to evaluate widely different DOM methods, including planning, imitation learning, and reinforcement learning. DaXBench combines recent advances in deformable object simulation with JAX, a high-performance computational framework. All DOM tasks in DaXBench are wrapped with the OpenAI Gym API for easy integration with DOM algorithms. We hope that DaXBench provides to the research community a comprehensive, standardized benchmark and a valuable tool to support the development and evaluation of new DOM methods. The code and video are available online * .We benchmark eight competitive DOM methods across different algorithmic paradigms, including sampling-based planning, reinforcement learning (RL), and imitation learning (IL). For planning methods, we consider model predictive control with the Cross Entropy Method (CEM-MPC)(Richards, 2005), differentiable model predictive control(Hu et al., 2020), and a combination of † These authors contributed equally. ‡ This work is partially completed at the SEA AI Lab. * The link of the project is https://github.com/AdaCompNUS/DaXBench. Published as a conference paper at ICLR 2023 the two. For RL domains, we consider Proximal Policy Optimization (PPO) (Schulman et al., 2017) with non-differentiable dynamics, and Short-Horizon Actor-Critic (SHAC) (Xu et al., 2022) and Analytic Policy Gradients (APG) (Freeman et al., 2021) with analytical gradients. For IL, Transporter networks (Seita et al., 2021) with non-differentiable dynamics and Imitation Learning via Differentiable physics (ILD) (Chen et al., 2023) are compared. Our experiments compare algorithms with and without analytic gradients on each task, providing insights into the benefits and challenges of differentiable-physics-based DOM methods.DaXBench provides a deformable object simulator, DaX, which combines recent advances in deformable object simulation algorithms(Xu et al., 2022;Chen et al., 2023)with the high-performance computational framework JAX(Bradbury et al., 2018). This integration allows for efficient autodifferentiation and parallelization across multiple accelerators, such as multiple GPUs. All task environments are wrapped with the OpenAI Gym API(Brockman et al., 2016), enabling seamless integration with DOM algorithms for fast and easy development. By providing a comprehensive and standardized simulation framework, we aim to facilitate algorithm development and advance the state of the art in DOM. Moreover, our experimental results show that the dynamics model in DaXBench enables direct sim-to-real transfer to a real robot for rope manipulation, indicating the potential applicability of our simulation platform to real-world problems. | DAXBENCH: BENCHMARKING DEFORMABLE OBJECT MANIPULATION WITH DIFFERENTIABLE PHYSICS |
d238407774 | Message-passing neural networks (MPNNs) are the leading architecture for deep learning on graph-structured data, in large part due to their simplicity and scalability. Unfortunately, it was shown that these architectures are limited in their expressive power. This paper proposes a novel framework called Equivariant Subgraph Aggregation Networks (ESAN) to address this issue. Our main observation is that while two graphs may not be distinguishable by an MPNN, they often contain distinguishable subgraphs. Thus, we propose to represent each graph as a set of subgraphs derived by some predefined policy, and to process it using a suitable equivariant architecture. We develop novel variants of the 1-dimensional Weisfeiler-Leman (1-WL) test for graph isomorphism, and prove lower bounds on the expressiveness of ESAN in terms of these new WL variants. We further prove that our approach increases the expressive power of both MPNNs and more expressive architectures. Moreover, we provide theoretical results that describe how design choices such as the subgraph selection policy and equivariant neural architecture affect our architecture's expressive power. To deal with the increased computational cost, we propose a subgraph sampling scheme, which can be viewed as a stochastic version of our framework. A comprehensive set of experiments on real and synthetic datasets demonstrates that our framework improves the expressive power and overall performance of popular GNN architectures. * Equal contribution, authors are in alphabetical order.Figure 1: We present a provably expressive graph learning framework based on representing graphs as bags of subgraphs and processing them with an equivariant architecture composed of GNNs and set networks. Left: A pair of graphs not distinguishable by the WL test. Right: The corresponding bags (multisets) of edge-deleted subgraphs, which can be distinguished by our framework.Our approach. In an effort to devise simple, intuitive and more flexible provably expressive graph architectures, we develop a novel framework, dubbed Equivariant Subgraph Aggregation Networks (ESAN), to enhance the expressive power of existing GNNs. Our solution emerges from the observation that while two graphs may not be distinguishable by an MPNN, it may be easy to find distinguishable subgraphs. More generally, instead of encoding multisets of node colors as done in MPNNs and the WL test, we opt for encoding bags (multisets) of subgraphs and show that such an encoding can lead to a better expressive power. Following that observation, we advocate representing each graph as a bag of subgraphs chosen according to some predefined policy, e.g., all graphs that can be obtained by removing one edge from the original graph.Figure 1illustrates this idea.Bags of subgraphs are highly structured objects whose symmetry arises from both the structure of each constituent graph as well as the multiset on the whole. We propose an equivariant architecture specifically tailored to capture this object's symmetry group. Specifically, we first formulate the symmetry group for a set of graphs as the direct product of the symmetry groups for sets and graphs. We then construct a neural network comprising layers that are equivariant to this group. Motivated byMaron et al. (2020), these layers employ two base graph encoders as subroutines: The first encoder implements a Siamese network processing each subgraph independently; The second acts as an information sharing module by processing the aggregation of the subgraphs. After being processed by several such layers, a set learning module aggregates the obtained subgraph representations into an invariant representation of the original graph that is used in downstream tasks.An integral component of our method, with major impacts on its complexity and expressivity, is the subgraph selection policy: a function that maps a graph to a bag of subgraphs, which is then processed by our equivariant neural network. In this paper, we explore four simple -yet powerful -subgraph selection policies: node-deleted subgraphs, edge-deleted subgraphs, and two variants of ego-networks. To alleviate the possible computational burden, we also introduce an efficient stochastic version of our method implemented by random sampling of subgraphs according to the aforementioned policies.We provide a thorough theoretical analysis of our approach. We first prove that our architecture can implement novel and provably stronger variants of the well-known WL test, capable of encoding the multiset of subgraphs according to the base graph encoder (e.g., WL for MPNNs). Furthermore, we study how the expressive power of our architecture depends on different main design choices like the underlying base graph encoder or the subgraph selection policy. Notably, we prove that our framework can separate 3-WL indistinguishable graphs using only a 1-WL graph encoder, and that it can enhance the expressive power of stronger architectures such as PPGN (Maron et al., 2019a).We then present empirical results on a wide range of synthetic and real datasets, using several existing GNNs as base encoders. Firstly, we study the expressive power of our approach using the synthetic datasets introduced by Abboud et al. (2020) and show that it achieves perfect accuracy.Published as a conference paper at ICLR 2022 Wengong Jin, Regina Barzilay, and Tommi Jaakkola. Junction tree variational autoencoder for molecular graph generation. In ICML, 2018. -supervised classification with graph convolutional networks. In ICLR, 2017.Dmitry B Kireev. Chemnet: a novel neural network based method for graph/property mapping. | EQUIVARIANT SUBGRAPH AGGREGATION NETWORKS |
d252918265 | Self-training (ST), or pseudo-labeling has sparked significant interest in the automatic speech recognition (ASR) community recently because of its success in harnessing unlabeled data. Unlike prior semi-supervised learning approaches that relied on iteratively regenerating pseudo-labels (PLs) from a trained model and using them to train a new model, recent state-of-the-art methods perform 'continuous training' where PLs are generated using a very recent version of the model being trained. Nevertheless, these approaches still rely on bootstrapping the ST using an initial supervised learning phase where the model is trained on labeled data alone. We believe this has the potential for over-fitting to the labeled dataset in low resource settings and that ST from the start of training should reduce over-fitting. In this paper we show how we can do this by dynamically controlling the evolution of PLs during the training process in ASR. To the best of our knowledge, this is the first study that shows the feasibility of generating PLs from the very start of the training. We are able to achieve this using two techniques that avoid instabilities which lead to degenerate models that do not generalize. Firstly, we control the evolution of PLs through a curriculum that uses the online changes in PLs to control the membership of the cache of PLs and improve generalization. Secondly, we find that by sampling transcriptions from the predictive distribution, rather than only using the best transcription, we can stabilize training further. With these techniques, our ST models match prior works without an external language model. * Work done during internship at Apple. | CONTINUOUS PSEUDO-LABELING FROM THE START |
d240070972 | There has been emerging interest in using transductive learning for adversarial robustness (Goldwasser et al., NeurIPS 2020; Wu et al., ICML 2020; Wang et al., ArXiv 2021). Compared to traditional defenses, these defense mechanisms "dynamically learn" the model based on test-time input; and theoretically, attacking these defenses reduces to solving a bilevel optimization problem, which poses difficulty in crafting adaptive attacks. In this paper, we examine these defense mechanisms from a principled threat analysis perspective. We formulate and analyze threat models for transductive-learning based defenses, and point out important subtleties. We propose the principle of attacking model space for solving bilevel attack objectives, and present Greedy Model Space Attack (GMSA), an attack framework that can serve as a new baseline for evaluating transductivelearning based defenses. Through systematic evaluation, we show that GMSA, even with weak instantiations, can break previous transductive-learning based defenses, which were resilient to previous attacks, such as AutoAttack. On the positive side, we report a somewhat surprising empirical result of "transductive adversarial training": Adversarially retraining the model using fresh randomness at the test time gives a significant increase in robustness against attacks we consider.Published as a conference paper at ICLR 2022 objective is transductive as U , the attacker's output, appears in both attack (V in L a ) and defense (U in Γ). We extend this threat model to multiple rounds, which is necessary when considering DENT (Wang et al., 2021) and RMC (Wu et al., 2020b). We point out important subtleties in the modeling that were unclear or overlooked in previous work.We then study adaptive attacks, that is to leverage the knowledge about Γ to construct attacks. Compared to situations considered in BPDA (Athalye et al., 2018), a transductive learner Γ is even further from being differentiable, and theoretically the attack objective is a bilevel optimization(Colson et al., 2007). To address these difficulties, our key observation is to consider the transferability of adversarial examples, and consider a robust version of (2): max U min U ∈N (U ) L a (Γ(U ), V ) (formula(6)), where we want to find a single attack set U to thwart a family of models, induced by U "around" U . This objective relaxes the attacker-defender constraint, and provides more information in dealing with nondifferentiability. To solve the robust objective, we propose Greedy Model Space Attack (GMSA), a general attack framework which attempts to solve the robust objective in a greedy manner. GMSA can serve as a new baseline for evaluating transductive-learning based defenses.We perform a systematic empirical study on various defenses. For RMC(Wu et al., 2020b), DENT (Wang et al., 2021), and URejectron (Goldwasser et al., 2020), we show that even weak instantiations of GMSA can break respective defenses. Specifically, for defenses based on adversarially training, we reduce the robust accuracy to that of adversarial training alone. We note that, under AutoAttack (Croce & Hein, 2020a), the state-of-the-art adaptive attack for the inductive threat model, some of these defenses have claimed to achieve substantial improvements compared to adversarial training alone. For example, Wang et al. show that DENT can improve the robustness of the state-of-the-art adversarial training defenses by more than 20% absolutely against AutoAttack on CIFAR-10. However, under our adaptive attacks, DENT only has minor improvement: less than 3% improvement over adversarial training alone. Our results thus demonstrates significant differences between attacking transductive-learning based defenses and attacking in the inductive setting, and significant difficulties in the use of transductive learning to improve adversarial robustness. On the positive side, we report a somewhat surprising empirical result of transductive adversarial training: Adversarially retraining the model using fresh private randomness on a new batch of test-time data gives a significant increase in robustness against all of our considered attacks. | TOWARDS EVALUATING THE ROBUSTNESS OF NEURAL NETWORKS LEARNED BY TRANSDUCTION |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.