_id stringlengths 4 10 | text stringlengths 0 18.4k | title stringlengths 0 8.56k |
|---|---|---|
d17262272 | Recently deep neural networks have received considerable attention due to their ability to extract and represent high-level abstractions in data sets. Deep neural networks such as fully-connected and convolutional neural networks have shown excellent performance on a wide range of recognition and classification tasks. However, their hardware implementations currently suffer from large silicon area and high power consumption due to the their high degree of complexity. The power/energy consumption of neural networks is dominated by memory accesses, the majority of which occur in fully-connected networks. In fact, they contain most of the deep neural network parameters. In this paper, we propose sparsely-connected networks, by showing that the number of connections in fullyconnected networks can be reduced by up to 90% while improving the accuracy performance on three popular datasets (MNIST, CIFAR10 and SVHN). We then propose an efficient hardware architecture based on linear-feedback shift registers to reduce the memory requirements of the proposed sparsely-connected networks. The proposed architecture can save up to 90% of memory compared to the conventional implementations of fully-connected neural networks. Moreover, implementation results show up to 84% reduction in the energy consumption of a single neuron of the proposed sparsely-connected networks compared to a single neuron of fully-connected neural networks. | SPARSELY-CONNECTED NEURAL NETWORKS: TO- WARDS EFFICIENT VLSI IMPLEMENTATION OF DEEP NEURAL NETWORKS |
d204800400 | The long-tail distribution of the visual world poses great challenges for deep learning based classification models on how to handle the class imbalance problem. Existing solutions usually involve class-balancing strategies, e.g. by loss re-weighting, data re-sampling, or transfer learning from head-to tail-classes, but most of them adhere to the scheme of jointly learning representations and classifiers. In this work, we decouple the learning procedure into representation learning and classification, and systematically explore how different balancing strategies affect them for long-tailed recognition. The findings are surprising: (1) data imbalance might not be an issue in learning high-quality representations; (2) with representations learned with the simplest instance-balanced (natural) sampling, it is also possible to achieve strong long-tailed recognition ability by adjusting only the classifier. We conduct extensive experiments and set new state-of-the-art performance on common long-tailed benchmarks like ImageNet-LT, Places-LT and iNaturalist, showing that it is possible to outperform carefully designed losses, sampling strategies, even complex modules with memory, by using a straightforward approach that decouples representation and classification. Our code is available at https | Published as a conference paper at ICLR 2020 DECOUPLING REPRESENTATION AND CLASSIFIER FOR LONG-TAILED RECOGNITION |
d251649215 | We consider learning causal relationships under conditional moment restrictions. Unlike causal inference under unconditional moment restrictions, conditional moment restrictions pose serious challenges for causal inference, especially in highdimensional settings. To address this issue, we propose a method that transforms conditional moment restrictions to unconditional moment restrictions through importance weighting, using a conditional density ratio estimator. Using this transformation, we successfully estimate nonparametric functions defined under conditional moment restrictions. Our proposed framework is general and can be applied to a wide range of methods, including neural networks. We analyze the estimation error, providing theoretical support for our proposed method. In experiments, we confirm the soundness of our proposed method.arXiv:2108.01312v2 [econ.EM] 28 Sep 2022Published as a conference paper at ICLR 2022In this paper, we propose transforming conditional moment restrictions into unconditional moment restrictions by importance weighting using the conditional density ratio, which is defined as the ratio of the conditional probability density, conditioned on the IVs, to the unconditional probability density. We show that the unconditional expectation of a random variable weighted by the conditional density ratio is equal to the conditional expectation. Further, we show that it is possible to estimate the conditional density ratio with the least-squares method with a neural network. Once the conditional density ratio is estimated, the usual method of moments, such as GMM, can be used straightforwardly. | LEARNING CAUSAL MODELS FROM CONDITIONAL MOMENT RESTRICTIONS BY IMPORTANCE WEIGHT- ING |
d256827458 | We introduce VA-DepthNet, a simple, effective, and accurate deep neural network approach for the single-image depth prediction (SIDP) problem. The proposed approach advocates using classical first-order variational constraints for this problem. While state-of-the-art deep neural network methods for SIDP learn the scene depth from images in a supervised setting, they often overlook the invaluable invariances and priors in the rigid scene space, such as the regularity of the scene. The paper's main contribution is to reveal the benefit of classical and well-founded variational constraints in the neural network design for the SIDP task. It is shown that imposing first-order variational constraints in the scene space together with popular encoder-decoder-based network architecture design provides excellent results for the supervised SIDP task. The imposed first-order variational constraint makes the network aware of the depth gradient in the scene space, i.e., regularity. The paper demonstrates the usefulness of the proposed approach via extensive evaluation and ablation analysis over several benchmark datasets, such as KITTI, NYU Depth V2, and SUN RGB-D. The VA-DepthNet at test time shows considerable improvements in depth prediction accuracy compared to the prior art and is accurate also at high-frequency regions in the scene space. At the time of writing this paper, our method-labeled as VA-DepthNet, when tested on the KITTI depth-prediction evaluation set benchmarks, shows state-of-the-art results, and is the top-performing published approach 1 2 . * Corresponding Author 1 kitti_depth_prediction_benchmark 2 For official code refer here arXiv:2302.06556v2 [cs.CV] 15 Feb 2023Published as a conference paper at ICLR 2023 2021). Popular recent methods for SIDP are mostly supervised. But even then, they are used less in real-world applications than geometric multiple view methods(Labbé & Michaud, 2019;Müller et al., 2022). Nonetheless, a good solution to SIDP is highly desirable in robotics(Yang et al., 2020), virtual-reality (Hoiem et al., 2005, augmented reality ), view synthesis (Hoiem et al., 2005 and other related vision tasks .In this paper, we advocate that despite the supervised approach being encouraging, SIDP advancement should not wholly rely on the increase of dataset sizes. Instead, geometric cues and scene priors could help improve the SIDP results. Not that scene priors have not been studied to improve SIDP accuracy in the past. For instance, Chen et al. (2016) uses pairwise ordinal relations between points to learn scene depth. Alternatively, Yin et al. (2019) uses surface normals as an auxiliary loss to improve performance. Other heuristic approaches, such as Qi et al. (2018), jointly exploit the depth-to-normal relation to recover scene depth and surface normals. Yet, such state-of-the-art SIDP methods have limitations: for example, the approach in Chen et al. (2016) -using ordinal relation to learn depth -over-smooths the depth prediction results, thereby failing to preserve high-frequency surface details. Conversely, Yin et al. (2019) relies on good depth map prediction from a deep network and the idea of virtual normal. The latter is computed by randomly sampling three noncollinear points with large distances. This is rather complex and heuristic in nature. Qi et al. (2018) uses depth and normal consistency, which is good, yet it requires good depth map initialization. | VA-DEPTHNET: A VARIATIONAL APPROACH TO SIN- GLE IMAGE DEPTH PREDICTION |
d257102833 | This paper shows that the heterogeneity in neuronal and synaptic dynamics reduces the spiking activity of a Recurrent Spiking Neural Network (RSNN) while improving prediction performance, enabling spike-efficient (unsupervised) learning. We analytically show that the diversity in neurons' integration/relaxation dynamics improves an RSNN's ability to learn more distinct input patterns (higher memory capacity), leading to improved classification and prediction performance. We further prove that heterogeneous Spike-Timing-Dependent-Plasticity (STDP) dynamics of synapses reduce spiking activity but preserve memory capacity. The analytical results motivate Heterogeneous RSNN design using Bayesian optimization to determine heterogeneity in neurons and synapses to improve E, defined as the ratio of spiking activity and memory capacity. The empirical results on time series classification and prediction tasks show that optimized HRSNN increases performance and reduces spiking activity compared to a homogeneous RSNN. | HETEROGENEOUS NEURONAL AND SYNAPTIC DYNAM- ICS FOR SPIKE-EFFICIENT UNSUPERVISED LEARNING: THEORY AND DESIGN PRINCIPLES |
d2768038 | We consider learning representations of entities and relations in KBs using the neural-embedding approach. We show that most existing models, including NTN (Socher et al., 2013) and TransE (Bordes et al., 2013b), can be generalized under a unified learning framework, where entities are low-dimensional vectors learned from a neural network and relations are bilinear and/or linear mapping functions. Under this framework, we compare a variety of embedding models on the link prediction task. We show that a simple bilinear formulation achieves new state-of-the-art results for the task (achieving a top-10 accuracy of 73.2% vs. 54.7% by TransE on Freebase). Furthermore, we introduce a novel approach that utilizes the learned relation embeddings to mine logical rules such as BornInCitypa, bq^CityInCountrypb, cq ùñ N ationalitypa, cq. We find that embeddings learned from the bilinear objective are particularly good at capturing relational semantics, and that the composition of relations is characterized by matrix multiplication. More interestingly, we demonstrate that our embedding-based rule extraction approach successfully outperforms a state-ofthe-art confidence-based rule mining approach in mining horn rules that involve compositional reasoning. | Under review as conference paper at ICLR 2015 EMBEDDING ENTITIES AND RELATIONS FOR LEARN- ING AND INFERENCE IN KNOWLEDGE BASES |
d221970551 | A data set sampled from a certain population is biased if the subgroups of the population are sampled at proportions that are significantly different from their underlying proportions. Training machine learning models on biased data sets requires correction techniques to compensate for the bias. We consider two commonlyused techniques, resampling and reweighting, that rebalance the proportions of the subgroups to maintain the desired objective function. Though statistically equivalent, it has been observed that resampling outperforms reweighting when combined with stochastic gradient algorithms. By analyzing illustrative examples, we explain the reason behind this phenomenon using tools from dynamical stability and stochastic asymptotics. We also present experiments from regression, classification, and off-policy prediction to demonstrate that this is a general phenomenon. We argue that it is imperative to consider the objective function design and the optimization algorithm together while addressing the sampling bias.Published as a conference paper at ICLR 2021 explicitly analyzable examples why resampling generates expected results while reweighting performs undesirably. Our theoretical analysis is based on two points of view, one from the dynamical stability perspective and the other from stochastic asymptotics.In addition to the theoretical analysis, we present experimental examples from three distinct categories (classification, regression, and off-policy prediction) to demonstrate that resampling outperforms reweighting in practice. This empirical study illustrates that this is a quite general phenomenon when models are trained using stochastic gradient type algorithms.Our theoretical analysis and experiments show clearly that adjusting only the loss functions is not sufficient for fixing the biased data problem. The output can be disastrous if one overlooks the optimization algorithm used in the training. In fact, recent understanding has shown that objective function design and optimization algorithm are closely related, for example optimization algorithms such as SGD play a key role in the generalizability of deep neural networks. Therefore in order to address the biased data issue, we advocate for considering data, model, and optimization as an integrated system. | Published as a conference paper at ICLR 2021 WHY RESAMPLING OUTPERFORMS REWEIGHTING FOR CORRECTING SAMPLING BIAS WITH STOCHASTIC GRA- DIENTS |
d254408495 | Changing how pre-trained models behave-e.g., improving their performance on a downstream task or mitigating biases learned during pre-training-is a common practice when developing machine learning systems. In this work, we propose a new paradigm for steering the behavior of neural networks, centered around task vectors. A task vector specifies a direction in the weight space of a pre-trained model, such that movement in that direction improves performance on the task. We build task vectors by subtracting the weights of a pre-trained model from the weights of the same model after fine-tuning on a task. We show that these task vectors can be modified and combined together through arithmetic operations such as negation and addition, and the behavior of the resulting model is steered accordingly. Negating a task vector decreases performance on the target task, with little change in model behavior on control tasks. Moreover, adding task vectors together can improve performance on multiple tasks at once. Finally, when tasks are linked by an analogy relationship of the form "A is to B as C is to D", combining task vectors from three of the tasks can improve performance on the fourth, even when no data from the fourth task is used for training. Overall, our experiments with several models, modalities and tasks show that task arithmetic is a simple, efficient and effective way of editing models. * Correspondence to gamaga@cs.washington.edu.1We use the term editing to refer to any intervention done to a model done after the pre-training stage. | Published as a conference paper at ICLR 2023 EDITING MODELS WITH TASK ARITHMETIC |
d6021932 | We present experiments demonstrating that some other form of capacity control, different from network size, plays a central role in learning multi-layer feedforward networks. We argue, partially through analogy to matrix factorization, that this is an inductive bias that can help shed light on deep learning. | IN SEARCH OF THE REAL INDUCTIVE BIAS: ON THE ROLE OF IMPLICIT REGULARIZATION IN DEEP LEARNING |
d257255248 | Building systems that achieve a deeper understanding of language is one of the central goals of natural language processing (NLP). Towards this goal, recent works have begun to train language models on narrative datasets which require extracting the most critical information by integrating across long contexts. However, it is still an open question whether these models are learning a deeper understanding of the text, or if the models are simply learning a heuristic to complete the task. This work investigates this further by turning to the one language processing system that truly understands complex language: the human brain. We show that training language models for deeper narrative understanding results in richer representations that have improved alignment to human brain activity. We further find that the improvements in brain alignment are larger for character names than for other discourse features, which indicates that these models are learning important narrative elements. Taken together, these results suggest that this type of training can indeed lead to deeper language understanding. These findings have consequences both for cognitive neuroscience by revealing some of the significant factors behind brain-NLP alignment, and for NLP by highlighting that understanding of long-range context can be improved beyond language modeling.We specifically investigate 4 pretrained language models (i.e., "base models") and 4 corresponding models obtained by training the base models on the BookSum dataset (Kryscinski et al., 2021) to improve the base language model's narrative understanding (i.e., "booksum models"). The Book-Sum dataset was selected because it is a summarization dataset that requires understanding complex interactions across long narratives. The 4 models were selected because their architectures were designed to integrate information across long contexts. We evaluate the alignment of the base and booksum models with fMRI recordings of 8 participants reading a chapter of a popular book wordby-word, made publicly available by Wehbe et al. (2014a). This dataset was chosen because it is one of the largest datasets of participants processing a narrative story (5176 words which corresponds to approximately 1300 samples of fMRI recordings per participant).Our main contributions are as follows:1. In Section 4, we show that training language models for deeper narrative understanding improves alignment to human brain activity. Also, when increasing the number of words fed to the models, up to 500 words, brain alignment increases. Lastly, for each model, we identify the layers where these improvements in brain alignment occur. 2. In Section 5, we show that improved brain alignment in Section 4 is not due to improved language modeling (LM) ability, a possible confounding factor. By disentangling LM ability's contribution to brain alignment, we present evidence that BookSum-trained models develop deeper language understanding. 3. In Section 6, we present a simple interpretability approach to study what brain-relevant information is gained by language models after training for deeper language understanding. Our results reveal that these models are learning richer representations across all tested discourse features (Characters, Emotions, Motions). Furthermore, they learn more about Characters than Emotions and Motions. This indicates that discourse features are a promising dimension to study brain alignment and deep language understanding.Combined, our contributions from Sections 4, 5, and 6 present evidence that models trained to summarize narratives indeed develop deeper language understanding. The first reason is that improved alignment to human brains' deep understanding of characters, emotions and motions suggests the model has developed richer representations of these entities and concepts. Second, we focus on brain regions suggested by previous research to underlie language comprehension in humans. Hence, improved brain alignment is not spuriously related to non-language brain activities. Third, we show that brain alignment improves only when we provide longer input contexts (20 to 1000 words) to the LMs, which may be important for deep contextual understanding. | Published as a conference paper at ICLR 2023 TRAINING LANGUAGE MODELS TO SUMMARIZE NARRATIVES IMPROVES BRAIN ALIGNMENT |
d247518913 | Searching for the architecture cells is a dominant paradigm in NAS. However, little attention has been devoted to the analysis of the cell-based search spaces even though it is highly important for the continual development of NAS. In this work, we conduct an empirical post-hoc analysis of architectures from the popular cellbased search spaces and find that the existing search spaces contain a high degree of redundancy: the architecture performance is minimally sensitive to changes at large parts of the cells, and universally adopted designs, like the explicit search for a reduction cell, significantly increase the complexities but have very limited impact on the performance. Across architectures found by a diverse set of search strategies, we consistently find that the parts of the cells that do matter for architecture performance often follow similar and simple patterns. By explicitly constraining cells to include these patterns, randomly sampled architectures can match or even outperform the state of the art. These findings cast doubts into our ability to discover truly novel architectures in the existing cell-based search spaces, and inspire our suggestions for improvement to guide future NAS research. Code is available at https: //github.com/xingchenwan/cell-based-NAS-analysis.Published as a conference paper at ICLR 2022 finding novel and high-performing architectures, not only in conventional CNNs but also in emerging NAS paradigms such as Transformers (which may also take the form of a cell-based design space). Second, opening the NAS black box enables us to distill the essence of the strong-performing NAS architectures beneath their surface of complexity. Unlike manually designed architectures where usually designers attribute performance to specific designs, currently owing to the apparent complexity of the design space, the NAS architectures, while all discovered in a similar or identical search space, are often compared to in terms of final performance on a standard dataset only (e.g. CIFAR-10 test error). This could be problematic, as we do not necessarily understand what NAS has discovered that led to the purported improvements, and the metric itself is a poor one on which external factors such as hyperparameter settings, variations in training/data augmentation protocols and even just noise could exert a greater influence than the architectural design itself(Yang et al., 2020a). However, by linking performance to specific designs, we could ascertain whether any performance differences stem from the architectures rather than the interfering factors. We aim to address this problem by presenting a post-hoc analysis of the well-performing architectures produced by technically diverse search methods. Specifically, we utilise explainable machine learning tools to open the NAS black box by inspecting the good-and bad-performing architectures produced by a wide range of NAS search methods in the dominant DARTS search space. We find: • Performances of architectures can often be disproportionately attributed to a small number of simple yet critical features that resemble known patterns in classical network designs; • Many designs almost universally adopted contribute to complexity but not performance; • The nominal complexity of the search spaces poorly reflects the actual diversity of the (highperforming) architectures discovered, the functional parts of which are often very similar despite the technical diversity in search methods and the seeming disparity in topology. In fact, with few simple and human-interpretable constraints, almost any randomly sampled architecture can perform on par or exceed those produced by the state-of-the-art NAS methods over varying network sizes and datasets (CIFAR-10/IMAGENET). Ultimately, these findings prompt us to rethink the suitability of the current standard protocol in evaluating NAS and the capability to find truly novel architectures within the search space. We finally provide suggestions for prospective new search spaces inspired by these findings. | Published as a conference paper at ICLR 2022 ON REDUNDANCY AND DIVERSITY IN CELL-BASED NEURAL ARCHITECTURE SEARCH |
d16563349 | Deep Belief Networks which are hierarchical generative models are effective tools for feature representation and extraction. Furthermore, DBNs can be used in numerous aspects of Machine Learning such as image denoising. In this paper, we propose a novel method for image denoising which relies on the DBNs' ability in feature representation. This work is based upon learning of the noise behavior. Generally, features which are extracted using DBNs are presented as the values of the last layer nodes. We train a DBN a way that the network totally distinguishes between nodes presenting noise and nodes presenting image content in the last later of DBN, i.e. the nodes in the last layer of trained DBN are divided into two distinct groups of nodes. After detecting the nodes which are presenting the noise, we are able to make the noise nodes inactive and reconstruct a noiseless image. In section 4 we explore the results of applying this method on the MNIST dataset of handwritten digits which is corrupted with additive white Gaussian noise (AWGN). A reduction of 65.9% in average mean square error (MSE) was achieved when the proposed method was used for the reconstruction of the noisy images. | Deep Belief Networks for Image Denoising |
d204509670 | Neural ordinary differential equations (ODEs) have been attracting increasing attention in various research domains recently. There have been some works studying optimization issues and approximation capabilities of neural ODEs, but their robustness is still yet unclear. In this work, we fill this important gap by exploring robustness properties of neural ODEs both empirically and theoretically. We first present an empirical study on the robustness of the neural ODE-based networks (ODENets) by exposing them to inputs with various types of perturbations and subsequently investigating the changes of the corresponding outputs. In contrast to conventional convolutional neural networks (CNNs), we find that the ODENets are more robust against both random Gaussian perturbations and adversarial attack examples. We then provide an insightful understanding of this phenomenon by exploiting a certain desirable property of the flow of a continuous-time ODE, namely that integral curves are non-intersecting. Our work suggests that, due to their intrinsic robustness, it is promising to use neural ODEs as a basic block for building robust deep network models. To further enhance the robustness of vanilla neural ODEs, we propose the time-invariant steady neural ODE (TisODE), which regularizes the flow on perturbed data via the time-invariant property and the imposition of a steady-state constraint. We show that the TisODE method outperforms vanilla neural ODEs and also can work in conjunction with other state-of-the-art architectural methods to build more robust deep networks.Published as a conference paper at ICLR 2020 to a multi-channel feature map, a neural ODE that serves as the nonlinear representation mapping (RM), and the fully-connected classifier (FCC) that generates a prediction vector based on the output of the RM. Figure 1: The architecture of an ODENet. The neural ODE block serves as a dimension-preserving nonlinear mapping.The robustness of a classification model can be evaluated through the lens of its performance on perturbed images. To comprehensively investigate the robustness of neural ODEs, we perturb original images with commonly-used perturbations, namely, random Gaussian noise(Szegedy et al., 2013)and harmful adversarial examples(Goodfellow et al., 2014;Madry et al., 2017). We conduct experiments in two common settings-training the model only on authentic non-perturbed images and training the model on authentic images as well as the Gaussian perturbed ones. We observe that ODENets are more robust compared to CNN models against all types of perturbations in both settings. We then provide an insightful understanding of such intriguing robustness of neural ODEs by exploiting a certain property of the flow(Dupont et al., 2019), namely that integral curves that start at distinct initial states are nonintersecting. The flow of a continuous-time ODE is defined as the family of solutions/paths traversed by the state, starting from different initial points, and an integral curve is a specific solution for a given initial point. The non-intersecting property indicates that an integral curve starting from some point is constrained by the integral curves starting from that point's neighborhood. Thus, in an ODENet, if a correctly classified datum is slightly perturbed, the integral curve associated to its perturbed version would not change too much from the original one. Consequently, the perturbed datum could still be correctly classified. Thus, there exists intrinsic robustness regularization in ODENets, which is absent from CNNs. | ON ROBUSTNESS OF NEURAL ORDINARY DIFFEREN- TIAL EQUATIONS |
d257279790 | Large pre-trained language models help to achieve state of the art on a variety of natural language processing (NLP) tasks, nevertheless, they still suffer from forgetting when incrementally learning a sequence of tasks. To alleviate this problem, recent works enhance existing models by sparse experience replay and local adaption, which yield satisfactory performance. However, in this paper we find that pre-trained language models like BERT have a potential ability to learn sequentially, even without any sparse memory replay. To verify the ability of BERT to maintain old knowledge, we adopt and re-finetune single-layer probe networks with the parameters of BERT fixed. We investigate the models on two types of NLP tasks, text classification and extractive question answering. Our experiments reveal that BERT can actually generate high quality representations for previously learned tasks in a long term, under extremely sparse replay or even no replay. We further introduce a series of novel methods to interpret the mechanism of forgetting and how memory rehearsal plays a significant role in task incremental learning, which bridges the gap between our new discovery and previous studies about catastrophic forgetting 1 . | CAN BERT REFRAIN FROM FORGETTING ON SEQUEN- TIAL TASKS? A PROBING STUDY |
d214014558 | We study reinforcement learning in settings where sampling an action from the policy must be done concurrently with the time evolution of the controlled system, such as when a robot must decide on the next action while still performing the previous action. Much like a person or an animal, the robot must think and move at the same time, deciding on its next action before the previous one has completed. In order to develop an algorithmic framework for such concurrent control problems, we start with a continuous-time formulation of the Bellman equations, and then discretize them in a way that is aware of system delays. We instantiate this new class of approximate dynamic programming methods via a simple architectural extension to existing value-based deep reinforcement learning algorithms. We evaluate our methods on simulated benchmark tasks and a large-scale robotic grasping task where the robot must "think while moving". Videos are available at | Published as a conference paper at ICLR 2020 THINKING WHILE MOVING: DEEP REINFORCEMENT LEARNING WITH CONCURRENT CONTROL |
d16305497 | Competing with top human players in the ancient game of Go has been a longterm goal of artificial intelligence. Go's high branching factor makes traditional search techniques ineffective, even on leading-edge hardware, and Go's evaluation function could change drastically with one stone change. Recent works[Maddison et al. (2015); Clark & Storkey(2015)] show that search is not strictly necessary for machine Go players. A pure pattern-matching approach, based on a Deep Convolutional Neural Network (DCNN) that predicts the next move, can perform as well as Monte Carlo Tree Search (MCTS)-based open source Go engines such as Pachi [Baudis & Gailly(2012)] if its search budget is limited. We extend this idea in our bot named darkforest, which relies on a DCNN designed for long-term predictions. Darkforest substantially improves the win rate for patternmatching approaches against MCTS-based approaches, even with looser search budgets. Against human players, the newest versions, darkfores2, achieve a stable 3d level on KGS Go Server as a ranked bot, a substantial improvement upon the estimated 4k-5k ranks for DCNN reported in Clark & Storkey (2015) based on games against other machine players. Adding MCTS to darkfores2 creates a much stronger player named darkfmcts3: with 5000 rollouts, it beats Pachi with 10k rollouts in all 250 games; with 75k rollouts it achieves a stable 5d level in KGS server, on par with state-of-the-art Go AIs (e.g., Zen, DolBaram, CrazyStone) except for AlphaGo[Silver et al. (2016)]; with 110k rollouts, it won the 3rd place in January KGS Go Tournament. | BETTER COMPUTER GO PLAYER WITH NEURAL NET- WORK AND LONG-TERM PREDICTION |
d189999405 | As the performance of computer systems stagnates due to the end of Moore's Law, there is a need for new models that can understand and optimize the execution of general purpose code. While there is a growing body of work on using Graph Neural Networks (GNNs) to learn static representations of source code, these representations do not understand how code executes at runtime. In this work, we propose a new approach using GNNs to learn fused representations of general source code and its execution. Our approach defines a multi-task GNN over low-level representations of source code and program state (i.e., assembly code and dynamic memory states), converting complex source code constructs and data structures into a simpler, more uniform format. We show that this leads to improved performance over similar methods that do not use execution and it opens the door to applying GNN models to new tasks that would not be feasible from static code alone. As an illustration of this, we apply the new model to challenging dynamic tasks (branch prediction and prefetching) from the SPEC CPU benchmark suite, outperforming the state-of-the-art by 26% and 45% respectively. Moreover, we use the learned fused graph embeddings to demonstrate transfer learning with high performance on an indirectly related algorithm classification task. * Work completed during an internship at Google. 1 As Moore's Law ends, prediction techniques in these fields have also stagnated. For example, the winner of the most recent branch prediction championship increased precision by 3.7% (Dundas, 2016). | Published as a conference paper at ICLR 2020 LEARNING EXECUTION THROUGH NEURAL CODE FUSION |
d246294455 | In Multiple Instance Learning (MIL), models are trained using bags of instances, where only a single label is provided for each bag. A bag label is often only determined by a handful of key instances within a bag, making it difficult to interpret what information a classifier is using to make decisions. In this work, we establish the key requirements for interpreting MIL models. We then go on to develop several model-agnostic approaches that meet these requirements. Our methods are compared against existing inherently interpretable MIL models on several datasets, and achieve an increase in interpretability accuracy of up to 30%. We also examine the ability of the methods to identify interactions between instances and scale to larger datasets, improving their applicability to real-world problems.1 Source code for this project is available at https://github.com/JAEarly/MILLI. The remainder of this work is laid out as follows. Section 2 provides relevant background knowledge and reviews existing MIL interpretability methods. Section 3 outlines the requirements for MIL interpretability and details our approaches that meet these requirements. Next, Section 4 provides our results and experiments. Section 5 discusses our findings, and Section 6 concludes.BACKGROUND AND RELATED WORKThe standard MIL assumption (SMIL) is a binary problem with positive and negative bags(Dietterich et al., 1997;Maron & Lozano-Pérez, 1998). A bag is positive if any of its instances are positive, otherwise it is negative. The assumptions of SMIL can be relaxed to allow more generalised versions of MIL, e.g., through extensions that include additional positive classes(Scott et al., 2005;Weidmann et al., 2003). SMIL also assumes there is no interaction between instances. However, recent work has highlighted that modelling relationships between instances is beneficial for performance(Tu et al., 2019;Zhou et al., 2009). Existing methods for interpreting MIL models often rely on the SMIL assumption, so cannot generalise to other MIL problems. In addition, existing methods are often model-specific, i.e., they only work for certain types of MIL models, which constrains the choice of model(Ribeiro et al., 2016a). Identifying the key instances in MIL bags is a form of local interpretability (Molnar, 2020), as the key instances are detected for a particular input to the model. Two of the key motivators for interpreting the decision-making process of a MIL system are reliability and trust -identifying the key instances allows an evaluation of the reliability of the system, which increases trust as the decision-making process becomes more transparent. In this work, we use the term interpretability rather than explainability to convey that the analysis remains tied to the models, i.e., these methods do not provide non-technical explanations in human terms.Under the SMIL assumption, which and what questions are equivalent -there are only two classes (positive and negative), and the only key instances are the instances that are positive, therefore once the key instances are identified, it is also known what outcome they support. However, when there are multiple positive classes, if instances from different classes co-occur in the same bags, answering which and what questions becomes two distinct problems. For example, some key instances will support one positive class, and some key instances will support another. Therefore, solely identifying which are the key instances does not answer the second question of what class they support. Existing methods, such as key instance detection (Liu et al., 2012), MIL attention (Ilse et al., 2018), and MIL graph neural networks (GNNs; Tu et al. (2019)) do not condition their output on a particular class, so it is not apparent what class each instance supports, i.e., they can only answer which questions. One existing method that can answer what questions is mi-Net (Wang et al., 2018), as it produces instance-level predictions as part of its processing. However, these instance-level predictions do not take account of interactions between the instances, so are often inaccurate. A related piece of work on MIL interpretability is Tibo et al. (2020), which considers interpretability within the scope of multi-multi-instance learning (MMIL; Tibo et al. (2017); Fuster et al. (2021)). In MMIL, the instances within a bag are arranged into into further bags, giving a hierarchical bags-of-bags structure. The interpretability techniques presented byTibo et al. (2020)are model-specific as they are only designed for MMIL networks. In this work, we aim to overcome the limitations of existing methods by developing model-agnostic methods that can answer both which and what questions.In single instance supervised learning, model-agnostic techniques have been developed to interpret models. Post-hoc local interpretability methods, such as Local Interpretable Model-agnostic Explanations (LIME;Ribeiro et al. (2016b)) and SHapley Additive exPlanations (SHAP; Lundberg & Lee (2017)), work by approximating the original predictive model with a locally faithful surrogate model that is inherently interpretable. The surrogate model learns from simplified inputs that represent perturbations of the original input that is being analysed. In this work, one of our proposed methods is a MIL-specific version of this approach.METHODOLOGYAt the start of this section, we outline the requirements for MIL interpretability (Section 3.1). In Section 3.2, we propose three methods that meet these requirements under the assumption that there are no interactions between instances (independent-instance methods). In Section 3.3 we remove this assumption and propose our local surrogate model-agnostic interpretability method for MIL. | Published as a conference paper at ICLR 2022 MODEL AGNOSTIC INTERPRETABILITY FOR MULTIPLE INSTANCE LEARNING |
d228063783 | Predicting the dynamics of neural network parameters during training is one of the key challenges in building a theoretical foundation for deep learning.A central obstacle is that the motion of a network in high-dimensional parameter space undergoes discrete finite steps along complex stochastic gradients derived from real-world datasets.We circumvent this obstacle through a unifying theoretical framework based on intrinsic symmetries embedded in a network's architecture that are present for any dataset.We show that any such symmetry imposes stringent geometric constraints on gradients and Hessians, leading to an associated conservation law in the continuous-time limit of stochastic gradient descent (SGD), akin to Noether's theorem in physics.We further show that finite learning rates used in practice can actually break these symmetry induced conservation laws.We apply tools from finite difference methods to derive modified gradient flow, a differential equation that better approximates the numerical trajectory taken by SGD at finite learning rates.We combine modified gradient flow with our framework of symmetries to derive exact integral expressions for the dynamics of certain parameter combinations.We empirically validate our analytic predictions for learning dynamics on VGG-16 trained on Tiny ImageNet.Overall, by exploiting symmetry, our work demonstrates that we can analytically describe the learning dynamics of various parameter combinations at finite learning rates and batch sizes for state of the art architectures trained on any dataset. | NEURAL MECHANICS: SYMMETRY AND BROKEN CON-SERVATION LAWS IN DEEP LEARNING DYNAMICS |
d248987713 | Knowledge Distillation (KD) is a typical method for training a lightweight student model with the help of a well-trained teacher model. However, most KD methods require access to either the teacher's training data or model parameter, which is unrealistic. To tackle this problem, recent works study KD under data-free and black-box settings. Nevertheless, these works require a large number of queries to the teacher model, which incurs significant monetary and computational costs. To address these problems, we propose a novel method called query-effIcient Datafree lEarning from blAck-box modeLs (IDEAL), which aims to query-efficiently learn from black-box model APIs to train a good student without any real data. In detail, IDEAL trains the student model in two stages: data generation and model distillation. Note that IDEAL does not require any query in the data generation stage and queries the teacher only once for each sample in the distillation stage. Extensive experiments on various real-world datasets show the effectiveness of the proposed IDEAL. For instance, IDEAL can improve the performance of the best baseline method DFME by 5.83% on CIFAR10 dataset with only 0.02× the query budget of DFME. * Equal contribution. † Work done during internship at Sony AI. | Published as a conference paper at ICLR 2023 IDEAL: QUERY-EFFICIENT DATA-FREE LEARNING FROM BLACK-BOX MODELS |
d3299269 | Training deep neural networks (DNNs) efficiently is a challenge due to the associated highly nonconvex optimization. The backpropagation (backprop) algorithm has long been the most widely used algorithm for gradient computation of parameters of DNNs and is used along with gradient descent-type algorithms for this optimization task. Recent work have shown the efficiency of block coordinate descent (BCD) type methods empirically for training DNNs. In view of this, we propose a novel algorithm based on the BCD method for training DNNs and provide its global convergence results built upon the powerful framework of the Kurdyka-Łojasiewicz (KL) property. Numerical experiments on standard datasets demonstrate its competitive efficiency against standard optimizers with backprop.RELATED WORKCarreira-Perpiñán & Wang (2014) and Zhang & Brand (2017) also suggest the use of BCD for training DNNs and observe empirically the per epoch efficiency where the training loss drops much faster than SGD. Multiple related work consider a similar scheme to ours. A very recent piece of work (Frerix et al., 2018) implements proximal steps for model parameter updates only but keep 1 arXiv:1803.09082v1 [stat.ML] | Workshop track -ICLR 2018 A PROXIMAL BLOCK COORDINATE DESCENT ALGO- RITHM FOR DEEP NEURAL NETWORK TRAINING |
d49666783 | Model-based reinforcement learning (RL) is considered to be a promising approach to reduce the sample complexity that hinders model-free RL. However, the theoretical understanding of such methods has been rather limited. This paper introduces a novel algorithmic framework for designing and analyzing model-based RL algorithms with theoretical guarantees. We design a meta-algorithm with a theoretical guarantee of monotone improvement to a local maximum of the expected reward. The meta-algorithm iteratively builds a lower bound of the expected reward based on the estimated dynamical model and sample trajectories, and then maximizes the lower bound jointly over the policy and the model. The framework extends the optimism-in-face-of-uncertainty principle to non-linear dynamical models in a way that requires no explicit uncertainty quantification. Instantiating our framework with simplification gives a variant of model-based RL algorithms Stochastic Lower Bounds Optimization (SLBO). Experiments demonstrate that SLBO achieves stateof-the-art performance when only one million or fewer samples are permitted on a range of continuous control benchmark tasks. 1 | ALGORITHMIC FRAMEWORK FOR MODEL-BASED DEEP REINFORCEMENT LEARNING WITH THEORETI- CAL GUARANTEES |
d14949670 | Natural language generation plays a critical role in any spoken dialogue system. We present a new approach to natural language generation using recurrent neural networks in an encoderdecoder framework. In contrast with previous work, our model uses both lexicalized and delexicalized versions of slot-value pairs for each dialogue act. This allows our model to learn from all available data, rather than being restricted to learning only from delexicalized slot-value pairs. We show that this helps our model generate more natural sentences with better grammar. We further improve our model's performance by initializing its weights from a pretrained language model. Human evaluation of our best-performing model indicates that it generates sentences which users find more natural and appealing. | Natural Language Generation in Dialogue using Lexicalized and Delexicalized Data |
d235294221 | Graph generative models are a highly active branch of machine learning. Given the steady development of new models of ever-increasing complexity, it is necessary to provide a principled way to evaluate and compare them. In this paper, we enumerate the desirable criteria for such a comparison metric and provide an overview of the status quo of graph generative model comparison in use today, which predominantly relies on the maximum mean discrepancy (MMD). We perform a systematic evaluation of MMD in the context of graph generative model comparison, highlighting some of the challenges and pitfalls researchers inadvertently may encounter. After conducting a thorough analysis of the behaviour of MMD on synthetically-generated perturbed graphs as well as on recently-proposed graph generative models, we are able to provide a suitable procedure to mitigate these challenges and pitfalls. We aggregate our findings into a list of practical recommendations for researchers to use when evaluating graph generative models. | Published as a conference paper at ICLR 2022 EVALUATION METRICS FOR GRAPH GENERATIVE MODELS: PROBLEMS, PITFALLS, AND PRACTICAL SOLUTIONS |
d9545399 | We systematically explore regularizing neural networks by penalizing low entropy output distributions. We show that penalizing low entropy output distributions, which has been shown to improve exploration in reinforcement learning, acts as a strong regularizer in supervised learning. Furthermore, we connect a maximum entropy based confidence penalty to label smoothing through the direction of the KL divergence. We exhaustively evaluate the proposed confidence penalty and label smoothing on 6 common benchmarks: image classification (MNIST and Cifar-10), language modeling (Penn Treebank), machine translation (WMT'14 English-to-German), and speech recognition (TIMIT and WSJ). We find that both label smoothing and the confidence penalty improve state-of-the-art models across benchmarks without modifying existing hyperparameters, suggesting the wide applicability of these regularizers. * Work done as part of the Google Brain Residency Program † Equal Contribution | REGULARIZING NEURAL NETWORKS BY PENALIZING CONFIDENT OUTPUT DISTRIBUTIONS |
d257505014 | We consider the task of generating realistic 3D shapes, which is useful for a variety of applications such as automatic scene generation and physical simulation. Compared to other 3D representations like voxels and point clouds, meshes are more desirable in practice, because (1) they enable easy and arbitrary manipulation of shapes for relighting and simulation, and (2) they can fully leverage the power of modern graphics pipelines which are mostly optimized for meshes. Previous scalable methods for generating meshes typically rely on sub-optimal post-processing, and they tend to produce overly-smooth or noisy surfaces without fine-grained geometric details. To overcome these shortcomings, we take advantage of the graph structure of meshes and use a simple yet very effective generative modeling method to generate 3D meshes. Specifically, we represent meshes with deformable tetrahedral grids, and then train a diffusion model on this direct parametrization. We demonstrate the effectiveness of our model on multiple generative tasks.Published as a conference paper at ICLR 2023 for computers, both voxels and point clouds are relatively hard for artists to edit, especially when the generated 3D shapes are complex and of low quality. Moreover, modern graphics pipelines are built and optimized for explicit geometry representations like meshes, making meshes one of the most desirable final 3D shape representations. While it is still possible to use methods like Poisson reconstruction to obtain surfaces from voxels and points clouds, the resulted surfaces are generally noisy and contain many topological artifacts, even with carefully tuned hyperparameters.To improve the representation flexibility, sign distance fields (SDFs) have been adopted to model shape surfaces, which enables us to use marching cubes [29] to extract the zero-surfaces and thus 3D meshes. However, SDFs are typically harder to learn as it requires a carefully designed sampling strategy and regularization. Because SDFs are usually parameterized with multi-layer perceptrons (MLPs) in which a smoothness prior is implicitly embedded, the generated shapes tend to be so smooth that sharp edges and important (and potentially semantic) details are lost. Moreover, SDFs are costly to render and therefore less suitable for downstream tasks like conditional generation with RGB images, which require an efficient differentiable renderer during inference.We instead aim to generate 3D shapes by directly producing 3D meshes, where surfaces are represented as a graph of triangular or polygon faces. With 3D meshes, all local surface information is completely included in the mesh vertices (along with the vertex connectivity), because the surface normal of any point on the shape surface is simply a nearest neighbor or some local linear combination of vertex normals. Such a regular structure with rich geometric details enables us to better model the data distribution and learn generative models that are more geometry-aware. In light of recent advances in score-based generative modeling[16,47]where powerful generative performance and effortless training are demonstrated, we propose to train diffusion models on these vertices to generate meshes. However, it is by no means a trivial task and poses two critical problems: (1) the numbers of vertices and faces are indefinite for general object categories, and (2) the underlying topology varies wildly and edges have to be generated at the same time. | Published as a conference paper at ICLR 2023 MESHDIFFUSION: SCORE-BASED GENERATIVE 3D MESH MODELING |
d232273817 | Many complex real-world tasks are composed of several levels of sub-tasks. Humans leverage these hierarchical structures to accelerate the learning process and achieve better generalization. In this work, we study the inductive bias and propose Ordered Memory Policy Network (OMPN) to discover subtask hierarchy by learning from demonstration. The discovered subtask hierarchy could be used to perform task decomposition, recovering the subtask boundaries in an unstructured demonstration. Experiments on Craft and Dial demonstrate that our model can achieve higher task decomposition performance under both unsupervised and weakly supervised settings, comparing with strong baselines. OMPN can also be directly applied to partially observable environments and still achieve higher task decomposition performance. Our visualization further confirms that the subtask hierarchy can emerge in our model 1 . | Published as a conference paper at ICLR 2021 LEARNING TASK DECOMPOSITION WITH ORDERED MEMORY POLICY NETWORK |
d252846193 | Work on fast weight programmers has demonstrated the effectiveness of key/value outer product-based learning rules for sequentially generating a weight matrix (WM) of a neural net (NN) by another NN or itself. However, the weight generation steps are typically not visually interpretable by humans, because the contents stored in the WM of an NN are not. Here we apply the same principle to generate natural images. The resulting fast weight painters (FPAs) learn to execute sequences of delta learning rules to sequentially generate images as sums of outer products of selfinvented keys and values, one rank at a time, as if each image was a WM of an NN. We train our FPAs in the generative adversarial networks framework, and evaluate on various image datasets. We show how these generic learning rules can generate images with respectable visual quality without any explicit inductive bias for images. While the performance largely lags behind the one of specialised state-ofthe-art image generators, our approach allows for visualising how synaptic learning rules iteratively produce complex connection patterns, yielding human-interpretable meaningful images. Finally, we also show that an additional convolutional U-Net (now popular in diffusion models) at the output of an FPA can learn one-step "denoising" of FPA-generated images to enhance their quality. Our code is public. 1 | Published as a conference paper at ICLR 2023 IMAGES AS WEIGHT MATRICES: SEQUENTIAL IMAGE GENERATION THROUGH SYNAPTIC LEARNING RULES |
d257019892 | With the increasing amount of multimedia data on modern mobile systems and IoT infrastructures, harnessing these rich multimodal data without breaching user privacy becomes a critical issue. Federated learning (FL) serves as a privacyconscious alternative to centralized machine learning. However, existing FL methods extended to multimodal data all rely on model aggregation on single modality level, which restrains the server and clients to have identical model architecture for each modality. This limits the global model in terms of both model complexity and data capacity, let alone task diversity. In this work, we propose Contrastive Representation Ensemble and Aggregation for Multimodal FL (CreamFL), a multimodal federated learning framework that enables training larger server models from clients with heterogeneous model architectures and data modalities, while only communicating knowledge on public dataset. To achieve better multimodal representation fusion, we design a global-local cross-modal ensemble strategy to aggregate client representations. To mitigate local model drift caused by two unprecedented heterogeneous factors stemming from multimodal discrepancy (modality gap and task gap), we further propose inter-modal and intra-modal contrasts to regularize local training, which complements information of the absent modality for uni-modal clients and regularizes local clients to head towards global consensus. Thorough evaluations and ablation studies on image-text retrieval and VQA tasks showcase the superiority of CreamFL over state-of-the-art FL methods. | MULTIMODAL FEDERATED LEARNING VIA CON- TRASTIVE REPRESENTATION ENSEMBLE |
d234096198 | Network pruning is an effective method to reduce the computational expense of over-parameterized neural networks for deployment on low-resource systems. Recent state-of-the-art techniques for retraining pruned networks such as weight rewinding and learning rate rewinding have been shown to outperform the traditional fine-tuning technique in recovering the lost accuracy (Renda et al., 2020), but so far it is unclear what accounts for such performance. In this work, we conduct extensive experiments to verify and analyze the uncanny effectiveness of learning rate rewinding. We find that the reason behind the success of learning rate rewinding is the usage of a large learning rate. Similar phenomenon can be observed in other learning rate schedules that involve large learning rates, e.g., the 1-cycle learning rate schedule (Smith & Topin, 2019). By leveraging the right learning rate schedule in retraining, we demonstrate a counter-intuitive phenomenon in that randomly pruned networks could even achieve better performance than methodically pruned networks (fine-tuned with the conventional approach). Our results emphasize the cruciality of the learning rate schedule in pruned network retraining -a detail often overlooked by practioners during the implementation of network pruning.However, the use of large networks exacerbate the gap between research and practice since real-world applications usually require running neural networks in low-resource environments for numerous purposes: reducing memory, latency, energy consumption, etc. To adopt those networks to resourceconstrained devices, network pruning(LeCun et al., 1990;Han et al., 2015;is often exploited to remove dispensable weights, filters and other structures from neural networks. The goal of pruning is to reduce overall computational cost and memory footprint without inducing significant drop in performance of the network.A common approach to mitigating performance drop after pruning is retraining: we continue to train the pruned models for some more epochs. In this paper, we are interested in approaches based on learning rate schedules to control the retraining. A well-known practice is fine-tuning, which aims to train the pruned model with a small fixed learning rate. More advanced learning rate schedules exist, which we generally refer to as retraining. The retraining step is a critical part in implementing network pruning, but it has been largely overlooked and tend to vary in each implementation including differences in learning rate schedules, retraining budget, hyperparameter choices, etc.Recently,Renda et al. (2020)proposed a state-of-the-art technique for retraining pruned networks namely learning rate rewinding (LRW). Specifically, instead of fine-tuning the pruned networks with a fixed learning rate, usually the last learning rate from the original training schedule(Han et al., 2015;Liu et al., 2019), the authors suggested using the learning rate schedule from the previous t epochs (i.e. rewinding). This seemingly subtle change in learning rate schedule led to an important result: LRW was shown to achieve comparable performance to more complex and computationally expensive | Published as a conference paper at ICLR 2021 NETWORK PRUNING THAT MATTERS: A CASE STUDY ON RETRAINING VARIANTS |
d4735139 | In dynamic malware analysis, programs are classified as malware or benign based on their execution logs. We propose a concept of applying monotonic classification models to the analysis process, to make the trained model's predictions consistent over execution time and provably stable to the injection of any noise or 'benign-looking' activity into the program's behavior. The predictions of such models change monotonically through the log in the sense that the addition of new lines into the log may only increase the probability of the file being found malicious, which make them suitable for real-time classification on a user's machine. We evaluate monotonic neural network models based on the work byChistyakov et al. (2017)and demonstrate that they provide stable and interpretable results. | Workshop track -ICLR 2018 MONOTONIC MODELS FOR REAL-TIME DYNAMIC MALWARE DETECTION |
d252110963 | Latent manifolds provide a compact characterization of neural population activity and of shared co-variability across brain areas. Nonetheless, existing statistical tools for extracting neural manifolds face limitations in terms of interpretability of latents with respect to task variables, and can be hard to apply to datasets with no trial repeats. Here we propose a novel probabilistic framework that allows for interpretable partitioning of population variability within and across areas in the context of naturalistic behavior. Our approach for task aligned manifold estimation (TAME-GP) extends a probabilistic variant of demixed PCA by (1) explicitly partitioning variability into private and shared sources, (2) using a Poisson noise model, and (3) introducing temporal smoothing of latent trajectories in the form of a Gaussian Process prior. This TAME-GP graphical model allows for robust estimation of task-relevant variability in local population responses, and of shared co-variability between brain areas. We demonstrate the efficiency of our estimator on within model and biologically motivated simulated data. We also apply it to neural recordings in a closed-loop virtual navigation task in monkeys, demonstrating the capacity of TAME-GP to capture meaningful intra-and inter-area neural variability with single trial resolution.Preprint. Under review. | A probabilistic framework for task-aligned intra-and inter-area neural manifold estimation |
d118696157 | Learning to take actions based on observations is a core requirement for artificial agents to be able to be successful and robust at their task. Reinforcement Learning (RL) is a well-known technique for learning such policies. However, current RL algorithms often have to deal with reward shaping, have difficulties generalizing to other environments and are most often sample inefficient. In this paper, we explore active inference and the free energy principle, a normative theory from neuroscience that explains how self-organizing biological systems operate by maintaining a model of the world and casting action selection as an inference problem. We apply this concept to a typical problem known to the RL community, the mountain car problem, and show how active inference encompasses both RL and learning from demonstrations. | BAYESIAN POLICY SELECTION USING ACTIVE INFERENCE |
d258079367 | Variational autoencoders (VAEs) are powerful generative modelling methods, however they suffer from blurry generated samples and reconstructions compared to the images they have been trained on. Significant research effort has been spent to increase the generative capabilities by creating more flexible models but often flexibility comes at the cost of higher complexity and computational cost. Several works have focused on altering the reconstruction term of the evidence lower bound (ELBO), however, often at the expense of losing the mathematical link to maximizing the likelihood of the samples under the modeled distribution. Here we propose a new formulation of the reconstruction term for the VAE that specifically penalizes the generation of blurry images while at the same time still maximizing the ELBO under the modeled distribution. We show the potential of the proposed loss on three different data sets, where it outperforms several recently proposed reconstruction losses for VAEs. | EXPLICITLY MINIMIZING THE BLUR ERROR OF VARI- ATIONAL AUTOENCODERS |
d256808752 | The vulnerability of deep neural networks (DNNs) to adversarial examples has attracted great attention in the machine learning community. The problem is related to non-flatness and non-smoothness of normally obtained loss landscapes. Training augmented with adversarial examples (a.k.a., adversarial training) is considered as an effective remedy. In this paper, we highlight that some collaborative examples, nearly perceptually indistinguishable from both adversarial and benign examples yet show extremely lower prediction loss, can be utilized to enhance adversarial training. A novel method is therefore proposed to achieve new state-of-the-arts in adversarial robustness. Code: https://github.com | Published as a conference paper at ICLR 2023 SQUEEZE TRAINING FOR ADVERSARIAL ROBUSTNESS |
d257220031 | Federated Learning (FL) aims to develop a centralized server that learns from distributed clients via communications without accessing the clients' local data. However, existing works mainly focus on federated learning in a single task scenario. with static data. In this paper, we introduce the continual federated learning (CFL) problem, where clients incrementally learn new tasks and history data cannot be stored due to certain reasons, such as limited storage and data retention policy 1 . Generative replay (GR) based methods are effective for continual learning without storing history data. However, we fail when trying to intuitively adapt GR models for this setting. By analyzing the behaviors of clients during training, we find the unstable training process caused by distributed training on non-IID data leads to a notable performance degradation. To address this problem, we propose our FedCIL model with two simple but effective solutions: 1. model consolidation and 2. consistency enforcement. Experimental results on multiple benchmark datasets demonstrate that our method significantly outperforms baselines.Published as a conference paper at ICLR 2023 traditional continual learning settings that only involve one model, our problem is more complex because there are multiple models including one server and many clients. | Published as a conference paper at ICLR 2023 BETTER GENERATIVE REPLAY FOR CONTINUAL FEDERATED LEARNING |
d257019561 | Pretrained language models have shown superior performance on many natural language processing tasks, yet they still struggle at multi-step formal reasoning tasks like grade school math problems. One key challenge of finetuning them to solve such math reasoning problems is that many existing datasets only contain one reference solution for each problem, despite the fact that there are often alternative solutions resembling different reasoning paths to the final answer. This way, the finetuned models are biased towards the limited reference solutions, which limits their generalization to unseen examples. To mitigate this issue, we propose to let the model perform sampling during training and learn from both selfsampled fully-correct solutions, which yield the correct answer upon execution, and partially-correct solutions, whose intermediate state matches an intermediate state of a known correct solution. We show that our use of self-sampled correct and partially-correct solutions can benefit learning and help guide the sampling process, leading to more efficient exploration of the solution space. Additionally, we explore various training objectives to support learning from multiple solutions per example and find they greatly affect the performance. Experiments on two math reasoning datasets show the effectiveness of our method compared to learning from a single reference solution with MLE, where we improve PASS@100 from 35.5% to 44.5% for GSM8K, and 27.6% to 36.2% PASS@80 for MathQA. Such improvements are also consistent across different model sizes. Our code is available at https://github.com/microsoft/TraceCodegen. | Published as a conference paper at ICLR 2023 LEARNING MATH REASONING FROM SELF-SAMPLED CORRECT AND PARTIALLY-CORRECT SOLUTIONS |
d8419990 | We investigate the problem of inducing word embeddings that are tailored for a particular bilexical relation. Our learning algorithm takes an existing lexical vector space and compresses it such that the resulting word embeddings are good predictors for a target bilexical relation. In experiments we show that task-specific embeddings can benefit both the quality and efficiency in lexical prediction tasks. | Under review as a workshop contribution at ICLR 2015 TAILORING WORD EMBEDDINGS FOR BILEXICAL PREDICTIONS: AN EXPERIMENTAL COMPARISON |
d232290796 | HardWare-aware Neural Architecture Search (HW-NAS) has recently gained tremendous attention by automating the design of deep neural networks deployed in more resource-constrained daily life devices. Despite its promising performance, developing optimal HW-NAS solutions can be prohibitively challenging as it requires cross-disciplinary knowledge in the algorithm, micro-architecture, and device-specific compilation. First, to determine the hardware-cost to be incorporated into the NAS process, existing works mostly adopt either pre-collected hardware-cost look-up tables or device-specific hardware-cost models. The former can be time-consuming due to the required knowledge of the device's compilation method and how to set up the measurement pipeline, while building the latter is often a barrier for non-hardware experts like NAS researchers. Both of them limit the development of HW-NAS innovations and impose a barrier-to-entry to non-hardware experts. Second, similar to generic NAS, it can be notoriously difficult to benchmark HW-NAS algorithms due to their significant required computational resources and the differences in adopted search spaces, hyperparameters, and hardware devices. To this end, we develop HW-NAS-Bench, the first public dataset for HW-NAS research which aims to democratize HW-NAS research to non-hardware experts and make HW-NAS research more reproducible and accessible. To design HW-NAS-Bench, we carefully collected the measured/estimated hardware performance (e.g., energy cost and latency) of all the networks in the search spaces of both NAS-Bench-201 and FBNet, on six hardware devices that fall into three categories (i.e., commercial edge devices, FPGA, and ASIC). Furthermore, we provide a comprehensive analysis of the collected measurements in HW-NAS-Bench to provide insights for HW-NAS research. Finally, we demonstrate exemplary user cases to (1) show that HW-NAS-Bench allows non-hardware experts to perform HW-NAS by simply querying our premeasured dataset and (2) verify that dedicated device-specific HW-NAS can indeed lead to optimal accuracy-cost trade-offs. The codes and all collected data are available at https://github.com/RICE-EIC/HW-NAS-Bench. | HW-NAS-BENCH: HARDWARE-AWARE NEURAL AR- CHITECTURE SEARCH BENCHMARK |
d67856619 | Transfer learning aims to solve the data sparsity for a target domain by applying information of the source domain. Given a sequence (e.g. a natural language sentence), the transfer learning, usually enabled by recurrent neural network (RNN), represents the sequential information transfer. RNN uses a chain of repeating cells to model the sequence data. However, previous studies of neural network based transfer learning simply represents the whole sentence by a single vector, which is unfeasible for seq2seq and sequence labeling. Meanwhile, such layer-wise transfer learning mechanisms lose the fine-grained cell-level information from the source domain. In this paper, we proposed the aligned recurrent transfer, ART, to achieve celllevel information transfer. ART is under the pre-training framework. Each cell attentively accepts transferred information from a set of positions in the source domain. Therefore, ART learns the cross-domain word collocations in a more flexible way. We conducted extensive experiments on both sequence labeling tasks (POS tagging, NER) and sentence classification (sentiment analysis). ART outperforms the state-of-the-arts over all experiments. | TRANSFER LEARNING FOR SEQUENCES VIA LEARN- ING TO COLLOCATE |
d237940779 | In NLP, a large volume of tasks involve pairwise comparison between two sequences (e.g., sentence similarity and paraphrase identification). Predominantly, two formulations are used for sentence-pair tasks: bi-encoders and cross-encoders. Bi-encoders produce fixed-dimensional sentence representations and are computationally efficient, however, they usually underperform cross-encoders. Crossencoders can leverage their attention heads to exploit inter-sentence interactions for better performance but they require task finetuning and are computationally more expensive. In this paper, we present a completely unsupervised sentence-pair model termed as TRANS-ENCODER that combines the two learning paradigms into an iterative joint framework to simultaneously learn enhanced bi-and crossencoders. Specifically, on top of a pre-trained language model (PLM), we start with converting it to an unsupervised bi-encoder, and then alternate between the bi-and cross-encoder task formulations. In each alternation, one task formulation will produce pseudo-labels which are used as learning signals for the other task formulation. We then propose an extension to conduct such self-distillation approach on multiple PLMs in parallel and use the average of their pseudo-labels for mutual-distillation. TRANS-ENCODER creates, to the best of our knowledge, the first completely unsupervised cross-encoder and also a state-of-the-art unsupervised bi-encoder for sentence similarity. Both the bi-encoder and cross-encoder formulations of TRANS-ENCODER outperform recently proposed state-of-the-art unsupervised sentence encoders such as Mirror- BERT (Liu et al., 2021) and SimCSE (Gao et al., 2021) by up to 5% on the sentence similarity benchmarks. Code and models are released at https://github.com/amzn/trans-encoder. Figure 1: A graphical illustration of the self-distillation learning scheme in TRANS-ENCODER. Notice that the blue boxes represent the same model architecture trained sequentially. explicitly model the interactions between sentences but could only compare them in the embedding space in a post hoc manner.In this work, we ask the question: can we leverage the advantages of both bi-and cross-encoders and bootstrap knowledge from them in an unsupervised manner? Our proposed TRANS-ENCODER addresses this question with the following intuition: As a starting point, we can use bi-encoder representations to tune a cross-encoder. With more powerful inter-sentence modelling, the crossencoder should resurface more knowledge from the PLMs than the bi-encoder given the same data. In turn, the more powerful cross-encoder can distil its knowledge back to the bi-encoder. We can repeat this cycle to iteratively bootstrap from both the bi-and cross-encoders.TRANS-ENCODERThe general idea of TRANS-ENCODER is simple yet extremely effective. In §2.1, we first transform an off-the-shelf PLM to a strong bi-encoder, serving as an initialisation point. Then, the bi-encoder produces pseudo-labels and the PLM subsequently learns from these pseudo-labels in a cross-encoder manner ( §2.2). Consecutively, the cross-encoder further produces more accurate pseudo-labels for bi-encoder learning ( §2.3). This self-distillation process is visualised inFig. 1. Then in, §2.4, we propose a further extension called mutual-distillation that stabilises the training process and boosts the encoder performance even more.TRANSFORM PLMS INTO EFFECTIVE BI-ENCODERSOff-the-shelf PLMs are unsatisfactory bi-encoders. 1 To have a reasonably good starting point, we leverage a simple contrastive tuning procedure to transform existing PLMs to bi-encoders. This approach is concurrently proposed in both(Gao et al., 2021). | TRANS-ENCODER: UNSUPERVISED SENTENCE-PAIR MODELLING THROUGH SELF-AND MUTUAL-DISTILLATIONS |
d259298795 | We show that participating in federated learning can be detrimental to group fairness. In fact, the bias of a few parties against under-represented groups (identified by sensitive attributes such as gender or race) can propagate through the network to all the parties in the network. We analyze and explain bias propagation in federated learning on naturally partitioned real-world datasets. Our analysis reveals that biased parties unintentionally yet stealthily encode their bias in a small number of model parameters, and throughout the training, they steadily increase the dependence of the global model on sensitive attributes. What is important to highlight is that the experienced bias in federated learning is higher than what parties would otherwise encounter in centralized training with a model trained on the union of all their data. This indicates that the bias is due to the algorithm. Our work calls for auditing group fairness in federated learning and designing learning algorithms that are robust to bias propagation. | Published as a conference paper at ICLR 2023 BIAS PROPAGATION IN FEDERATED LEARNING |
d235421834 | Randomized Smoothing (RS) is a promising method for obtaining robustness certificates by evaluating a base model under noise. In this work, we: (i) theoretically motivate why ensembles are a particularly suitable choice as base models for RS, and (ii) empirically confirm this choice, obtaining state-of-the-art results in multiple settings. The key insight of our work is that the reduced variance of ensembles over the perturbations introduced in RS leads to significantly more consistent classifications for a given input. This, in turn, leads to substantially increased certifiable radii for samples close to the decision boundary. Additionally, we introduce key optimizations which enable an up to 55-fold decrease in sample complexity of RS for predetermined radii, thus drastically reducing its computational overhead. Experimentally, we show that ensembles of only 3 to 10 classifiers consistently improve on their strongest constituting model with respect to their average certified radius (ACR) by 5% to 21% on both CIFAR10 and ImageNet, achieving a new state-of-theart ACR of 0.86 and 1.11, respectively. We release all code and models required to reproduce our results at https. Enhancing certified robustness of smoothed classifiers via weighted model ensembling. ArXiv preprint, abs/2005.09363, 2020.Hongbin Liu, Jinyuan Jia, and Neil Zhenqiang Gong. Pointguard: Provably robust 3d point cloud | Published as a conference paper at ICLR 2022 BOOSTING RANDOMIZED SMOOTHING WITH VARIANCE REDUCED CLASSIFIERS |
d195833273 | This paper studies the expressive power of graph neural networks falling within the message-passing framework (GNN mp ). Two results are presented. First, GNN mp are shown to be Turing universal under sufficient conditions on their depth, width, node attributes, and layer expressiveness. Second, it is discovered that GNN mp can lose a significant portion of their power when their depth and width is restricted. The proposed impossibility statements stem from a new technique that enables the repurposing of seminal results from distributed computing and leads to lower bounds for an array of decision, optimization, and estimation problems involving graphs. Strikingly, several of these problems are deemed impossible unless the product of a GNN mp 's depth and width exceeds a polynomial of the graph size; this dependence remains significant even for tasks that appear simple or when considering approximation. near-quadratic lower bounds for the congest model. arXiv preprint arXiv:1705.05646, 2017. | Published as a conference paper at ICLR 2020 WHAT GRAPH NEURAL NETWORKS CANNOT LEARN: DEPTH VS WIDTH |
d252734947 | Despite the use of large vision and language models (VLMs) in many downstream applications, it is unclear how well they encode the compositional relationships between objects and attributes. Here, we create the Attribution, Relation, and Order (ARO) benchmark to systematically evaluate the ability of VLMs to understand different types of relationships, attributes, and order information. ARO consists of Visual Genome Attribution, to test the understanding of objects' properties; Visual Genome Relation, to test for relational understanding; and COCO-Order & Flickr30k-Order, to test for order sensitivity in VLMs. ARO is orders of magnitude larger than previous benchmarks of compositionality, with more than 50,000 test cases. We present the settings in which state-of-the-art VLMs behave like bagsof-words-i.e. when they have poor relational understanding, can blunder when linking objects to their attributes, and demonstrate a severe lack of order sensitivity. VLMs are predominantly trained and evaluated on large scale datasets with rich compositional structure in the images and captions. Yet, training on these datasets has not been enough to address the lack of compositional understanding, and evaluating on these datasets has failed to surface this deficiency. To understand why these limitations emerge and are not represented in the standard tests, we zoom into the training and evaluation procedures. We demonstrate that it is possible to perform well on image-text retrieval over existing datasets without using the composition and order information. This further motivates the value of using ARO to benchmark VLMs. Given that contrastive pretraining optimizes for retrieval on large datasets with similar shortcuts, we hypothesize that this can explain why the models do not need to learn to represent compositional information. This finding suggests a natural solution: composition-aware hard negative mining. We show that a simple-to-implement modification of contrastive learning significantly improves the performance on tasks requiring an understanding of order and compositionality. | Published as a conference paper at ICLR 2023 WHEN AND WHY VISION-LANGUAGE MODELS BE- HAVE LIKE BAGS-OF-WORDS, AND WHAT TO DO ABOUT IT? |
d222142219 | Recent increases in the computational demands of deep neural networks (DNNs), combined with the observation that most input samples require only simple models, have sparked interest in input-adaptive multi-exit architectures, such as MSDNets or Shallow-Deep Networks. These architectures enable faster inferences and could bring DNNs to low-power devices, e.g. in the Internet of Things (IoT). However, it is unknown if the computational savings provided by this approach are robust against adversarial pressure. In particular, an adversary may aim to slow down adaptive DNNs by increasing their average inference time-a threat analogous to the denial-of-service attacks from the Internet. In this paper, we conduct a systematic evaluation of this threat by experimenting with three generic multi-exit DNNs (based on VGG16, MobileNet, and ResNet56) and a custom multi-exit architecture, on two popular image classification benchmarks (CIFAR-10 and Tiny ImageNet). To this end, we show that adversarial sample-crafting techniques can be modified to cause slowdown, and we propose a metric for comparing their impact on different architectures. We show that a slowdown attack reduces the efficacy of multi-exit DNNs by 90%-100%, and it amplifies the latency by 1.5-5× in a typical IoT deployment. We also show that it is possible to craft universal, reusable perturbations and that the attack can be effective in realistic black-box scenarios, where the attacker has limited knowledge about the victim. Finally, we show that adversarial training provides limited protection against slowdowns. These results suggest that further research is needed for defending multi-exit architectures against this emerging threat. * Indicates equal contribution, ordered alphabetically. | A Panda? No, It's a Sloth: Slowdown Attacks on Adaptive Multi-Exit Neural Network Inference A PANDA? NO, IT'S A SLOTH: SLOWDOWN ATTACKS ON ADAPTIVE MULTI-EXIT NEURAL NETWORK INFERENCE |
d53366105 | Published as a conference paper at ICLR 2019 | |
d232307121 | Machine learning (ML) models that learn and predict properties of computer programs are increasingly being adopted and deployed. In this work, we investigate principled ways to adversarially perturb a computer program to fool such learned models, and thus determine their adversarial robustness. We use program obfuscations, which have conventionally been used to avoid attempts at reverse engineering programs, as adversarial perturbations. These perturbations modify programs in ways that do not alter their functionality but can be crafted to deceive an ML model when making a decision. We provide a general formulation for an adversarial program that allows applying multiple obfuscation transformations to a program in any language. We develop first-order optimization algorithms to efficiently determine two key aspects -which parts of the program to transform, and what transformations to use. We show that it is important to optimize both these aspects to generate the best adversarially perturbed program. Due to the discrete nature of this problem, we also propose using randomized smoothing to improve the attack loss landscape to ease optimization. We evaluate our work on Python and Java programs on the problem of program summarization. 1 We show that our best attack proposal achieves a 52% improvement over a state-of-the-art attack generation approach for programs trained on a SEQ2SEQ model. We further show that our formulation is better at training models that are robust to adversarial attacks. | Published as a conference paper at ICLR 2021 GENERATING ADVERSARIAL COMPUTER PROGRAMS USING OPTIMIZED OBFUSCATIONS |
d252968113 | We introduce Token Merging (ToMe), a simple method to increase the throughput of existing ViT models without needing to train. ToMe gradually combines similar tokens in a transformer using a general and light-weight matching algorithm that is as fast as pruning while being more accurate. Off-the-shelf, ToMe can 2× the throughput of state-of-the-art ViT-L @ 512 and ViT-H @ 518 models on images and 2.2× the throughput of ViT-L on video with only a 0.2-0.3% accuracy drop in each case. ToMe can also easily be applied during training, improving in practice training speed up to 2× for MAE fine-tuning on video. Training with ToMe further minimizes accuracy drop, leading to 2× the throughput of ViT-B on audio for only a 0.4% mAP drop. Qualitatively, we find that ToMe merges object parts into one token, even over multiple frames of video. Overall, ToMe's accuracy and speed are competitive with state-of-the-art on images, video, and audio. * Work done during an internship at Meta AI. Code at . Flashattention: Fast and memory-efficient exact attention with io-awareness. arXiv:2205.14135 [cs.LG], 2022. Natsev, et al. The kinetics human action video dataset. arXiv:1705.06950 [cs.CV], 2017. | Published as a conference paper at ICLR 2023 TOKEN MERGING: YOUR VIT BUT FASTER |
d233394003 | Large optimization problems with hard constraints arise in many settings, yet classical solvers are often prohibitively slow, motivating the use of deep networks as cheap "approximate solvers." Unfortunately, naive deep learning approaches typically cannot enforce the hard constraints of such problems, leading to infeasible solutions. In this work, we present Deep Constraint Completion and Correction (DC3), an algorithm to address this challenge. Specifically, this method enforces feasibility via a differentiable procedure, which implicitly completes partial solutions to satisfy equality constraints and unrolls gradient-based corrections to satisfy inequality constraints. We demonstrate the effectiveness of DC3 in both synthetic optimization tasks and the real-world setting of AC optimal power flow, where hard constraints encode the physics of the electrical grid. In both cases, DC3 achieves near-optimal objective values while preserving feasibility. * These authors contributed equally.Published as a conference paper at ICLR 2021 • AC optimal power flow. We show how the general DC3 framework can be used to optimize power flows on the electrical grid. This difficult non-convex optimization task must be solved at scale and is especially critical for renewable energy adoption. Our results greatly improve upon the performance of general-purpose deep learning methods on this task.RELATED WORKOur approach is situated within the broader literature on fast optimization methods, and draws inspiration from literature on implicit layers and on incorporating constraints into neural networks. We briefly describe each of these areas and their relationship to the present work.Fast optimization methods. Many classical optimization methods have been proposed to improve the practical efficiency of solving optimization problems. These include general techniques such as constraint and variable elimination (i.e., the removal of non-active constraints or redundant variables, respectively), as well as problem-specific techniques (e.g., KKT factorization techniques in the case of convex quadratic programs)(Nocedal & Wright, 2006). Our present work builds upon aspects of this literature, applying concepts from variable elimination to reduce the number of degrees of freedom associated with the optimization problems we wish to solve.In addition to the classical optimization literature, there has been a large body of literature in deep learning that has sought to approximate or speed up optimization models. As described in reviews on topics such as combinatorial optimization (Bengio et al., 2020) and optimal power flow (Hasan et al., 2020), ML methods to speed up optimization models have thus far taken two main approaches. The first class of approaches, akin to work on surrogate modeling(Koziel & Leifsson, 2013), has involved training machine learning models to map directly from optimization inputs to full solutions. However, such approaches have often struggled to produce solutions that are both feasible and (near-)optimal. The second class of approaches has instead focused on employing machine learning approaches alongside or in the loop of optimization models, e.g., to learn warm-start points (see, e.g., Baker (2019) and Dong et al.(2020)) or to enable constraint elimination techniques by predicting active constraints (see, e.g., Misra et al.(2018)). We view our work as part of the former set of approaches, but drawing important inspiration from the latter: that employing structural knowledge about the optimization model is paramount to achieving both feasibility and optimality.Constraints in neural networks. While deep learning is often thought of as wholly unconstrained, in reality, it is quite common to incorporate (simple) constraints within deep learning procedures. For instance, softmax layers encode simplex constraints, sigmoids instantiate upper and lower bounds, ReLUs encode projections onto the positive orthant, and convolutional layers enforce translational equivariance (an idea taken further in general group-equivariant networks (Cohen & Welling, 2016)). Recent work has also focused on embedding specialized kinds of constraints into neural networks, such as conservation of energy (see, e.g., Greydanus et al.(2019)and Beucler et al.(2019)), and homogeneous linear inequality constraints(Frerix et al., 2020). However, while these represent common "special cases," there has to date been little work on building more general hard constraints into deep learning models. | Published as a conference paper at ICLR 2021 DC3: A LEARNING METHOD FOR OPTIMIZATION WITH HARD CONSTRAINTS |
d253107402 | Message Passing Neural Networks (MPNNs) are a widely used class of Graph Neural Networks (GNNs). The limited representational power of MPNNs inspires the study of provably powerful GNN architectures. However, knowing one model is more powerful than another gives little insight about what functions they can or cannot express. It is still unclear whether these models are able to approximate specific functions such as counting certain graph substructures, which is essential for applications in biology, chemistry and social network analysis. Motivated by this, we propose to study the counting power of Subgraph MPNNs, a recent and popular class of powerful GNN models that extract rooted subgraphs for each node, assign the root node a unique identifier and encode the root node's representation within its rooted subgraph. Specifically, we prove that Subgraph MPNNs fail to count more-than-4-cycles at node level, implying that node representations cannot correctly encode the surrounding substructures like ring systems with more than four atoms. To overcome this limitation, we propose I 2 -GNNs to extend Subgraph MPNNs by assigning different identifiers for the root node and its neighbors in each subgraph. I 2 -GNNs' discriminative power is shown to be strictly stronger than Subgraph MPNNs and partially stronger than the 3-WL test. More importantly, I 2 -GNNs are proven capable of counting all 3, 4, 5 and 6-cycles, covering common substructures like benzene rings in organic chemistry, while still keeping linear complexity. To the best of our knowledge, it is the first linear-time GNN model that can count 6-cycles with theoretical guarantees. We validate its counting power in cycle counting tasks and demonstrate its competitive performance in molecular prediction benchmarks.arXiv:2210.13978v3 [cs.LG] 8 May 2023Published as a conference paper at ICLR 2023 test's power of counting general graph substructures. Graph substructures are important as they are closely related to tasks in chemistry[50,51,52], biology[54]and social network analysis[53]. Particularly, cycles play an essential role in organic chemistry. Different types of rings impact the compounds' stability, aromaticity and other chemical properties. Therefore, studying the approximating power of counting substructures, especially cycles, provides a fine-grained and intuitive description of models' representational power and gives insight to real-world practices.Nevertheless, the difficulty of counting cycles is usually underestimated. Although[8]claim that ID-GNNs can count arbitrary cycles at node level, the proof turns out to be incorrect, since it confuses walks with paths (a cycle is a closed path without repeated nodes while walks allow repeated nodes). In fact, even powerful 2-FWL test with cubic complexity can only count up to 7-cycles[12,11]. The difficulty makes us question whether existing powerful models, such as ID-GNNs, can count cycles properly.ID-GNNs can be categorized into a new class of GNNs named Subgraph GNNs[28,32,7,8,29,30]. The core idea is to decompose a graph into a bag of subgraphs and encode the graph by aggregating subgraph representations, though the strategy of extracting subgraphs varies. See [9, 55] for detailed discussions. Subgraph GNNs have demonstrated their impressive performance by achieving stateof-the-art results on multiple open benchmarks. Theoretically, the discriminative power of existing Subgraph GNNs is known to be strictly stronger than WL test and weaker than 3-WL test[9]. However, it is fair to say we still do not know the approximation power of Subgraph GNNs in terms of counting substructures.Main contributions. In our work, we propose to study the representational power of Subgraph GNNs via the ability to count a specific class of substructures-cycles and paths, because they are the bases to represent some important substructures such as ring systems in chemistry. We focus on Subgraph MPNNs, a subclass of Subgraph GNNs covering[28,7,8]. Our main contribution include • We prove that Subgraph MPNNs can count 3-cycles and 4-cycles, but cannot count 5-cycle or any longer cycles at node level. This result is unsatisfying because only a small portion of ring systems are 4-cycles. It also negates the previous proposition that ID-GNNs can use node representations to count arbitrary cycles[8]. | Published as a conference paper at ICLR 2023 BOOSTING THE CYCLE COUNTING POWER OF GRAPH NEURAL NETWORKS WITH I 2 -GNNS |
d3619097 | Deep reinforcement learning has demonstrated increasing capabilities for continuous control problems, including agents that can move with skill and agility through their environment. An open problem in this setting is that of developing good strategies for integrating or merging policies for multiple skills, where each individual skill is a specialist in a specific skill and its associated state distribution. We extend policy distillation methods to the continuous action setting and leverage this technique to combine expert policies, as evaluated in the domain of simulated bipedal locomotion across different classes of terrain. We also introduce an input injection method for augmenting an existing policy network to exploit new input features. Lastly, our method uses transfer learning to assist in the efficient acquisition of new skills. The combination of these methods allows a policy to be incrementally augmented with new skills. We compare our progressive learning and integration via distillation (PLAID) method against three alternative baselines. * These authors contributed equally to this work. | Published as a conference paper at ICLR 2018 PROGRESSIVE REINFORCEMENT LEARNING WITH DISTILLATION FOR MULTI-SKILLED MOTION CONTROL |
d202712891 | We study the problem of defending deep neural network approaches for image classification from physically realizable attacks. First, we demonstrate that the two most scalable and effective methods for learning robust models, adversarial training with PGD attacks and randomized smoothing, exhibit limited effectiveness against three of the highest profile physical attacks. Next, we propose a new abstract adversarial model, rectangular occlusion attacks, in which an adversary places a small adversarially crafted rectangle in an image, and develop two approaches for efficiently computing the resulting adversarial examples. Finally, we demonstrate that adversarial training using our new attack yields image classification models that exhibit high robustness against the physically realizable attacks we study, offering the first effective generic defense against such attacks. 1 | Published as a conference paper at ICLR 2020 DEFENDING AGAINST PHYSICALLY REALIZABLE AT- TACKS ON IMAGE CLASSIFICATION |
d170078913 | Neural networks have succeeded in many reasoning tasks. Empirically, these tasks require specialized network structures, e.g., Graph Neural Networks (GNNs) perform well on many such tasks, but less structured networks fail. Theoretically, there is limited understanding of why and when a network structure generalizes better than others, although they have equal expressive power. In this paper, we develop a framework to characterize which reasoning tasks a network can learn well, by studying how well its computation structure aligns with the algorithmic structure of the relevant reasoning process. We formally define this algorithmic alignment and derive a sample complexity bound that decreases with better alignment. This framework offers an explanation for the empirical success of popular reasoning models, and suggests their limitations. As an example, we unify seemingly different reasoning tasks, such as intuitive physics, visual question answering, and shortest paths, via the lens of a powerful algorithmic paradigm, dynamic programming (DP). We show that GNNs align with DP and thus are expected to solve these tasks. On several reasoning tasks, our theory is supported by empirical results.However, there is limited understanding of the relation between the generalization ability and network structure for reasoning. What tasks can a neural network (sample efficiently) learn to reason about? Answering this question is crucial for understanding the empirical success and limitations of existing models, and for designing better models for new reasoning tasks.This paper is an initial work towards answering this fundamental question, by developing a theoretical framework to characterize what tasks a neural network can reason about. We build on a simple observation that reasoning processes resemble algorithms. Hence, we study how well a reasoning algorithm aligns with the computation graph of the network. Intuitively, if they align well, the network only needs to learn simple algorithm steps to simulate the reasoning process, which leads to better sample efficiency. We formalize this intuition with a numeric measure of algorithmic alignment, 1 | Published as a conference paper at ICLR 2020 WHAT CAN NEURAL NETWORKS REASON ABOUT? |
d256827816 | Modern semantic segmentation methods devote much effect to adjusting image feature representations to improve the segmentation performance in various ways, such as architecture design, attention mechnism, etc. However, almost all those methods neglect the particularity of class weights (in the classification layer) in segmentation models. In this paper, we notice that the class weights of categories that tend to share many adjacent boundary pixels lack discrimination, thereby limiting the performance. We call this issue Boundary-caused Class Weights Confusion (BCWC). We try to focus on this problem and propose a novel method named Embedded Conditional Random Field (E-CRF) to alleviate it. E-CRF innovatively fuses the CRF into the CNN network as an organic whole for more effective end-to-end optimization. The reasons are two folds. It utilizes CRF to guide the message passing between pixels in high-level features to purify the feature representation of boundary pixels, with the help of inner pixels belonging to the same object. More importantly, it enables optimizing class weights from both scale and direction during backpropagation. We make detailed theoretical analysis to prove it. Besides, superpixel is integrated into E-CRF and served as an auxiliary to exploit the local object prior for more reliable message passing. Finally, our proposed method yields impressive results on ADE20K, Cityscapes, and Pascal Context datasets.Published as a conference paper at ICLR 2023 Previous works mainly aim to improve boundary pixel segmentation, but they seldom explicitly take class weights confusion i.e., BCWC, into consideration 1 . | E-CRF: EMBEDDED CONDITIONAL RANDOM FIELD FOR BOUNDARY-CAUSED CLASS WEIGHTS CONFU- SION IN SEMANTIC SEGMENTATION |
d209476324 | In this paper, we propose an end-to-end deep learning model, called E2Efold, for RNA secondary structure prediction which can effectively take into account the inherent constraints in the problem. The key idea of E2Efold is to directly predict the RNA base-pairing matrix, and use an unrolled algorithm for constrained programming as the template for deep architectures to enforce constraints. With comprehensive experiments on benchmark datasets, we demonstrate the superior performance of E2Efold: it predicts significantly better structures compared to previous SOTA | Published as a conference paper at ICLR 2020 RNA SECONDARY STRUCTURE PREDICTION BY LEARNING UNROLLED ALGORITHMS |
d252780545 | Recent advances in pre-training vision-language models like CLIP(Radford et al., 2021)have shown great potential in learning transferable visual representations. Nonetheless, for downstream inference, CLIP-like models suffer from either 1) degraded accuracy and robustness in the case of inaccurate text descriptions during retrieval-based inference (the challenge for zero-shot protocol); or 2) breaking the well-established vision-language alignment (the challenge for linear probing). To address them, we propose Decomposed Feature Prompting (DeFo). DeFo leverages a flexible number of learnable embeddings as textual input while maintaining the vision-language dual-model architecture, which enables the model to learn decomposed visual features with the help of feature-level textual prompts. We further use an additional linear layer to perform classification, allowing a scalable size of language inputs. Our empirical study shows DeFo's significance in improving the vision-language models. For example, DeFo obtains 73.2% test accuracy on ImageNet with a ResNet-50 backbone without tuning any pretrained weights of both the vision and language encoder, outperforming zero-shot CLIP by a large margin of 15.0%, and outperforming state-of-the-art vision-language prompt tuning method by 7.6%. | Under review as a conference paper at ICLR 2023 LEARNING TO DECOMPOSE VISUAL FEATURES WITH LATENT TEXTUAL PROMPTS |
d232232790 | Motivated by the rising abundance of observational data with continuous treatments, we investigate the problem of estimating the average dose-response curve (ADRF). Available parametric methods are limited in their model space, and previous attempts in leveraging neural network to enhance model expressiveness relied on partitioning continuous treatment into blocks and using separate heads for each block; this however produces in practice discontinuous ADRFs. Therefore, the question of how to adapt the structure and training of neural network to estimate ADRFs remains open. This paper makes two important contributions. First, we propose a novel varying coefficient neural network (VCNet) that improves model expressiveness while preserving continuity of the estimated ADRF. Second, to improve finite sample performance, we generalize targeted regularization to obtain a doubly robust estimator of the whole ADRF curve. * Equal contribution and corresponding authors. Small. Non-parametric methods for doubly robust estimation of continuous treatment effects. 's Disease Neuroimaging Initiative. A functional varying-coefficient single-index model for functional response data. . Causal effect inference with deep latent-variable models. . Stacked convolutional autoencoders for hierarchical feature extraction. In International conference on artificial neural networks, pp. 52-59. Springer, 2011. David Newman. Bag of words data set, 2008. BN Prichard and PM Gillam. Assessment of propranolol in angina pectoris. clinical dose response curve and effect on electrocardiogram at rest and on exercise. British heart journal, 33(4):473, 1971. Danilo Jimenez Rezende and Shakir Mohamed. Variational inference with normalizing flows. arXiv preprint arXiv:1505.05770, 2015. . A hypercube-based encoding for evolving large-scale neural networks. Artificial life, 15(2):185-212, 2009. Timothy J Threlfall and Dallas R English. Sun exposure and pterygium of the eye: a dose-response curve. American journal of ophthalmology, 128(3):280-287, 1999. | Published as a conference paper at ICLR 2021 VCNET AND FUNCTIONAL TARGETED REGULARIZA- TION FOR LEARNING CAUSAL EFFECTS OF CONTINU- OUS TREATMENTS |
d211132680 | We address the problem of discovering 3D parts for objects in unseen categories. | Published as a conference paper at ICLR 2020 LEARNING TO GROUP: A BOTTOM-UP FRAMEWORK FOR 3D PART DISCOVERY IN UNSEEN CATEGORIES |
d238634570 | Normalizing flows have shown great success as general-purpose density estimators. However, many real world applications require the use of domain-specific knowledge, which normalizing flows cannot readily incorporate. We propose embedded-model flows (EMF), which alternate general-purpose transformations with structured layers that embed domain-specific inductive biases. These layers are automatically constructed by converting user-specified differentiable probabilistic models into equivalent bijective transformations. We also introduce gated structured layers, which allow bypassing the parts of the models that fail to capture the statistics of the data. We demonstrate that EMFs can be used to induce desirable properties such as multimodality and continuity. Furthermore, we show that EMFs enable a high performance form of variational inference where the structure of the prior model is embedded in the variational architecture. In our experiments, we show that this approach outperforms a large number of alternative methods in common structured inference problems. | EMBEDDED-MODEL FLOWS: COMBINING THE INDUC- TIVE BIASES OF MODEL-FREE DEEP LEARNING AND EXPLICIT PROBABILISTIC MODELING |
d256808454 | The transferability of adversarial examples across deep neural networks (DNNs) is the crux of many black-box attacks. Many prior efforts have been devoted to improving the transferability via increasing the diversity in inputs of some substitute models. In this paper, by contrast, we opt for the diversity in substitute models and advocate to attack a Bayesian model for achieving desirable transferability. Deriving from the Bayesian formulation, we develop a principled strategy for possible finetuning, which can be combined with many off-the-shelf Gaussian posterior approximations over DNN parameters. Extensive experiments have been conducted to verify the effectiveness of our method, on common benchmark datasets, and the results demonstrate that our method outperforms recent state-of-the-arts by large margins (roughly 19% absolute increase in average attack success rate on Ima-geNet), and, by combining with these recent methods, further performance gain can be obtained. Our code: https://github.com/qizhangli/MoreBayesian-attack. | Published as a conference paper at ICLR 2023 MAKING SUBSTITUTE MODELS MORE BAYESIAN CAN ENHANCE TRANSFERABILITY OF ADVERSARIAL EX- AMPLES |
d246210292 | Information sharing is key in building team cognition and enables coordination and cooperation. High-performing human teams also benefit from acting strategically with hierarchical levels of iterated communication and rationalizability, meaning a human agent can reason about the actions of their teammates in their decisionmaking. Yet, the majority of prior work in Multi-Agent Reinforcement Learning (MARL) does not support iterated rationalizability and only encourage inter-agent communication, resulting in a suboptimal equilibrium cooperation strategy. In this work, we show that reformulating an agent's policy to be conditional on the policies of its neighboring teammates inherently maximizes Mutual Information (MI) lower-bound when optimizing under Policy Gradient (PG). Building on the idea of decision-making under bounded rationality and cognitive hierarchy theory, we show that our modified PG approach not only maximizes local agent rewards but also implicitly reasons about MI between agents without the need for any explicit ad-hoc regularization terms. Our approach, InfoPG, outperforms baselines in learning emergent collaborative behaviors and sets the state-of-the-art in decentralized cooperative MARL tasks. Our experiments validate the utility of InfoPG by achieving higher sample efficiency and significantly larger cumulative reward in several complex cooperative multi-agent domains. * Co-first authors. These authors contributed equally to this work.arXiv:2201.08484v4 [cs.MA] 24 Jun 2022Published as a conference paper at ICLR 2022 agent and the collective goal is to maximize the globally averaged return over all agents. Nevertheless, under an F-Dec setting, agents seek to maximize their own reward, which does not necessarily imply the maximization of the team long-term return since agents do not inherently understand coordination.Recently, strong empirical evidence has shown that Mutual Information (MI) is a statistic that correlates with the degree of collaboration between pairs of agents (Trendafilov et al., 2015). Researchers have shown that information redundancy is minimized among agents by maximizing the joint entropy of agents' decisions, which in turn, improves the overall performance in MARL (Malakar et al., 2012). Therefore, recent work in MARL has sought to integrate entropy regularization terms as means of maximizing MI among interacting agents (Kim et al., 2020; Wang et al., 2019;Jaques et al., 2019). The formulaic calculation of MI relies upon the estimation of action-conditional distributions. In most prior work, agents are equipped with conventional state-conditional policies, and researchers employ techniques, such as variational inference, for estimating and optimizing an action-conditional policy distribution to quantify MI(Wen et al., 2019;Kim et al., 2020). However, agents are not explicitly given the ability to reason about their teammates' action-decisions and, instead, have to learn implicitly from sparse rewards or hand-engineered regularization and auxiliary loss terms.Contributions -In this work, we propose a novel information-theoretic, fully-distributed cooperative MARL framework, called InfoPG, by reformulating an agent's policy to be directly conditional on the policies of its instantaneous neighbors during Policy Gradient (PG) optimization. We study cooperative MARL under the assumption of bounded rational agents and leverage action-conditional policies into PG objective function to accommodate our assumption. By leveraging the k-level reasoning (Ho & Su, 2013) paradigm from cognitive hierarchy theory, we propose a cooperative MARL framework in which naive, nonstrategic agents are improved to sophisticated agents that iteratively reason about the rationality of their teammates for decision-making. InfoPG implicitly increases MI among agents' k-level action-conditional policies to promote cooperativity. To learn collaborative behavior, we build InfoPG on a communicative fully-decentralized structure where agents learn to achieve consensus in their actions and maximize their shared utility by communicating with their physical neighbors over a potentially time-varying communication graph. We show the effectiveness of InfoPG across multiple, complex cooperative environments by empirically assessing its performance against several baselines. The primary contributions of our work are as follows:1. We derive InfoPG, an information-theoretic PG framework that leverages cognitive hierarchy and action-conditional policies for maximizing MI among agents and maximizing agents' individual rewards. We derive an analytical lower-bond for MI estimated during InfoPG and provide mathematical reasoning underlying InfoPG's performance.2. We propose a fully-decentralized graph-based communication and k-level reasoning structure to enable theory of mind for coordinating agents and maximizing their shared utility.3. We propose a generalized variant of InfoPG and derive an MI upper-bound to modulate MI among agents depending on cooperativity of agents and environment feedback. We demonstrate the utility of this generalization in solving an instance of the Byzantine Generals Problem (BGP), in a fully decentralized setting.4. We present quantitative results that show InfoPG sets the SOTA performance in learning emergent cooperative behaviors by converging faster and accumulating higher team rewards. | Published as a conference paper at ICLR 2022 ITERATED REASONING WITH MUTUAL INFORMATION IN COOPERATIVE AND BYZANTINE DECENTRALIZED TEAMING |
d259108523 | This paper aims to deal with the ignored real-world complexities in prior work on human motion forecasting, emphasizing the social properties of multi-person motion, the diversity of motion and social interactions, and the complexity of articulated motion. To this end, we introduce a novel task of stochastic multi-person 3D motion forecasting. We propose a dual-level generative modeling framework that separately models independent individual motion at the local level and social interactions at the global level. Notably, this dual-level modeling mechanism can be achieved within a shared generative model, through introducing learnable latent codes that represent intents of future motion and switching the codes' modes of operation at different levels. Our framework is general; we instantiate it with different generative models, including generative adversarial networks and diffusion models, and various multi-person forecasting models. Extensive experiments on CMU-Mocap, MuPoTS-3D, and SoMoF benchmarks show that our approach produces diverse and accurate multi-person predictions, significantly outperforming the state of the art. * Yu-Xiong Wang and Liang-Yan Gui contributed equally to this work. arXiv:2306.05421v1 [cs.CV] 8 Jun 2023 Published as a conference paper at ICLR 2023 (b) Multi-Person Fidelity (a) Single-Person Fidelity (c) Overall Diversity Starting Poses Ending Poses Figure 1: Illustration of the multifaceted challenges in the proposed task of stochastic multi-person 3D motion forecasting. (a) Single-person fidelity: for each person, the predicted pose and trajectory should be realistic and consistent with each other, e.g., to avoid foot floating and skating. (b) Multiperson fidelity: multi-person motion in a scene inherently involves social interactions, e.g., to avoid motion collisions. (c) Overall diversity: long-term human motion is uncertain and stochastic; we address this intrinsic multi-modality, while existing work (Wang et al., 2021b; Adeli et al., 2020; 2021; Guo et al., 2022) simplifies to deterministic prediction. the complexity of articulated motion. As shown in Figure 1, this task requires the predicted multiple motion sequences of multi-person to satisfy the following conditions -(a) single-person fidelity: for example, all single-person motion should be continuous, and the articulated properties should be preserved under the physical rules; (b) multi-person fidelity: the predicted motion should be socially-aware, with the consideration of the interactions between predictions from different people; and (c) overall diversity: the movement of the human body should be as varied as possible, but within the constraints of conditions (a) and (b).Due to the substantially increased complexity of our task, it becomes challenging to optimize all three objectives simultaneously. We observe simply extending existing work such as on deterministic motion forecasting cannot address the proposed task. This difficulty motivates us to adopt a divideand-conquer strategy, together with the observation that single-person fidelity and multi-person fidelity can be viewed as relatively independent goals, while there is an inherent trade-off between fidelity and diversity. Therefore, we propose a Dual-level generative modeling framework for Multi-person Motion Forecasting (DuMMF). At the local level, we model motion for different people independently under relaxed conditions, thus satisfying single-person fidelity and diversity. Meanwhile, at the global level, we model social interactions by considering the correlation between all motion, thereby further improving multi-person fidelity. Notably, this dual-level modeling mechanism can be achieved within a shared generative model, through simply switching the modes of operation of the motion intent codes (i.e., latent codes of the generative model) at different levels. By optimizing these codes with level-specific objectives, we produce diverse and realistic multi-person predictions.Our contributions can be summarized as follows. (a) To the best of our knowledge, we are the first to investigate the task of stochastic multi-person 3D motion forecasting. (b) We propose a simple yet effective dual-level learning framework to address this task. (c) We introduce discrete learnable social intents at dual levels to improve the realism and diversity of predictions. (d) Our framework is general and can be operationalized with various generative models, including generative adversarial networks and diffusion models, and different types of multi-person motion forecasting models. Notably, it can be generalized to challenging more-person (e.g., 18-person) scenarios that are unseen during training.RELATED WORKStochastic Human Motion Forecasting. There have been many advances in stochastic human motion forecasting, many of which (Walker et al., 2017; Yan et al., 2018;Barsoum et al., 2018)are based on the adaptation and improvement of deep generative models such as variational autoencodersPublished as a conference paper at ICLR 2023 | STOCHASTIC MULTI-PERSON 3D MOTION FORECAST- ING |
d67364912 | Recent efforts on combining deep models with probabilistic graphical models are promising in providing flexible models that are also easy to interpret. We propose a variational message-passing algorithm for variational inference in such models. We make three contributions. First, we propose structured inference networks that incorporate the structure of the graphical model in the inference network of variational auto-encoders (VAE). Second, we establish conditions under which such inference networks enable fast amortized inference similar to VAE. Finally, we derive a variational message passing algorithm to perform efficient naturalgradient inference while retaining the efficiency of the amortized inference. By simultaneously enabling structured, amortized, and natural-gradient inference for deep structured models, our method simplifies and generalizes existing methods. * Equal contributions. Wu Lin is now at | Published as a conference paper at ICLR 2018 VARIATIONAL MESSAGE PASSING WITH STRUCTURED INFERENCE NETWORKS |
d220403547 | We investigate two causes for adversarial vulnerability in deep neural networks: bad data and (poorly) trained models. When trained with SGD, deep neural networks essentially achieve zero training error, even in the presence of label noise, while also exhibiting good generalization on natural test data, something referred to as benign overfitting [2, 10]. However, these models are vulnerable to adversarial attacks. We identify label noise as one of the causes for adversarial vulnerability, and provide theoretical and empirical evidence in support of this. Surprisingly, we find several instances of label noise in datasets such as MNIST and CIFAR, and that robustly trained models incur training error on some of these, i.e. they don't fit the noise. However, removing noisy labels alone does not suffice to achieve adversarial robustness. Standard training procedures bias neural networks towards learning "simple" classification boundaries, which may be less robust than more complex ones. We observe that adversarial training does produce more complex decision boundaries. We conjecture that in part the need for complex decision boundaries arises from sub-optimal representation learning. By means of simple toy examples, we show theoretically how the choice of representation can drastically affect adversarial robustness. | How benign is benign overfitting? |
d4109354 | Workshop track -ICLR 2018 COMPARING FIXED AND ADAPTIVE COMPUTATION TIME FOR RECURRENT NEURAL NETWORKS | |
d231855774 | Neural ordinary differential equations (Neural ODEs) are a new family of deeplearning models with continuous depth. However, the numerical estimation of the gradient in the continuous case is not well solved: existing implementations of the adjoint method suffer from inaccuracy in reverse-time trajectory, while the naive method and the adaptive checkpoint adjoint method (ACA) have a memory cost that grows with integration time. In this project, based on the asynchronous leapfrog (ALF) solver, we propose the Memory-efficient ALF Integrator (MALI), which has a constant memory cost w.r.t number of solver steps in integration similar to the adjoint method, and guarantees accuracy in reverse-time trajectory (hence accuracy in gradient estimation). We validate MALI in various tasks: on image recognition tasks, to our knowledge, MALI is the first to enable feasible training of a Neural ODE on ImageNet and outperform a well-tuned ResNet, while existing methods fail due to either heavy memory burden or inaccuracy; for time series modeling, MALI significantly outperforms the adjoint method; and for continuous generative models, MALI achieves new state-of-the-art performance.Code is available at httpsmachine learning framework for solving high-dimensional mean field game and mean field control problems. | MALI: A MEMORY EFFICIENT AND REVERSE ACCU- RATE INTEGRATOR FOR NEURAL ODES |
d10691002 | We show that deep narrow Boltzmann machines are universal approximators of probability distributions on the activities of their visible units, provided they have sufficiently many hidden layers, each containing the same number of units as the visible layer. Besides from this existence statement, we provide upper and lower bounds on the sufficient number of layers and parameters. These bounds show that deep narrow Boltzmann machines are at least as compact universal approximators as restricted Boltzmann machines and narrow sigmoid belief networks, with respect to the currently available bounds for those models. | Deep Narrow Boltzmann Machines are Universal Approximators |
d3515469 | Permutations and matchings are core building blocks in a variety of latent variable models, as they allow us to align, canonicalize, and sort data. Learning in such models is difficult, however, because exact marginalization over these combinatorial objects is intractable. In response, this paper introduces a collection of new methods for end-to-end learning in such models that approximate discrete maximum-weight matching using the continuous Sinkhorn operator. Sinkhorn operator is attractive because it functions as a simple, easy-to-implement analog of the softmax operator. With this, we can define the Gumbel-Sinkhorn method, an extension of the Gumbel-Softmax method(Jang et al., 2016;Maddison et al., 2016)to distributions over latent matchings. We demonstrate the effectiveness of our method by outperforming competitive baselines on a range of qualitatively different tasks: sorting numbers, solving jigsaw puzzles, and identifying neural signals in worms. | Published as a conference paper at ICLR 2018 LEARNING LATENT PERMUTATIONS WITH GUMBEL- SINKHORN NETWORKS |
d247158892 | A standard hardware bottleneck when training deep neural networks is GPU memory. The bulk of memory is occupied by caching intermediate tensors for gradient computation in the backward pass. We propose a novel method to reduce this footprint -Dropping Intermediate Tensors (DropIT). DropIT drops min-k elements of the intermediate tensors and approximates gradients from the sparsified tensors in the backward pass. Theoretically, DropIT reduces noise on estimated gradients and therefore has a higher rate of convergence than vanilla-SGD. Experiments show that we can drop up to 90% of the intermediate tensor elements in fullyconnected and convolutional layers while achieving higher testing accuracy for Visual Transformers and Convolutional Neural Networks on various tasks (e.g. , classification, object detection, instance segmentation). Our code and models are available at https://github.com/chenjoya/dropit. * Equal contribution. | Published as a conference paper at ICLR 2023 DROPIT: DROPPING INTERMEDIATE TENSORS FOR MEMORY-EFFICIENT DNN TRAINING |
d256358479 | Existing Deep Reinforcement Learning (DRL) algorithms suffer from sample inefficiency. Generally, episodic control-based approaches are solutions that leverage highly-rewarded past experiences to improve sample efficiency of DRL algorithms. However, previous episodic control-based approaches fail to utilize the latent information from the historical behaviors (e.g., state transitions, topological similarities, etc.) and lack scalability during DRL training. This work introduces Neural Episodic Control with State Abstraction (NECSA), a simple but effective state abstraction-based episodic control containing a more comprehensive episodic memory, a novel state evaluation, and a multi-step state analysis. We evaluate our approach to the MuJoCo and Atari tasks in OpenAI gym domains. The experimental results indicate that NECSA achieves higher sample efficiency than the state-of-the-art episodic control-based approaches. Our data and code are available at the project website 1 . | Published as a conference paper at ICLR 2023 NEURAL EPISODIC CONTROL WITH STATE ABSTRAC- TION |
d238407848 | Biological spiking neural networks (SNNs) can temporally encode information in their outputs, e.g. in the rank order in which neurons fire, whereas artificial neural networks (ANNs) conventionally do not. As a result, models of SNNs for neuromorphic computing are regarded as potentially more rapid and efficient than ANNs when dealing with temporal input. On the other hand, ANNs are simpler to train, and usually achieve superior performance. Here we show that temporal coding such as rank coding (RC) inspired by SNNs can also be applied to conventional ANNs such as LSTMs, and leads to computational savings and speedups. In our RC for ANNs, we apply backpropagation through time using the standard real-valued activations, but only from a strategically early time step of each sequential input example, decided by a threshold-crossing event. Learning then incorporates naturally also when to produce an output, without other changes to the model or the algorithm. Both the forward and the backward training pass can be significantly shortened by skipping the remaining input sequence after that first event. RC-training also significantly reduces time-to-insight during inference, with a minimal decrease in accuracy. The desired speed-accuracy trade-off is tunable by varying the threshold or a regularization parameter that rewards output entropy. We demonstrate these in two toy problems of sequence classification, and in a temporally-encoded MNIST dataset where our RC model achieves 99.19% accuracy after the first input time-step, outperforming the state of the art in temporal coding with SNNs, as well as in spoken-word classification of Google Speech Commands, outperforming non-RC-trained early inference with LSTMs. | SPIKE-INSPIRED RANK CODING FOR FAST AND ACCU- RATE RECURRENT NEURAL NETWORKS |
d257280243 | Recently, there has been a growing surge of interest in enabling machine learning systems to generalize well to Out-of-Distribution (OOD) data. Most efforts are devoted to advancing optimization objectives that regularize models to capture the underlying invariance; however, there often are compromises in the optimization process of these OOD objectives: i) Many OOD objectives have to be relaxed as penalty terms of Empirical Risk Minimization (ERM) for the ease of optimization, while the relaxed forms can weaken the robustness of the original objective; ii) The penalty terms also require careful tuning of the penalty weights due to the intrinsic conflicts between ERM and OOD objectives. Consequently, these compromises could easily lead to suboptimal performance of either the ERM or OOD objective. To address these issues, we introduce a multi-objective optimization (MOO) perspective to understand the OOD optimization process, and propose a new optimization scheme called PAreto Invariant Risk Minimization (PAIR). PAIR improves the robustness of OOD objectives by cooperatively optimizing with other OOD objectives, thereby bridging the gaps caused by the relaxations. Then PAIR approaches a Pareto optimal solution that trades off the ERM and OOD objectives properly. Extensive experiments on challenging benchmarks, WILDS, show that PAIR alleviates the compromises and yields top OOD performances. 1 * Work done during an internship at Tencent AI Lab. 1 Code is available at https://github.com/LFhase/PAIR.Published as a conference paper at ICLR 2023To address these issues, we propose a new optimization scheme for OOD generalization, called PAreto Invariant Risk Minimization (PAIR), which includes a new optimizer (PAIR-o) and a new model selection criteria (PAIR-s). Owing to the MOO formulation, PAIR-o allows for cooperative optimization with other OOD objectives to improve the robustness of practical OOD objectives. Despite the huge gaps between IRMv1 and IRM, we show that incorporating VREx (Krueger et al., 2021) into IRMv1 provably recovers the causal invariance (Arjovsky et al., 2019) for some group of problem instances (Sec. 3.2). When given robust OOD objectives, PAIR-o finds a descent path with adaptive penalty weights, which leads to a Pareto optimal solution that trades off ERM and OOD performance properly (Sec. 4). In addition, the MOO analysis also motivates PAIR-s, which facilitates the OOD model selection by considering the trade-offs between ERM and OOD objectives.We conducted extensive experiments on challenging OOD benchmarks. Empirical results show that PAIR-o successfully alleviates the objective conflicts and empowers IRMv1 to achieve high perfor- | PARETO INVARIANT RISK MINIMIZATION: TOWARDS MITIGATING THE OPTIMIZATION DILEMMA IN OUT- OF-DISTRIBUTION GENERALIZATION |
d256827141 | In recent years, deep network pruning has attracted significant attention in order to enable the rapid deployment of AI into small devices with computation and memory constraints. Pruning is often achieved by dropping redundant weights, neurons, or layers of a deep network while attempting to retain a comparable test performance. Many deep pruning algorithms have been proposed with impressive empirical success. However, existing approaches lack a quantifiable measure to estimate the compressibility of a sub-network during each pruning iteration and thus may underprune or over-prune the model. In this work, we propose PQ Index (PQI) to measure the potential compressibility of deep neural networks and use this to develop a Sparsity-informed Adaptive Pruning (SAP) algorithm. Our extensive experiments corroborate the hypothesis that for a generic pruning procedure, PQI decreases first when a large model is being effectively regularized and then increases when its compressibility reaches a limit that appears to correspond to the beginning of underfitting. Subsequently, PQI decreases again when the model collapse and significant deterioration in the performance of the model start to occur. Additionally, our experiments demonstrate that the proposed adaptive pruning algorithm with proper choice of hyper-parameters is superior to the iterative pruning algorithms such as the lottery ticket-based pruning methods, in terms of both compression efficiency and robustness. Our code is available here. | Published as a conference paper at ICLR 2023 PRUNING DEEP NEURAL NETWORKS FROM A SPARSITY PERSPECTIVE |
d198917339 | The problem of verifying whether a textual hypothesis holds the truth based on the given evidence, also known as fact verification, plays an important role in the study of natural language understanding and semantic representation. However, existing studies are mainly restricted to dealing with unstructured evidence (e.g., natural language sentences and documents, news, etc), while verification under structured evidence, such as tables, graphs, and databases, remains unexplored. This paper specifically aims to study the fact verification given semistructured data as evidence. To this end, we construct a large-scale dataset called TABFACT with 16k Wikipedia tables as evidence for 118k human-annotated natural language statements, which are labeled as either ENTAILED or REFUTED. | TABFACT: A LARGE-SCALE DATASET FOR TABLE- BASED FACT VERIFICATION |
d256194556 | Adversarial training suffers from the issue of robust overfitting, which seriously impairs its generalization performance. Data augmentation, which is effective at preventing overfitting in standard training, has been observed by many previous works to be ineffective in mitigating overfitting in adversarial training. This work proves that, contrary to previous findings, data augmentation alone can significantly boost accuracy and robustness in adversarial training. We find that the hardness and the diversity of data augmentation are important factors in combating robust overfitting. In general, diversity can improve both accuracy and robustness, while hardness can boost robustness at the cost of accuracy within a certain limit and degrade them both over that limit. To mitigate robust overfitting, we first propose a new crop transformation, Cropshift, which has improved diversity compared to the conventional one (Padcrop). We then propose a new data augmentation scheme, based on Cropshift, with much improved diversity and well-balanced hardness. Empirically, our augmentation method achieves the state-of-the-art accuracy and robustness for data augmentations in adversarial training. Furthermore, when combined with weight averaging it matches, or even exceeds, the performance of the best contemporary regularization methods for alleviating robust overfitting. Code is available at: https://github.com/TreeLLi/DA-Alone-Improves-AT. | Published as a conference paper at ICLR 2023 DATA AUGMENTATION ALONE CAN IMPROVE ADVER- SARIAL TRAINING |
d247011539 | Denoising diffusion probabilistic models have been recently proposed to generate high-quality samples by estimating the gradient of the data density. The framework defines the prior noise as a standard Gaussian distribution, whereas the corresponding data distribution may be more complicated than the standard Gaussian distribution, which potentially introduces inefficiency in denoising the prior noise into the data sample because of the discrepancy between the data and the prior. In this paper, we propose PriorGrad to improve the efficiency of the conditional diffusion model for speech synthesis (for example, a vocoder using a mel-spectrogram as the condition) by applying an adaptive prior derived from the data statistics based on the conditional information. We formulate the training and sampling procedures of PriorGrad and demonstrate the advantages of an adaptive prior through a theoretical analysis. Focusing on the speech synthesis domain, we consider the recently proposed diffusion-based speech generative models based on both the spectral and time domains and show that PriorGrad achieves faster convergence and inference with superior performance, leading to an improved perceptual quality and robustness to a smaller network capacity, and thereby demonstrating the efficiency of a data-dependent adaptive prior. * Work done during an internship at Microsoft Research Asia † Corresponding Authors Published as a conference paper at ICLR 2022 However, although the diffusion-based speech synthesis models have achieved high-quality speech audio generation, they exhibit potential inefficiency, which may necessitate advanced strategies. For example, the model suffers from a significantly slow convergence during training, and a prohibitively large training computation time is required to learn the approximate reverse diffusion process. We investigate the diffuion-based models and observe the discrepancy between the real data distribution and the choice of the prior. Existing diffusion-based models define a standard Gaussian as the prior distribution and design a non-parametric diffusion process that procedurally destroys the signal into the prior noise. The deep neural network is trained to approximate the reverse diffusion process by estimating the gradient of the data density. Although applying the standard Gaussian as the prior is simple without any assumptions on the target data, it also introduces inefficiency. For example, in time-domain waveform data, the signal has extremely high variability between different segments such as voiced and unvoiced parts. Jointly modeling the voiced and unvoiced segments with the same standard Gaussian prior may be difficult for the model to cover all modes of the data, leading to training inefficiencies and potentially spurious diffusion trajectories.Given the previous reasoning, we assessed the following question: For a conditional diffusion-based model, can we formulate a more informative prior without incorporating additional computational or parameter complexity? To investigate this, we propose a simple yet effective method, called PriorGrad, that uses adaptive noise by directly computing the mean and variance for the forward diffusion process prior, based on the conditional information. Specifically, using a conditional speech synthesis model, we propose structuring the prior distribution based on the conditional data, such as a mel-spectrogram for the vocoder Kong et al., 2021) and a phoneme for the acoustic model(Jeong et al., 2021). By computing the statistics from the conditional data at the frame level (vocoder) or phoneme-level (acoustic model) granularity and mapping them as the mean and variance of the Gaussian prior, we can structure the noise that is similar to the target data distribution at an instance level, easing the burden of learning the reverse diffusion process.We implemented PriorGrad based on the recently proposed diffusion-based speech generative models (Kong et al., 2021;Jeong et al., 2021), and conducted experiments on the LJSpeech (Ito & Johnson, 2017) dataset. The experimental results demonstrate the benefits of Prior-Grad, such as a significantly faster model convergence during training, improved perceptual quality, and an improved tolerance to a reduction in network capacity. Our contributions are as follows: | Published as a conference paper at ICLR 2022 PRIORGRAD: IMPROVING CONDITIONAL DENOISING DIFFUSION MODELS WITH DATA-DEPENDENT ADAP- TIVE PRIOR |
d252596186 | Vessel segmentation in medical images is one of the important tasks in the diagnosis of vascular diseases and therapy planning. Although learning-based segmentation approaches have been extensively studied, a large amount of groundtruth labels are required in supervised methods and confusing background structures make neural networks hard to segment vessels in an unsupervised manner. To address this, here we introduce a novel diffusion adversarial representation learning (DARL) model that leverages a denoising diffusion probabilistic model with adversarial learning, and apply it to vessel segmentation. In particular, for self-supervised vessel segmentation, DARL learns the background signal using a diffusion module, which lets a generation module effectively provide vessel representations. Also, by adversarial learning based on the proposed switchable spatially-adaptive denormalization, our model estimates synthetic fake vessel images as well as vessel segmentation masks, which further makes the model capture vessel-relevant semantic information. Once the proposed model is trained, the model generates segmentation masks in a single step and can be applied to general vascular structure segmentation of coronary angiography and retinal images. Experimental results on various datasets show that our method significantly outperforms existing unsupervised and self-supervised vessel segmentation methods. * co-first authors arXiv:2209.14566v2 [eess.IV] | DIFFUSION ADVERSARIAL REPRESENTATION LEARN- ING FOR SELF-SUPERVISED VESSEL SEGMENTATION |
d258841675 | Existing feature distillation methods commonly adopt the One-to-one Representation Matching between any pre-selected teacher-student layer pair. In this paper, we present N-to-One Representation Matching (NORM), a new two-stage knowledge distillation method, which relies on a simple Feature Transform (FT) module consisting of two linear layers. In view of preserving the intact information learnt by the teacher network, during training, our FT module is merely inserted after the last convolutional layer of the student network. The first linear layer projects the student representation to a feature space having N times feature channels than the teacher representation from the last convolutional layer, and the second linear layer contracts the expanded output back to the original feature space. By sequentially splitting the expanded student representation into N non-overlapping feature segments having the same number of feature channels as the teacher's, they can be readily forced to approximate the intact teacher representation simultaneously, formulating a novel many-to-one representation matching mechanism conditioned on a single teacher-student layer pair. After training, such an FT module will be naturally merged into the subsequent fully connected layer thanks to its linear property, introducing no extra parameters or architectural modifications to the student network at inference. Extensive experiments on different visual recognition benchmarks demonstrate the leading performance of our method. For instance, the ResNet18|MobileNet|ResNet50-1/4 model trained by NORM reaches 72.14%|74.26%|68.03% top-1 accuracy on the ImageNet dataset when using a pretrained ResNet34|ResNet50|ResNet50 model as the teacher, achieving an absolute improvement of 2.01%|4.63%|3.03% against the individually trained counterpart. Code is available at https://github.com/OSVAI/NORM. * XL, LL and CL contributed to the basic method implementations. XL conducted the main experiments. AY proposed the original idea, supervised the project and led the paper writing. † Corresponding author.Published as a conference paper at ICLR 2023We evaluate the performance of NORM on different visual recognition benchmarks. On the CIFAR-100 dataset, the student models trained by NORM show a mean accuracy improvement of 2.88% over 7 teacher-student pairs of the same type network architectures. Over 6 teacher-student pairs of different type network architectures, the mean accuracy improvement reaches 5.81%, and the maximal gain is 6.92%. Leading results are obtained on the large-scale ImageNet dataset. With NORM, the ResNet18|MobileNet|ResNet50-1/4 model reaches 72.14%|74.26%|68.03% top-1 accuracy when using a pre-trained ResNet34|ResNet50|ResNet50 model as the teacher, showing 2.01%|4.63%|3.03% absolute gain to the baseline model. Thanks to its simplicity and compatibility, we show that improved performance could be further attained by combining NORM with other popular distillation strategies like logits based supervision (Hinton et al., 2015) and contrastive learning (Tian et al., 2020).RELATED WORKTwo-stage KD methods. This category of KD methods first assumes that a pre-trained teacher network is available, and then uses its learnt representation as extra supervision to guide the training . Rethinking atrous convolution for semantic image segmentation. arXiv preprint arXiv:1706.05587, 2017. . Distpro: Searching a fast knowledge distillation process via meta optimization. arXiv preprint arXiv:2204.05547, 2022.Xiaohan Ding, Yuchen Guo, Guiguang Ding, and Jungong Han. Acnet: Strengthening the kernel skeletons for powerful cnn via asymmetric convolution blocks. In ICCV, 2019. | Published as a conference paper at ICLR 2023 NORM: KNOWLEDGE DISTILLATION VIA N-TO-ONE REPRESENTATION MATCHING |
d257426979 | Learning policies via preference-based reward learning is an increasingly popular method for customizing agent behavior, but has been shown anecdotally to be prone to spurious correlations and reward hacking behaviors. While much prior work focuses on causal confusion in reinforcement learning and behavioral cloning, we focus on a systematic study of causal confusion and reward misidentification when learning from preferences. In particular, we perform a series of sensitivity and ablation analyses on several benchmark domains where rewards learned from preferences achieve minimal test error but fail to generalize to outof-distribution states-resulting in poor policy performance when optimized. We find that the presence of non-causal distractor features, noise in the stated preferences, and partial state observability can all exacerbate reward misidentification. We also identify a set of methods with which to interpret misidentified learned rewards. In general, we observe that optimizing misidentified rewards drives the policy off the reward's training distribution, resulting in high predicted (learned) rewards but low true rewards. These findings illuminate the susceptibility of preference learning to reward misidentification and causal confusion-failure to consider even one of many factors can result in unexpected, undesirable behavior. | Published as a conference paper at ICLR 2023 CAUSAL CONFUSION AND REWARD MISIDENTIFICA- TION IN PREFERENCE-BASED REWARD LEARNING |
d231861390 | We study the challenging task of neural network quantization without end-toend retraining, called Post-training Quantization (PTQ). PTQ usually requires a small subset of training data but produces less powerful quantized models than Quantization-Aware Training (QAT). In this work, we propose a novel PTQ framework, dubbed BRECQ, which pushes the limits of bitwidth in PTQ down to INT2 for the first time. BRECQ leverages the basic building blocks in neural networks and reconstructs them one-by-one. In a comprehensive theoretical study of the second-order error, we show that BRECQ achieves a good balance between crosslayer dependency and generalization error. To further employ the power of quantization, the mixed precision technique is incorporated in our framework by approximating the inter-layer and intra-layer sensitivity. Extensive experiments on various handcrafted and searched neural architectures are conducted for both image classification and object detection tasks. And for the first time we prove that, without bells and whistles, PTQ can attain 4-bit ResNet and MobileNetV2 comparable with QAT and enjoy 240× faster production of quantized models. Codes Published as a conference paper at ICLR 2021 error based on the Gauss-Newton matrix. We show that the second-order error can be transformed into network final outputs but suffer from bad generalization. To achieve the best tradeoff, we adopt an intermediate choice, block reconstruction. In addition, our contributions are threefold: | Published as a conference paper at ICLR 2021 BRECQ: PUSHING THE LIMIT OF POST-TRAINING QUANTIZATION BY BLOCK RECONSTRUCTION |
d248476477 | We propose a novel 3d shape representation for 3d shape reconstruction from a single image. Rather than predicting a shape directly, we train a network to generate a training set which will be fed into another learning algorithm to define the shape. The nested optimization problem can be modeled by bi-level optimization. Specifically, the algorithms for bi-level optimization are also being used in meta learning approaches for few-shot learning. Our framework establishes a link between 3D shape analysis and few-shot learning. We combine training data generating networks with bi-level optimization algorithms to obtain a complete framework for which all components can be jointly trained. We improve upon recent work on standard benchmarks for 3d shape reconstruction. | Published as a conference paper at ICLR 2022 TRAINING DATA GENERATING NETWORKS: SHAPE RECONSTRUCTION VIA BI-LEVEL OPTIMIZATION |
d253098549 | The formalization of existing mathematical proofs is a notoriously difficult process. Despite decades of research on automation and proof assistants, writing formal proofs remains arduous and only accessible to a few experts. While previous studies to automate formalization focused on powerful search algorithms, no attempts were made to take advantage of available informal proofs. In this work, we introduce Draft, Sketch, and Prove (DSP), a method that maps informal proofs to formal proof sketches, and uses the sketches to guide an automated prover by directing its search to easier sub-problems. We investigate two relevant setups where informal proofs are either written by humans or generated by a language model. Our experiments and ablation studies show that large language models are able to produce wellstructured formal sketches that follow the same reasoning steps as the informal proofs. Guiding an automated prover with these sketches enhances its performance from 20.9% to 39.3% on a collection of mathematical competition problems.Figure 1: Draft, Sketch, and Prove. Starting with an informal statement, our framework yields a formal proof through a three-stage process: drafting informal proofs, mapping them into formal sketches, and proving the remaining conjectures. Concretely, an informal statement is a mathematical problem described in a mixture of natural and mathematical languages (e.g., formulae in L A T E X). Then, we use a large language model to autoformalize each informal proof into a formal sketch, which is a skeleton of the formal proof with open conjectures left unproven (indicated by the <proof> blocks). The formal sketch mirrors the structure of the informal proof. Finally, the open conjectures/gaps inside each formal sketch are proved by an off-the-shelf prover. † Equal contributions as leading authors. Correspondence to: qj213@cam.ac.uk. ‡ Equal contributions as senior authors. | Published as a conference paper at ICLR 2023 DRAFT, SKETCH, AND PROVE: GUIDING FORMAL THEOREM PROVERS WITH INFORMAL PROOFS |
d18323900 | Several interesting generative learning algorithms involve a complex probability distribution over many random variables, involving intractable normalization constants or latent variable marginalization. Some of them may not have even an analytic expression for the unnormalized probability function and no tractable approximation. This makes it difficult to estimate the quality of these models, once they have been trained, or to monitor their quality (e.g. for early stopping) while training. A previously proposed method is based on constructing a non-parametric density estimator of the model's probability function from samples generated by the model. We revisit this idea, propose a more efficient estimator, and prove that it provides a lower bound on the true test log-likelihood and an unbiased estimator as the number of generated samples goes to infinity, although one that incorporates the effect of poor mixing. We further propose a biased variant of the estimator that can be used reliably with a finite number of samples for the purpose of model comparison. | Bounding the Test Log-Likelihood of Generative Models |
d2536452 | We propose a method for learning latent representations of the factors of variation in data. By augmenting deep autoencoders with a supervised cost and an additional unsupervised cost, we create a semi-supervised model that can discover and explicitly represent factors of variation beyond those relevant for categorization. We use a novel unsupervised covariance penalty (XCov) to disentangle factors like handwriting style for digits and subject identity in faces. We demonstrate this on the MNIST handwritten digit database, the Toronto Faces Database (TFD) and the Multi-PIE dataset by generating manipulated instances of the data. Furthermore, we demonstrate these deep networks can extrapolate 'hidden' variation in the supervised signal using the Toronto Faces Database. | DISCOVERING HIDDEN FACTORS OF VARIATION IN DEEP NETWORKS |
d257482840 | We consider the problem of training a deep neural network on a given classification task, e.g., ImageNet-1K (IN1K), so that it excels at both the training task as well as at other (future) transfer tasks. These two seemingly contradictory properties impose a trade-off between improving the model's generalization and maintaining its performance on the original task. Models trained with self-supervised learning tend to generalize better than their supervised counterparts for transfer learning; yet, they still lag behind supervised models on IN1K. In this paper, we propose a supervised learning setup that leverages the best of both worlds. We extensively analyze supervised training using multi-scale crops for data augmentation and an expendable projector head, and reveal that the design of the projector allows us to control the trade-off between performance on the training task and transferability. We further replace the last layer of class weights with class prototypes computed on the fly using a memory bank and derive two models: t-ReX that achieves a new state of the art for transfer learning and outperforms top methods such as DINO and PAWS on IN1K, and t-ReX* that matches the highly optimized RSB-A1 model on IN1K while performing better on transfer tasks. Code and pretrained models: https://europe.naverlabs.com/t-rex | Published as a conference paper at ICLR 2023 NO REASON FOR NO SUPERVISION: IMPROVED GENERALIZATION IN SUPERVISED MODELS |
d237347130 | Since the introduction of the transformer model byVaswani et al. (2017), a fundamental question has yet to be answered: how does a model achieve extrapolation at inference time for sequences that are longer than it saw during training? We first show that extrapolation can be enabled by simply changing the position representation method, though we find that current methods do not allow for efficient extrapolation. We therefore introduce a simpler and more efficient position method, Attention with Linear Biases (ALiBi). ALiBi does not add positional embeddings to word embeddings; instead, it biases query-key attention scores with a penalty that is proportional to their distance. We show that this method trains a 1.3 billion parameter model on input sequences of length 1024 that extrapolates to input sequences of length 2048, achieving the same perplexity as a sinusoidal position embedding model trained on inputs of length 2048 but training 11% faster and using 11% less memory. ALiBi's inductive bias towards recency also leads it to outperform multiple strong position methods on the WikiText-103 benchmark. 1 1 Code & models: https://github.com/ofirpress/attention_with_linear_biases 2Figure 7in the appendix plots training speed, in words per second, against L. | Published as a conference paper at ICLR 2022 TRAIN SHORT, TEST LONG: ATTENTION WITH LINEAR BIASES ENABLES INPUT LENGTH EXTRAPOLATION |
d53113692 | Neural language models have been widely used in various NLP tasks, including machine translation, next word prediction and conversational agents. However, it is challenging to deploy these models on mobile devices due to their slow prediction speed, where the bottleneck is to compute top candidates in the softmax layer. In this paper, we introduce a novel softmax layer approximation algorithm by exploiting the clustering structure of context vectors. Our algorithm uses a light-weight screening model to predict a much smaller set of candidate words based on the given context, and then conducts an exact softmax only within that subset. Training such a procedure end-to-end is challenging as traditional clustering methods are discrete and non-differentiable, and thus unable to be used with back-propagation in the training process. Using the Gumbel softmax, we are able to train the screening model end-to-end on the training set to exploit data distribution. The algorithm achieves an order of magnitude faster inference than the original softmax layer for predicting top-k words in various tasks such as beam search in machine translation or next words prediction. For example, for machine translation task on German to English dataset with around 25K vocabulary, we can achieve 20.4 times speed up with 98.9% precision@1 and 99.3% preci-sion@5 with the original softmax layer prediction, while state-of-the-art (Zhang et al., 2018) only achieves 6.7x speedup with 98.7% precision@1 and 98.1% pre-cision@5 for the same task. | LEARNING TO SCREEN FOR FAST SOFTMAX INFER- ENCE ON LARGE VOCABULARY NEURAL NETWORKS |
d248006494 | We introduce compositional soft prompting (CSP), a parameter-efficient learning technique to improve the zero-shot compositionality of large-scale pretrained vision-language models (VLMs) like CLIP. We develop CSP for compositional zero-shot learning, the task of predicting unseen attribute-object compositions (e.g., old cat and young tiger). VLMs have a flexible text encoder that can represent arbitrary classes as natural language prompts but they often underperform taskspecific architectures on the compositional zero-shot benchmark datasets. CSP treats the attributes and objects that define classes as learnable tokens of vocabulary. During training, the vocabulary is tuned to recognize classes that compose tokens in multiple ways (e.g., old cat and white cat). At test time, we recompose the learned attribute-object vocabulary in new combinations to recognize novel classes. We show that CSP outperforms the CLIP on benchmark datasets by an average of 10.9 percentage points on AUC. CSP also outperforms CoOp, a soft prompting method that fine-tunes the prefix context tokens, by an average of 5.8 percentage points on AUC. We perform additional experiments to show that CSP improves generalization to higher-order attribute-attribute-object compositions (e.g., old white cat) and combinations of pretrained attributes and fine-tuned objects. The code is available at https | Published as a conference paper at ICLR 2023 LEARNING TO COMPOSE SOFT PROMPTS FOR COMPOSITIONAL ZERO-SHOT LEARNING |
d252668546 | Dedicated neural network (NN) architectures have been designed to handle specific data types (such as CNN for images or RNN for text), which ranks them among state-of-the-art methods for dealing with these data. Unfortunately, no architecture has been found for dealing with tabular data yet, for which tree ensemble methods (tree boosting, random forests) usually show the best predictive performances. In this work, we propose a new sparse initialization technique for (potentially deep) multilayer perceptrons (MLP): we first train a tree-based procedure to detect feature interactions and use the resulting information to initialize the network, which is subsequently trained via standard stochastic gradient strategies. Numerical experiments on several tabular data sets show that this new, simple and easy-to-use method is a solid concurrent, both in terms of generalization capacity and computation time, to default MLP initialization and even to existing complex deep learning solutions. In fact, this wise MLP initialization raises the resulting NN methods to the level of a valid competitor to gradient boosting when dealing with tabular data. Besides, such initializations are able to preserve the sparsity of weights introduced in the first layers of the network through training. This fact suggests that this new initializer operates an implicit regularization during the NN training, and emphasizes that the first layers act as a sparse feature extractor (as for convolutional layers in CNN). poorly on tabular inputs, for which tree ensemble methods remain the gold standards (Grinsztajn et al., 2022).Tree ensemble methods Tree-based methods are widely used in the ML community, especially for processing tabular data. Two main approaches exist depending on whether the tree building process is parallel (e.g. Random Forest, RF, see Breiman, 2001a) or sequential (e.g. Gradient Boosting Decision Trees, GBDT, see Friedman, 2001). In these tree ensemble procedures, the final prediction relies on averaging predictions of randomized decision trees, coding for particular partitions of the input space. The two most successful and most widely used implementations of these methods are XGBoost and LightGBM (see Chen and Guestrin, 2016; Ke et al., 2017) which both rely on the sequential GBDT approach.Neural networks Neural Networks (NN) are efficient methods to unveil the patterns of spatial or temporal data, such as images (Krizhevsky et al., 2012) or texts (Liu et al., 2016). Their performance results notably from the fact that several architectures directly encode relevant structures in the input: convolutional neural networks (CNN, LeCun et al., 1995) use convolutions to detect spatiallyinvariant patterns in images, and recurrent neural networks (RNN, Rumelhart et al., 1985) use a hidden temporal state to leverage the natural order of a text. However, a dedicated natural architecture has yet to be introduced to deal with tabular data. Indeed, designing such an architecture would require to detect and leverage the structure of the relations between variables, which is much easier for images or text (spatial or temporal correlation) than for tabular data (unconstrained covariance structure).NN initialization and trainingIn the absence of a suitable architecture for handling tabular data, the Multi-Layer Perceptron (MLP) architecture (Rumelhart et al., 1986) remains the obvious choice due to its generalist nature. Apart from the large number of parameters, one difficulty of MLP training arises from the non-convexity of the loss function (see, e.g., Sun, 2020). In such situations, the initialization of the network parameters (weights and biases) are of the utmost importance, since it can influence both the optimization stability and the quality of the minimum found. Typically, such initializations are drawn according to independent uniform distributions with a variance decreasing w.r.t. the size of the layer (He et al., 2015). Therefore, one may wonder how to capitalize on methods that are inherently capable of recognizing patterns in tabular data (e.g., tree-based methods) to propose a new NN architecture suitable for tabular data and an initialization procedure that leads to faster convergence and better generalization performance.2 | S T - I N N |
d7301499 | Transferring knowledge from prior source tasks in solving a new target task can be useful in several learning applications. The application of transfer poses two serious challenges which have not been adequately addressed. First, the agent should be able to avoid negative transfer, which happens when the transfer hampers or slows down the learning instead of helping it. Second, the agent should be able to selectively transfer, which is the ability to select and transfer from different and multiple source tasks for different parts of the state space of the target task. We propose A2T (Attend, Adapt and Transfer), an attentive deep architecture which adapts and transfers from these source tasks. Our model is generic enough to effect transfer of either policies or value functions. Empirical evaluations on different learning algorithms show that A2T is an effective architecture for transfer by being able to avoid negative transfer while transferring selectively from multiple source tasks in the same domain. | Published as a conference paper at ICLR 2017 ATTEND, ADAPT AND TRANSFER: ATTENTIVE DEEP ARCHITECTURE FOR ADAPTIVE TRANSFER FROM MULTIPLE SOURCES IN THE SAME DOMAIN |
d231627780 | Given (small amounts of) time-series' data from a high-dimensional, fine-grained, multiscale dynamical system, we propose a generative framework for learning an effective, lower-dimensional, coarse-grained dynamical model that is predictive of the fine-grained system's long-term evolution but also of its behavior under different initial conditions. We target fine-grained models as they arise in physical applications (e.g. molecular dynamics, agent-based models), the dynamics of which are strongly non-stationary but their transition to equilibrium is governed by unknown slow processes which are largely inaccessible by brute-force simulations. Approaches based on domain knowledge heavily rely on physical insight in identifying temporally slow features and fail to enforce the long-term stability of the learned dynamics. On the other hand, purely statistical frameworks lack interpretability and rely on large amounts of expensive simulation data (long and multiple trajectories) as they cannot infuse domain knowledge. The generative framework proposed achieves the aforementioned desiderata by employing a flexible prior on the complex plane for the latent, slow processes, and an intermediate layer of physics-motivated latent variables that reduces reliance on data and imbues inductive bias. In contrast to existing schemes, it does not require the a priori definition of projection operators or encoders and addresses simultaneously the tasks of dimensionality reduction and model estimation. We demonstrate its efficacy and accuracy in multiscale physical systems of particle dynamics where probabilistic, long-term predictions of phenomena not contained in the training data are produced. | PHYSICS-AWARE, PROBABILISTIC MODEL ORDER RE- DUCTION WITH GUARANTEED STABILITY |
d232257725 | The ability to learn continually without forgetting the past tasks is a desired attribute for artificial learning systems. Existing approaches to enable such learning in artificial neural networks usually rely on network growth, importance based weight update or replay of old data from the memory. In contrast, we propose a novel approach where a neural network learns new tasks by taking gradient steps in the orthogonal direction to the gradient subspaces deemed important for the past tasks. We find the bases of these subspaces by analyzing network representations (activations) after learning each task with Singular Value Decomposition (SVD) in a single shot manner and store them in the memory as Gradient Projection Memory (GPM). With qualitative and quantitative analyses, we show that such orthogonal gradient descent induces minimum to no interference with the past tasks, thereby mitigates forgetting. We evaluate our algorithm on diverse image classification datasets with short and long sequences of tasks and report better or on-par performance compared to the state-of-the-art approaches 1 . 1 Our code is available at https | Published as a conference paper at ICLR 2021 GRADIENT PROJECTION MEMORY FOR CONTINUAL LEARNING |
d235652429 | Data augmentation is a key element of deep learning pipelines, as it informs the network during training about transformations of the input data that keep the label unchanged. Manually finding adequate augmentation methods and parameters for a given pipeline is however rapidly cumbersome. In particular, while intuition can guide this decision for images, the design and choice of augmentation policies remains unclear for more complex types of data, such as neuroscience signals. Besides, class-dependent augmentation strategies have been surprisingly unexplored in the literature, although it is quite intuitive: changing the color of a car image does not change the object class to be predicted, but doing the same to the picture of an orange does. This paper investigates gradient-based automatic data augmentation algorithms amenable to class-wise policies with exponentially larger search spaces. Motivated by supervised learning applications using EEG signals for which good augmentation policies are mostly unknown, we propose a new differentiable relaxation of the problem. In the class-agnostic setting, results show that our new relaxation leads to optimal performance with faster training than competing gradient-based methods, while also outperforming gradient-free methods in the class-wise setting. This work proposes also novel differentiable augmentation operations relevant for sleep stage classification. | CADDA: Class-wise Automatic Differen- tiable Data Augmentation for EEG Signals |
d254043898 | Neural networks with sinusoidal activations have been proposed as an alternative to networks with traditional activation functions. Despite their promise, particularly for learning implicit models, their training behavior is not yet fully understood, leading to a number of empirical design choices that are not well justified. In this work, we first propose a simplified version of such sinusoidal neural networks, which allows both for easier practical implementation and simpler theoretical analysis. We then analyze the behavior of these networks from the neural tangent kernel perspective and demonstrate that their kernel approximates a low-pass filter with an adjustable bandwidth. Finally, we utilize these insights to inform the sinusoidal network initialization, optimizing their performance for each of a series of tasks, including learning implicit models and solving differential equations. | Published as a conference paper at ICLR 2023 SIMPLE INITIALIZATION AND PARAMETRIZATION OF SINUSOIDAL NETWORKS VIA THEIR KERNEL BAND- WIDTH |
d6395440 | There are two main approaches to the distributed representation of words: lowdimensional deep learning embeddings and high-dimensional distributional models, in which each dimension corresponds to a context word. In this paper, we combine these two approaches by learning embeddings based on distributionalmodel vectors -as opposed to one-hot vectors as is standardly done in deep learning. We show that the combined approach has better performance on a word relatedness judgment task. | Distributional Models and Deep Learning Embeddings: Combining the Best of Both Worlds |
d211132734 | The universal approximation theorem, in one of its most general versions, says that if we consider only continuous activation functions σ, then a standard feedforward neural network with one hidden layer is able to approximate any continuous multivariate function f to any given approximation threshold ε, if and only if σ is non-polynomial. In this paper, we give a direct algebraic proof of the theorem. Furthermore we shall explicitly quantify the number of hidden units required for approximation. Specifically, if X ⊆ R n is compact, then a neural network with n input units, m output units, and a single hidden layer with n+d d hidden units (independent of m and ε), can uniformly approximate any polynomial function f : X → R m whose total degree is at most d for each of its m coordinate functions. In the general case that f is any continuous function, we show there exists some N ∈ O(ε −n ) (independent of m), such that N hidden units would suffice to approximate f . We also show that this uniform approximation property (UAP) still holds even under seemingly strong conditions imposed on the weights. We highlight several consequences: (i) For any δ > 0, the UAP still holds if we restrict all non-bias weights w in the last layer to satisfy |w| < δ. (ii) There exists some λ > 0 (depending only on f and σ), such that the UAP still holds if we restrict all non-bias weights w in the first layer to satisfy |w| > λ. (iii) If the non-bias weights in the first layer are fixed and randomly chosen from a suitable range, then the UAP holds with probability 1. | A CLOSER LOOK AT THE APPROXIMATION CAPABILI- TIES OF NEURAL NETWORKS |
d6679306 | Multiple instance learning (MIL) can reduce the need for costly annotation in tasks such as semantic segmentation by weakening the required degree of supervision. We propose a novel MIL formulation of multi-class semantic segmentation learning by a fully convolutional network. In this setting, we seek to learn a semantic segmentation model from just weak image-level labels. The model is trained endto-end to jointly optimize the representation while disambiguating the pixel-image label assignment. Fully convolutional training accepts inputs of any size, does not need object proposal pre-processing, and offers a pixelwise loss map for selecting latent instances. Our multi-class MIL loss exploits the further supervision given by images with multiple labels. We evaluate this approach through preliminary experiments on the PASCAL VOC segmentation data. | Under review as a workshop contribution at ICLR 2015 FULLY CONVOLUTIONAL MULTI-CLASS MULTIPLE INSTANCE LEARNING |
d222177084 | We propose Deep Autoencoding Predictive Components (DAPC) -a selfsupervised representation learning method for sequence data, based on the intuition that useful representations of sequence data should exhibit a simple structure in the latent space. We encourage this latent structure by maximizing an estimate of predictive information of latent feature sequences, which is the mutual information between past and future windows at each time step. In contrast to the mutual information lower bound commonly used by contrastive learning, the estimate of predictive information we adopt is exact under a Gaussian assumption. Additionally, it can be computed without negative sampling. To reduce the degeneracy of the latent space extracted by powerful encoders and keep useful information from the inputs, we regularize predictive information learning with a challenging masked reconstruction loss. We demonstrate that our method recovers the latent space of noisy dynamical systems, extracts predictive features for forecasting tasks, and improves automatic speech recognition when used to pretrain the encoder on large amounts of unlabeled data. * Work done during internship at Salesforce Research. | REPRESENTATION LEARNING FOR SEQUENCE DATA WITH DEEP AUTOENCODING PREDICTIVE COMPO- NENTS |
d249209758 | We introduce CriticSMC, a new algorithm for planning as inference built from a composition of sequential Monte Carlo with learned Soft-Q function heuristic factors. These heuristic factors, obtained from parametric approximations of the marginal likelihood ahead, more effectively guide SMC towards the desired target distribution, which is particularly helpful for planning in environments with hard constraints placed sparsely in time. Compared with previous work, we modify the placement of such heuristic factors, which allows us to cheaply propose and evaluate large numbers of putative action particles, greatly increasing inference and planning efficiency. CriticSMC is compatible with informative priors, whose density function need not be known, and can be used as a model-free control algorithm. Our experiments on collision avoidance in a high-dimensional simulated driving task show that CriticSMC significantly reduces collision rates at a low computational cost while maintaining realism and diversity of driving behaviors across vehicles and environment scenarios. arXiv:2205.15460v2 [stat.ML] 21 Jan 2023 Published as a conference paper at ICLR 2023 (a) SMC (10 particles) (b) SMC (10k particles) (c) CriticSMC (10 particles) | Published as a conference paper at ICLR 2023 CRITIC SEQUENTIAL MONTE CARLO |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.