_id stringlengths 4 10 | text stringlengths 0 18.4k | title stringlengths 0 8.56k |
|---|---|---|
d22332081 | Learning a better representation with neural networks is a challenging problem, which was tackled extensively from different prospectives in the past few years. In this work, we focus on learning a representation that could be used for a clustering task and introduce two novel loss components that substantially improve the quality of produced clusters, are simple to apply to an arbitrary model and cost function, and do not require a complicated training procedure. We evaluate them on two most common types of models, Recurrent Neural Networks and Convolutional Neural Networks, showing that the approach we propose consistently improves the quality of KMeans clustering in terms of Adjusted Mutual Information score and outperforms previously proposed methods. | Forced to Learn: Discovering Disentangled Representations Without Exhaustive Labels |
d18946424 | We introduce a new method for training deep Boltzmann machines jointly. Prior methods of training DBMs require an initial learning pass that trains the model greedily, one layer at a time, or do not perform well on classification tasks. In our approach, we train all layers of the DBM simultaneously, using a novel training procedure called multi-prediction training. The resulting model can either be interpreted as a single generative model trained to maximize a variational approximation to the generalized pseudolikelihood, or as a family of recurrent networks that share parameters and may be approximately averaged together using a novel technique we call the multi-inference trick. We show that our approach performs competitively for classification and outperforms previous methods in terms of accuracy of approximate inference and classification with missing inputs. | Joint Training of Deep Boltzmann Machines for Classification |
d236924584 | Neural networks and other machine learning models compute continuous representations, while humans communicate mostly through discrete symbols. Reconciling these two forms of communication is desirable for generating human-readable interpretations or learning discrete latent variable models, while maintaining endto-end differentiability. Some existing approaches (such as the Gumbel-Softmax transformation) build continuous relaxations that are discrete approximations in the zero-temperature limit, while others (such as sparsemax transformations and the Hard Concrete distribution) produce discrete/continuous hybrids. In this paper, we build rigorous theoretical foundations for these hybrids, which we call "mixed random variables." Our starting point is a new "direct sum" base measure defined on the face lattice of the probability simplex. From this measure, we introduce new entropy and Kullback-Leibler divergence functions that subsume the discrete and differential cases and have interpretations in terms of code optimality. Our framework suggests two strategies for representing and sampling mixed random variables, an extrinsic ("sample-and-project") and an intrinsic one (based on face stratification). We experiment with both approaches on an emergent communication benchmark and on modeling MNIST and Fashion-MNIST data with variational auto-encoders with mixed latent variables. Our code is publicly available. | Published as a conference paper at ICLR 2022 SPARSE COMMUNICATION VIA MIXED DISTRIBUTIONS |
d257102348 | Abstraction is a desirable capability for deep learning models, which means to induce abstract concepts from concrete instances and flexibly apply them beyond the learning context. At the same time, there is a lack of clear understanding about both the presence and further characteristics of this capability in deep learning models. In this paper, we introduce a systematic probing framework to explore the abstraction capability of deep learning models from a transferability perspective. A set of controlled experiments are conducted based on this framework, providing strong evidence that two probed pre-trained language models (PLMs), T5 and GPT2, have the abstraction capability. We also conduct in-depth analysis, thus shedding further light: (1) the whole training phase exhibits a "memorize-thenabstract" two-stage process; (2) the learned abstract concepts are gathered in a few middle-layer attention heads, rather than evenly distributed throughout the model;(3) the probed abstraction capabilities exhibit robustness against concept mutations, and are more robust to low-level/source-side mutations than high-level/target-side ones; (4) generic pre-training is critical to the emergence of abstraction capability, and PLMs exhibit better abstraction with larger model sizes and data scales. * Work done during an internship at Microsoft Research.Abstract ConceptsInput Output… …Input Output… …Surface PatternsInput Output… …Abstract ConceptsFigure 1: Motivating example: the abstract concepts learned in task A can be effectively reused in task B, but surface patterns are useless. Unused patterns or concepts are whitened after the update.generally reused. We consider designing multiple tasks with shared abstract concepts and totally different surface patterns, then tracing whether the learning on one task can boost the performance on another.Figure 1demonstrates a motivating example.Published as a conference paper at ICLR 2023 Aiming Task Train Set ഥ Probing Task Test Set ഥ Transfer Set Abstract Concepts Same Task-Specific Characteristics Different Contrast Task Contrast Set ഥ Abstract Concepts Broken Main Exp ⟹ Pretrain Finetune Test ഥ Control Exp ⇑ ഥ | Published as a conference paper at ICLR 2023 DOES DEEP LEARNING LEARN TO ABSTRACT? A SYSTEMATIC PROBING FRAMEWORK |
d246431036 | Intrinsic interpretability of graph neural networks (GNNs) is to find a small subset of the input graph's features -rationale -which guides the model prediction. Unfortunately, the leading rationalization models often rely on data biases, especially shortcut features, to compose rationales and make predictions without probing the critical and causal patterns. Moreover, such data biases easily change outside the training distribution. As a result, these models suffer from a huge drop in interpretability and predictive performance on out-of-distribution data. In this work, we propose a new strategy of discovering invariant rationale (DIR) to construct intrinsically interpretable GNNs. It conducts interventions on the training distribution to create multiple interventional distributions. Then it approaches the causal rationales that are invariant across different distributions while filtering out the spurious patterns that are unstable. Experiments on both synthetic and realworld datasets validate the superiority of our DIR in terms of interpretability and generalization ability on graph classification over the leading baselines. Code and datasets are available at https://github.com/Wuyxin/DIR-GNN. : Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805, 2018. -lsc: A large-scale challenge for machine learning on graphs. arXiv preprint arXiv:2103.09430, 2021. Pritish Kamath, Akilesh Tangella, Danica J. Sutherland, and Nathan Srebro. Does invariant risk minimization capture invariance? In Arindam Banerjee and Kenji Fukumizu (eds.), AISTATS, 2021. Diederik P. Kingma and Jimmy Ba. Adam: A method for stochastic optimization. deep models for semantic compositionality over a sentiment treebank. In EMNLP, pp. 1631-1642, 2013. Damien Teney, Ehsan Abbasnejad, and Anton van den Hengel. Unshuffling data for improved generalization. arXiv, 2002.11894, 2020. Jin Tian, Changsung Kang, and Judea Pearl. A characterization of interventional distributions in semi-markovian causal models. In AAAI, pp. 1239-1244, 2006. G.E. van der Maaten, L.J.P.; Hinton. Visualizing high-dimensional data using t-sne. Journal of Machine Learning Research 9:2579-2605, 2008. Tyler J VanderWeele. A three-way decomposition of a total effect into direct, indirect, and interactive effects. Epidemiology (Cambridge, Mass.), 24(2):224, 2013. | Published as a conference paper at ICLR 2022 DISCOVERING INVARIANT RATIONALES FOR GRAPH NEURAL NETWORKS |
d220871147 | We develop an approach to growing deep network architectures over the course of training, driven by a principled combination of accuracy and sparsity objectives. Unlike existing pruning or architecture search techniques that operate on full-sized models or supernet architectures, our method can start from a small, simple seed architecture and dynamically grow and prune both layers and filters. By combining a continuous relaxation of discrete network structure optimization with a scheme for sampling sparse subnetworks, we produce compact, pruned networks, while also drastically reducing the computational expense of training. For example, we achieve 49.7% inference FLOPs and 47.4% training FLOPs savings compared to a baseline ResNet-50 on ImageNet, while maintaining 75.2% top-1 accuracy -all without any dedicated fine-tuning stage. Experiments across CIFAR, ImageNet, PASCAL VOC, and Penn Treebank, with convolutional networks for image classification and semantic segmentation, and recurrent networks for language modeling, demonstrate that we both train faster and produce more efficient networks than competing architecture pruning or search methods. | Published as a conference paper at ICLR 2021 GROWING EFFICIENT DEEP NETWORKS BY STRUCTURED CONTINUOUS SPARSIFICATION |
d213085920 | Many applications of machine learning require a model to make accurate predictions on test examples that are distributionally different from training ones, while task-specific labels are scarce during training. An effective approach to this challenge is to pre-train a model on related tasks where data is abundant, and then fine-tune it on a downstream task of interest. While pre-training has been effective in many language and vision domains, it remains an open question how to effectively use pre-training on graph datasets. In this paper, we develop a new strategy and self-supervised methods for pre-training Graph Neural Networks (GNNs). The key to the success of our strategy is to pre-train an expressive GNN at the level of individual nodes as well as entire graphs so that the GNN can learn useful local and global representations simultaneously. We systematically study pre-training on multiple graph classification datasets. We find that naïve strategies, which pre-train GNNs at the level of either entire graphs or individual nodes, give limited improvement and can even lead to negative transfer on many downstream tasks. In contrast, our strategy avoids negative transfer and improves generalization significantly across downstream tasks, leading up to 9.4% absolute improvements in ROC-AUC over non-pre-trained models and achieving state-of-the-art performance for molecular property prediction and protein function prediction. * Equal contribution. Project website, data and code: Attribute Masking Supervised Attribute Prediction Structural Similarity Prediction Structure prediction Context Prediction (b) Categorization of our pre-training methods Graph space Node space Graph embeddings Node embeddings Linear classifier Figure 1: (a.i) When only node-level pre-training is used, nodes of different shapes (semantically different nodes) can be well separated, however, node embeddings are not composable, and thus resulting graph embeddings (denoted by their classes, + and −) that are created by pooling node-level embeddings are not separable. (a.ii) With graph-level pre-training only, graph embeddings are well separated, however the embeddings of individual nodes do not necessarily capture their domainspecific semantics. (a.iii) High-quality node embeddings are such that nodes of different types are well separated, while at the same time, the embedding space is also composable. This allows for accurate and robust representations of entire graphs and enables robust transfer of pre-trained models to a variety of downstream tasks. (b) Categorization of pre-training methods for GNNs. Crucially, our methods, i.e., Context Prediction, Attribute Masking, and graph-level supervised pre-training (Supervised Attribute Prediction) enable both node-level and graph-level pre-training.matter of increasing the number of labeled pre-training datasets that are from the same domain as the downstream task. Instead, it requires substantial domain expertise to carefully select examples and target labels that are correlated with the downstream task of interest. Otherwise, the transfer of knowledge from related pre-training tasks to a new downstream task can harm generalization, which is known as negative transfer(Rosenstein et al., 2005)and significantly limits the applicability and reliability of pre-trained models.Present work. Here, we focus on pre-training as an approach to transfer learning in Graph Neural Networks (GNNs)(Kipf & Welling, 2017;Hamilton et al., 2017a;Ying et al., 2018b;Xu et al., 2019;2018)for graph-level property prediction. Our work presents two key contributions.(1) We conduct the first systematic large-scale investigation of strategies for pre-training GNNs. For that, we build two large new pre-training datasets, which we share with the community: a chemistry dataset with 2 million graphs and a biology dataset with 395K graphs. We also show that large domain-specific datasets are crucial to investigate pre-training and that existing downstream benchmark datasets are too small to evaluate models in a statistically reliable way.(2) We develop an effective pretraining strategy for GNNs and demonstrate its effectiveness and its ability for out-of-distribution generalization on hard transfer-learning problems. | Published as a conference paper at ICLR 2020 STRATEGIES FOR PRE-TRAINING GRAPH NEURAL NETWORKS |
d252408513 | Knowledge-intensive tasks, such as open-domain question answering (QA), require access to a large amount of world or domain knowledge. A common approach for knowledge-intensive tasks is to employ a retrieve-then-read pipeline that first retrieves a handful of relevant contextual documents from an external corpus such as Wikipedia and then predicts an answer conditioned on the retrieved documents. In this paper, we present a novel perspective for solving knowledge-intensive tasks by replacing document retrievers with large language model generators. We call our method generate-then-read (GENREAD), which first prompts a large language model to generate contextual documents based on a given question, and then reads the generated documents to produce the final answer. Furthermore, we propose a novel clustering-based prompting method that selects distinct prompts, in order to generate diverse documents that cover different perspectives, leading to better recall over acceptable answers. We conduct extensive experiments on three different knowledge-intensive tasks, including open-domain QA, fact checking, and dialogue system. Notably, GENREAD achieves 71.6 and 54.4 exact match scores on TriviaQA and WebQ, significantly outperforming the state-of-the-art retrieve-thenread pipeline DPR-FiD by +4.0 and +3.9, without retrieving any documents from any external knowledge source. Lastly, we demonstrate the model performance can be further improved by combining retrieval and generation. Our code and generated documents can be found at https://github.com/wyu97/GenRead. § Unless otherwise specified, we use the text-davinci-002 version of InstructGPT in our experiments. * Work done during internship at Microsoft Cognitive Service Research group. | Published as a conference paper at ICLR 2023 GENERATE RATHER THAN RETRIEVE: LARGE LANGU- AGE MODELS ARE STRONG CONTEXT GENERATORS |
d257632075 | This paper addresses learning end-to-end models for time series data that include a temporal alignment step via dynamic time warping (DTW). Existing approaches to differentiable DTW either differentiate through a fixed warping path or apply a differentiable relaxation to the min operator found in the recursive steps used to solve the DTW problem. We instead propose a DTW layer based around bilevel optimisation and deep declarative networks, which we name DecDTW. By formulating DTW as a continuous, inequality constrained optimisation problem, we can compute gradients for the solution of the optimal alignment (with respect to the underlying time series) using implicit differentiation. An interesting byproduct of this formulation is that DecDTW outputs the optimal warping path between two time series as opposed to a soft approximation, recoverable from Soft-DTW. We show that this property is particularly useful for applications where downstream loss functions are defined on the optimal alignment path itself. This naturally occurs, for instance, when learning to improve the accuracy of predicted alignments against ground truth alignments. We evaluate DecDTW on two such applications, namely the audio-to-score alignment task in music information retrieval and the visual place recognition task in robotics, demonstrating state-of-the-art results in both. | Published as a conference paper at ICLR 2023 DEEP DECLARATIVE DYNAMIC TIME WARPING FOR END-TO-END LEARNING OF ALIGNMENT PATHS |
d233004606 | There has been increasing interest in building deep hierarchy-aware classifiers that aim to quantify and reduce the severity of mistakes, and not just reduce the number of errors. The idea is to exploit the label hierarchy (e.g., the WordNet ontology) and consider graph distances as a proxy for mistake severity. Surprisingly, on examining mistake-severity distributions of the top-1 prediction, we find that current state-of-the-art hierarchy-aware deep classifiers do not always show practical improvement over the standard cross-entropy baseline in making better mistakes. The reason for the reduction in average mistake-severity can be attributed to the increase in low-severity mistakes, which may also explain the noticeable drop in their accuracy. To this end, we use the classical Conditional Risk Minimization (CRM) framework for hierarchy aware classification. Given a cost matrix and a reliable estimate of likelihoods (obtained from a trained network), CRM simply amends mistakes at inference time; it needs no extra hyperparameters, and requires adding just a few lines of code to the standard cross-entropy baseline. It significantly outperforms the state-of-the-art and consistently obtains large reductions in the average hierarchical distance of top-k predictions across datasets, with very little loss in accuracy. CRM, because of its simplicity, can be used with any off-the-shelf trained model that provides reliable likelihood estimates. | Published as a conference paper at ICLR 2021 NO COST LIKELIHOOD MANIPULATION AT TEST TIME FOR MAKING BETTER MISTAKES IN DEEP NETWORKS |
d246294519 | We propose to identify directions invariant to a given classifier so that these directions can be controlled in tasks such as style transfer. While orthogonal decomposition is directly identifiable when the given classifier is linear, we formally define a notion of orthogonality in the non-linear case. We also provide a surprisingly simple method for constructing the orthogonal classifier (a classifier utilizing directions other than those of the given classifier). Empirically, we present three use cases where controlling orthogonal variation is important: style transfer, domain adaptation, and fairness. The orthogonal classifier enables desired style transfer when domains vary in multiple aspects, improves domain adaptation with label shifts and mitigates the unfairness as a predictor. The code is available at . Shape-texture debiased neural network training. ArXiv, abs/2010.05981, 2021.Etai Littwin and Lior Wolf. The multiverse loss for robust transfer learning. . Evading the simplicity bias: Training a diverse set of models discovers solutions with superior ood generalization. ArXiv, abs/2105.05612, 2021. | Published as a conference paper at ICLR 2022 CONTROLLING DIRECTIONS ORTHOGONAL TO A CLASSIFIER |
d229331643 | Numerical experiments demonstrate that deep neural network classifiers progressively separate class distributions around their mean, achieving linear separability on the training set, and increasing the Fisher discriminant ratio. We explain this mechanism with two types of operators. We prove that a rectifier without biases applied to sign-invariant tight frames can separate class means and increase Fisher ratios. On the opposite, a soft-thresholding on tight frames can reduce withinclass variabilities while preserving class means. Variance reduction bounds are proved for Gaussian mixture models. For image classification, we show that separation of class means can be achieved with rectified wavelet tight frames that are not learned. It defines a scattering transform. Learning 1 × 1 convolutional tight frames along scattering channels and applying a soft-thresholding reduces within-class variabilities. The resulting scattering network reaches the classification accuracy of ResNet-18 on CIFAR-10 and ImageNet, with fewer layers and no learned biases. | SEPARATION AND CONCENTRATION IN DEEP NET- WORKS |
d236635216 | Time series data introduces two key challenges for explainability methods: firstly, observations of the same feature over subsequent time steps are not independent, and secondly, the same feature can have varying importance to model predictions over time. In this paper, we propose Windowed Feature Importance in Time (WinIT), a feature removal based explainability approach to address these issues. Unlike existing feature removal explanation methods, WinIT explicitly accounts for the temporal dependence between different observations of the same feature in the construction of its importance score. Furthermore, WinIT captures the varying importance of a feature over time, by summarizing its importance over a window of past time steps. We conduct an extensive empirical study on synthetic and real-world data, compare against a wide range of leading explainability methods, and explore the impact of various evaluation strategies. Our results show that WinIT achieves significant gains over existing methods, with more consistent performance across different evaluation metrics. | Published as a conference paper at ICLR 2023 TEMPORAL DEPENDENCIES IN FEATURE IMPORTANCE FOR TIME SERIES PREDICTION |
d256389851 | Since the introduction of Vision Transformers, the landscape of many computer vision tasks (e.g., semantic segmentation), which has been overwhelmingly dominated by CNNs, recently has significantly revolutionized. However, the computational cost and memory requirement render these methods unsuitable on the mobile device, especially for the high-resolution per-pixel semantic segmentation task. In this paper, we introduce a new method squeeze-enhanced Axial Transformer (SeaFormer) for mobile semantic segmentation. Specifically, we design a generic attention block characterized by the formulation of squeeze Axial and detail enhancement. It can be further used to create a family of backbone architectures with superior cost-effectiveness. Coupled with a light segmentation head, we achieve the best trade-off between segmentation accuracy and latency on the ARM-based mobile devices on the ADE20K and Cityscapes datasets. Critically, we beat both the mobile-friendly rivals and Transformer-based counterparts with better performance and lower latency without bells and whistles. Beyond semantic segmentation, we further apply the proposed SeaFormer architecture to image classification problem, demonstrating the potentials of serving as a versatile mobile-friendly backbone. Our code and models are made publicly available at https://github.com/fudan-zvg/SeaFormer. | Published as a conference paper at ICLR 2023 SEAFORMER: SQUEEZE-ENHANCED AXIAL TRANS- FORMER FOR MOBILE SEMANTIC SEGMENTATION |
d231648154 | Graph-structured data ubiquitously appears in science and engineering. Graph neural networks (GNNs) are designed to exploit the relational inductive bias exhibited in graphs; they have been shown to outperform other forms of neural networks in scenarios where structure information supplements node features. The most common GNN architecture aggregates information from neighborhoods based on message passing. Its generality has made it broadly applicable. In this paper, we focus on a special, yet widely used, type of graphs-DAGs-and inject a stronger inductive bias-partial ordering-into the neural network design. We propose the directed acyclic graph neural network, DAGNN, an architecture that processes information according to the flow defined by the partial order. DAGNN can be considered a framework that entails earlier works as special cases (e.g., models for trees and models updating node representations recurrently), but we identify several crucial components that prior architectures lack. We perform comprehensive experiments, including ablation studies, on representative DAG datasets (i.e., source code, neural architectures, and probabilistic graphical models) and demonstrate the superiority of DAGNN over simpler DAG architectures as well as general graph architectures. | Published as a conference paper at ICLR 2021 DIRECTED ACYCLIC GRAPH NEURAL NETWORKS |
d232046284 | The privacy leakage of the model about the training data can be bounded in the differential privacy mechanism. However, for meaningful privacy parameters, a differentially private model degrades the utility drastically when the model comprises a large number of trainable parameters. In this paper, we propose an algorithm Gradient Embedding Perturbation (GEP) towards training differentially private deep models with decent accuracy. Specifically, in each gradient descent step, GEP first projects individual private gradient into a non-sensitive anchor subspace, producing a low-dimensional gradient embedding and a small-norm residual gradient. Then, GEP perturbs the low-dimensional embedding and the residual gradient separately according to the privacy budget. Such a decomposition permits a small perturbation variance, which greatly helps to break the dimensional barrier of private learning. With GEP, we achieve decent accuracy with reasonable computational cost and modest privacy guarantee for deep models. Especially, with privacy bound = 8, we achieve 74.9% test accuracy on CIFAR10 and 95.1% test accuracy on SVHN, significantly improving over existing results. * Authors contribute equally to this work. | Published as a conference paper at ICLR 2021 DO NOT LET PRIVACY OVERBILL UTILITY: GRADIENT EMBEDDING PERTURBATION FOR PRIVATE LEARNING |
d247594025 | Sequential training from task to task is becoming one of the major objects in deep learning applications such as continual learning and transfer learning. Nevertheless, it remains unclear under what conditions the trained model's performance improves or deteriorates. To deepen our understanding of sequential training, this study provides a theoretical analysis of generalization performance in a solvable case of continual learning. We consider neural networks in the neural tangent kernel (NTK) regime that continually learn target functions from task to task, and investigate the generalization by using an established statistical mechanical analysis of kernel ridge-less regression. We first show characteristic transitions from positive to negative transfer. More similar targets above a specific critical value can achieve positive knowledge transfer for the subsequent task while catastrophic forgetting occurs even with very similar targets. Next, we investigate a variant of continual learning which supposes the same target function in multiple tasks. Even for the same target, the trained model shows some transfer and forgetting depending on the sample size of each task. We can guarantee that the generalization error monotonically decreases from task to task for equal sample sizes while unbalanced sample sizes deteriorate the generalization. We respectively refer to these improvement and deterioration as self-knowledge transfer and forgetting, and empirically confirm them in realistic training of deep neural networks as well. | Published as a conference paper at ICLR 2022 LEARNING CURVES FOR CONTINUAL LEARNING IN NEURAL NETWORKS: SELF-KNOWLEDGE TRANSFER AND FORGETTING |
d15538683 | Recent work has shown deep neural networks (DNNs) to be highly susceptible to well-designed, small perturbations at the input layer, or so-called adversarial examples. Taking images as an example, such distortions are often imperceptible, but can result in 100% mis-classification for a state of the art DNN. We study the structure of adversarial examples and explore network topology, pre-processing and training strategies to improve the robustness of DNNs. We perform various experiments to assess the removability of adversarial examples by corrupting with additional noise and pre-processing with denoising autoencoders (DAEs). We find that DAEs can remove substantial amounts of the adversarial noise. However, when stacking the DAE with the original DNN, the resulting network can again be attacked by new adversarial examples with even smaller distortion. As a solution, we propose Deep Contractive Network, a model with a new end-to-end training procedure that includes a smoothness penalty inspired by the contractive autoencoder (CAE). This increases the network robustness to adversarial examples, without a significant performance penalty. | TOWARDS DEEP NEURAL NETWORK ARCHITECTURES ROBUST TO ADVERSARIAL EXAMPLES |
d258967945 | Large pre-trained language models (PLMs) have demonstrated strong performance on natural language understanding (NLU) tasks through fine-tuning. However, fine-tuned models still suffer from overconfident predictions, especially in out-of-domain settings. In this paper, we tackle the problem of calibrating finetuned language models. We demonstrate that the PLMs are well-calibrated on the masked language modeling task with robust predictive confidence under domain shift, yet the fine-tuned models fail to retain such property due to catastrophic forgetting, which impacts the calibration on the downstream classification task. In light of these observations, we evaluate the calibration of several methods that preserve pre-trained features and show that preserving pre-trained features can improve the calibration of fine-tuned language models. Among these methods, our proposed method that encourages the fine-tuned model to learn generative representations with auxiliary language modeling objective achieves competitive accuracy and the lowest expected calibration error compared to several strong baselines under both in-domain and out-of-domain settings on three downstream NLU tasks. * Corresponding authors. | Published as a conference paper at ICLR 2023 PRESERVING PRE-TRAINED FEATURES HELPS CALIBRATE FINE-TUNED LANGUAGE MODELS |
d2479619 | Recent work has established an empirically successful framework for adapting learning rates for stochastic gradient descent (SGD). This effectively removes all needs for tuning, while automatically reducing learning rates over time on stationary problems, and permitting learning rates to grow appropriately in nonstationary tasks. Here, we extend the idea in three directions, addressing proper minibatch parallelization, including reweighted updates for sparse or orthogonal gradients, improving robustness on non-smooth loss functions, in the process replacing the diagonal Hessian estimation procedure that may not always be available by a robust finite-difference approximation. The final algorithm integrates all these components, has linear complexity and is hyper-parameter free. | Adaptive learning rates and parallelization for stochastic, sparse, non-smooth gradients |
d7774489 | Deep reinforcement learning has achieved many impressive results in recent years. However, tasks with sparse rewards or long horizons continue to pose significant challenges. To tackle these important problems, we propose a general framework that first learns useful skills in a pre-training environment, and then leverages the acquired skills for learning faster in downstream tasks. Our approach brings together some of the strengths of intrinsic motivation and hierarchical methods: the learning of useful skill is guided by a single proxy reward, the design of which requires very minimal domain knowledge about the downstream tasks. Then a high-level policy is trained on top of these skills, providing a significant improvement of the exploration and allowing to tackle sparse rewards in the downstream tasks. To efficiently pre-train a large span of skills, we use Stochastic Neural Networks combined with an information-theoretic regularizer. Our experiments 1 show 2 that this combination is effective in learning a wide span of interpretable skills in a sample-efficient way, and can significantly boost the learning performance uniformly across a wide range of downstream tasks. | Published as a conference paper at ICLR 2017 STOCHASTIC NEURAL NETWORKS FOR HIERARCHICAL REINFORCEMENT LEARNING |
d3635880 | Model pruning has become a useful technique that improves the computational efficiency of deep learning, making it possible to deploy solutions in resourcelimited scenarios. A widely-used practice in relevant work assumes that a smallernorm parameter or feature plays a less informative role at the inference time. In this paper, we propose a channel pruning technique for accelerating the computations of deep convolutional neural networks (CNNs) that does not critically rely on this assumption. Instead, it focuses on direct simplification of the channel-tochannel computation graph of a CNN without the need of performing a computationally difficult and not-always-useful task of making high-dimensional tensors of CNN structured sparse. Our approach takes two stages: first to adopt an end-toend stochastic training method that eventually forces the outputs of some channels to be constant, and then to prune those constant channels from the original neural network by adjusting the biases of their impacting layers such that the resulting compact model can be quickly fine-tuned. Our approach is mathematically appealing from an optimization perspective and easy to reproduce. We experimented our approach through several image learning benchmarks and demonstrate its interesting aspects and competitive performance.Published as a conference paper at ICLR 2018 focus on setting the dense weights of convolutions or linear maps to be structured sparse, we propose here a method adopting a new conception to achieve in effect the same goal. | Published as a conference paper at ICLR 2018 RETHINKING THE SMALLER-NORM-LESS- INFORMATIVE ASSUMPTION IN CHANNEL PRUNING OF CONVOLUTION LAYERS |
d1793573 | Deep learning has recently led to great successes in tasks such as image recognition (e.gKrizhevsky et al., 2012). However, deep networks are still outmatched by the power and versatility of the brain, perhaps in part due to the richer neuronal computations available to cortical circuits. The challenge is to identify which neuronal mechanisms are relevant, and to find suitable abstractions to model them. Here, we show how aspects of spike timing, long hypothesized to play a crucial role in cortical information processing, could be incorporated into deep networks to build richer, versatile representations.We introduce a neural network formulation based on complex-valued neuronal units that is not only biologically meaningful but also amenable to a variety of deep learning frameworks. Here, units are attributed both a firing rate and a phase, the latter indicating properties of spike timing. We show how this formulation qualitatively captures several aspects thought to be related to neuronal synchrony, including gating of information processing and dynamic binding of distributed object representations. Focusing on the latter, we demonstrate the potential of the approach in several simple experiments. Thus, neuronal synchrony could be a flexible mechanism that fulfills multiple functional roles in deep networks. | Neuronal Synchrony in Complex-Valued Deep Networks |
d256459658 | Prompt tuning with large-scale pretrained vision-language models empowers open-vocabulary predictions trained on limited base categories, e.g., object classification and detection. In this paper, we propose compositional prompt tuning with motion cues: an extended prompt tuning paradigm for compositional predictions of video data. In particular, we present Relation Prompt (RePro) for Open-vocabulary Video Visual Relation Detection (Open-VidVRD), where conventional prompt tuning is easily biased to certain subject-object combinations and motion patterns. To this end, RePro addresses the two technical challenges of Open-VidVRD: 1) the prompt tokens should respect the two different semantic roles of subject and object, and 2) the tuning should account for the diverse spatio-temporal motion patterns of the subject-object compositions. Without bells and whistles, our RePro achieves a new state-of-the-art performance on two Vid-VRD benchmarks of not only the base training object and predicate categories, but also the unseen ones. Extensive ablations also demonstrate the effectiveness of the proposed compositional and multi-mode design of prompts. | Published as a conference paper at ICLR 2023 COMPOSITIONAL PROMPT TUNING WITH MOTION CUES FOR OPEN-VOCABULARY VIDEO RELATION DETECTION |
d231627730 | Using a high Update-To-Data (UTD) ratio, model-based methods have recently achieved much higher sample efficiency than previous model-free methods for continuous-action DRL benchmarks. In this paper, we introduce a simple modelfree algorithm, Randomized Ensembled Double Q-Learning (REDQ), and show that its performance is just as good as, if not better than, a state-of-the-art modelbased algorithm for the MuJoCo benchmark. Moreover, REDQ can achieve this performance using fewer parameters than the model-based method, and with less wall-clock run time. REDQ has three carefully integrated ingredients which allow it to achieve its high performance: (i) a UTD ratio 1; (ii) an ensemble of Q functions; (iii) in-target minimization across a random subset of Q functions from the ensemble. Through carefully designed experiments, we provide a detailed analysis of REDQ and related model-free algorithms. To our knowledge, REDQ is the first successful model-free DRL algorithm for continuous-action spaces using a UTD ratio 1. * Equal contribution, in alphabetical order. † Correspondence to: Keith Ross <keithwross@nyu.edu>. 1 arXiv:2101.05982v1 [cs.LG] 15 Jan 2021 Preprint, under review.zero for most of training, even when the UTD is very high. Furthermore, by adjusting the number of randomly selected Q-functions for in-target minimization, REDQ can control the average Q-function bias. In comparison with standard ensemble averaging and with SAC with a higher UTD, REDQ has much lower std of Q-function bias while maintaining an average bias that is negative but close to zero throughout most of training, resulting in significantly better learning performance. We perform an ablation study, and show that REDQ is very robust to choices of hyperparameters, and can work well with a small ensemble and a small number of Q functions in the in-target minimization. We also provide a theoretical analysis, providing additional insights into REDQ. Finally, we consider combining the REDQ algorithm with an online feature extractor network (OFENet) (Ota et al., 2020) to further improve performance, particularly for the more challenging environments Ant and Humanoid. We achieve more than 7x the sample efficiency of SAC to reach a score of 5000 for both Ant and Humanoid. In Humanoid, REDQ-OFE also greatly outperforms MBPO, reaching a score of 5000 at 150K interactions, which is 3x MBPO's score at that point.To ensure our comparisons are fair, and to ensure our results are reproducible(Henderson et al., 2018;Islam et al., 2017;Duan et al., 2016), we provide open source code. For all algorithmic comparisons, we use the same codebase (except for MBPO, for which we use the authors' code).RANDOMIZED ENSEMBLED DOUBLE Q-LEARNING (REDQ)Janner et al.(2019)proposed Model-Based Policy Optimization (MBPO), which was shown to be much more sample efficient than popular model-free algorithms such as SAC and PPO for the MuJoCo environments. MBPO learns a model, and generates "fake data" from its model as well as "real data" through environment interactions. It then performs parameter updates using both the fake and the real data. One of the distinguishing features of MBPO is that it has a UTD ratio 1 for updating its Q functions, enabling MBPO to achieve high sample efficiency. | Preprint, under review. RANDOMIZED ENSEMBLED DOUBLE Q-LEARNING: LEARNING FAST WITHOUT A MODEL |
d219531210 | Recent progress in pre-trained neural language models has significantly improved the performance of many natural language processing (NLP) tasks. In this paper we propose a new model architecture DeBERTa (Decoding-enhanced BERT with disentangled attention) that improves the BERT and RoBERTa models using two novel techniques. The first is the disentangled attention mechanism, where each word is represented using two vectors that encode its content and position, respectively, and the attention weights among words are computed using disentangled matrices on their contents and relative positions, respectively. Second, an enhanced mask decoder is used to incorporate absolute positions in the decoding layer to predict the masked tokens in model pre-training. In addition, a new virtual adversarial training method is used for fine-tuning to improve models' generalization. We show that these techniques significantly improve the efficiency of model pre-training and the performance of both natural language understand (NLU) and natural langauge generation (NLG) downstream tasks. Compared to RoBERTa-Large, a DeBERTa model trained on half of the training data performs consistently better on a wide range of NLP tasks, achieving improvements on MNLI by +0.9% (90.2% vs. 91.1%), on SQuAD v2.0 by +2.3% (88.4% vs. 90.7%) and RACE by +3.6% (83.2% vs. 86.8%). Notably, we scale up DeBERTa by training a larger version that consists of 48 Transform layers with 1.5 billion parameters. The significant performance boost makes the single DeBERTa model surpass the human performance on the SuperGLUE benchmark(Wang et al., 2019a)for the first time in terms of macro-average score (89.9 versus 89.8), and the ensemble DeBERTa model sits atop the SuperGLUE leaderboard as of January 6, 2021, outperforming the human baseline by a decent margin (90.3 versus 89.8). The pre-trained DeBERTa models and the source code were released at: https://github.com/microsoft/DeBERTa 1 . | Published as a conference paper at ICLR 2021 DEBERTA: DECODING-ENHANCED BERT WITH DIS- ENTANGLED ATTENTION |
d238744365 | Recently, developed a new neural network architecture based on ∞ -distance functions, which naturally possesses certified ∞ robustness by its construction. Despite the novel design and theoretical foundation, so far the model only achieved comparable performance to conventional networks. In this paper, we make the following two contributions: (i) We demonstrate that ∞ -distance nets enjoy a fundamental advantage in certified robustness over conventional networks (under typical certification approaches); (ii) With an improved training process we are able to significantly boost the certified accuracy of ∞ -distance nets. Our training approach largely alleviates the optimization problem that arose in the previous training scheme, in particular, the unexpected large Lipschitz constant due to the use of a crucial trick called p -relaxation. The core of our training approach is a novel objective function that combines scaled cross-entropy loss and clipped hinge loss with a decaying mixing coefficient. Experiments show that using the proposed training strategy, the certified accuracy of ∞ -distance net can be dramatically improved from 33.30% to 40.06% on CIFAR-10 ( = 8/255), meanwhile outperforming other approaches in this area by a large margin. Our results clearly demonstrate the effectiveness and potential of ∞ -distance net for certified robustness. Codes are available at https://github.com/zbh2047/L inf-dist-net-v2.Published as a conference paper at ICLR 2022 point according to the output margin. The whole procedure only requires a forward pass without any additional computation. The authors further showed that the model family has strong expressive power, e.g., a large enough ∞ -distance net can approximate any 1-Lipschitz function on a bounded domain. Unfortunately, however, the empirical model performance did not well reflect the theoretical advantages. As shown in , it is necessary to use a conventional multi-layer perception (MLP) 1 on top of an ∞ -distance net backbone to achieve better performance compared to the baseline methods. It makes both the training and the certification procedure complicated. More importantly, it calls into question whether the ∞ -distance net is really a better model configuration than conventional architectures in the regime of certified robustness.In this paper, we give an affirmative answer by showing that ∞ -distance net itself suffices for good performance and can be well learned using an improved training strategy. We first mathematically prove that under mild assumptions of the dataset, there exists an ∞ -distance net with reasonable size by construction that achieves perfect certified robustness. This result indicates the strong expressive power of ∞ -distance nets in robustness certification, and shows a fundamental advantage over conventional networks under typical certification approaches (which do not possess such expressive power according toMirman et al. (2021)). However, it seems to contradict the previous empirical observations, suggesting that the model may fail to find an optimal solution and further motivating us to revisit the optimization process designed in .Due to the non-smoothness of the ∞ -distance function, Zhang et al.(2021)developed several training tricks to overcome the optimization difficulty. A notable trick is called the p -relaxation, in which p -distance neurons are used during optimization to give a smooth approximation of ∞ -distance. However, we find that the relaxation on neurons unexpectedly relaxes the Lipschitz constant of the network to an exponentially large value, making the objective function no longer maximize the robust accuracy and leading to sub-optimal solutions. | Published as a conference paper at ICLR 2022 BOOSTING THE CERTIFIED ROBUSTNESS OF L-INFINITY DISTANCE NETS |
d3513418 | Due to their complex nature, it is hard to characterize the ways in which machine learning models can misbehave or be exploited when deployed. Recent work on adversarial examples, i.e. inputs with minor perturbations that result in substantially different model predictions, is helpful in evaluating the robustness of these models by exposing the adversarial scenarios where they fail. However, these malicious perturbations are often unnatural, not semantically meaningful, and not applicable to complicated domains such as language. In this paper, we propose a framework to generate natural and legible adversarial examples by searching in semantic space of dense and continuous data representation, utilizing the recent advances in generative adversarial networks. We present generated adversaries to demonstrate the potential of the proposed approach for black-box classifiers in a wide range of applications such as image classification, textual entailment, and machine translation. We include experiments to show that the generated adversaries are natural, legible to humans, and useful in evaluating and analyzing black-box classifiers. | Under review as a conference paper at ICLR 2018 GENERATING NATURAL ADVERSARIAL EXAMPLES |
d213597045 | Deep reinforcement learning (RL) agents often fail to generalize to unseen environments (yet semantically similar to trained agents), particularly when they are trained on high-dimensional state spaces, such as images. In this paper, we propose a simple technique to improve a generalization ability of deep RL agents by introducing a randomized (convolutional) neural network that randomly perturbs input observations. It enables trained agents to adapt to new domains by learning robust features invariant across varied and randomized environments. Furthermore, we consider an inference method based on the Monte Carlo approximation to reduce the variance induced by this randomization. We demonstrate the superiority of our method across 2D CoinRun, 3D DeepMind Lab exploration and 3D robotics control tasks: it significantly outperforms various regularization and data augmentation methods for the same purpose. Code is available at . Deep direct reinforcement learning for financial signal representation and trading. IEEE transactions on neural networks and learning systems, 28(3):653-664, 2016.Terrance DeVries and Graham W Taylor. Improved regularization of convolutional neural networks with cutout. arXiv preprint arXiv:1708.04552, 2017. , et al. Impala: Scalable distributed deep-rl with importance weighted actor-learner architectures. In ICML, 2018. | Published as a conference paper at ICLR 2020 NETWORK RANDOMIZATION: A SIMPLE TECHNIQUE FOR GENERALIZATION IN DEEP REINFORCEMENT LEARNING |
d52938034 | We propose and study a method for learning interpretable representations for the task of regression. Features are represented as networks of multi-type expression trees comprised of activation functions common in neural networks in addition to other elementary functions. Differentiable features are trained via gradient descent, and the performance of features in a linear model is used to weight the rate of change among subcomponents of each representation. The search process maintains an archive of representations with accuracy-complexity trade-offs to assist in generalization and interpretation. We compare several stochastic optimization approaches within this framework. We benchmark these variants on 100 open-source regression problems in comparison to state-of-the-art machine learning approaches. Our main finding is that this approach produces the highest average test scores across problems while producing representations that are orders of magnitude smaller than the next best performing method (gradient boosting). We also report a negative result in which attempts to directly optimize the disentanglement of the representation result in more highly correlated features. that offer insight by virtue of their simplicity, in a similar vein to models built from first-principles (e.g.(Tibshirani, 1996;Schmidt & Lipson, 2009)). Akin to the latter group, our goal is to discover the simplest description of a process whose predictions generalize as well as possible.Good representations should also disentangle the factors of variation(Bengio et al., 2013)in the data, in order to ease model interpretation. Disentanglement implies functional modularity; i.e., sub-components of the network should encapsulate behaviors that model a sub-process of the task. In this sense, stochastic methods such as evolutionary computation (EC) appear well-motivated, as they are premised on the identification and propagation of building blocks of solutions(Holland, 1975). Experiments with EC applied to networks suggest it pressures networks to be modular(Huizinga et al., 2014;Kashtan & Alon, 2005). Although the identification functional building blocks of solutions sounds ideal, we have no way of knowing a priori whether a given problem will admit the identification of building blocks of solutions via heuristic search(Oppacher, 2014). Our goal in this paper is thus to empirically assess the performance of several SO approaches in a system designed to produce intelligible representations from NN building blocks for regression.In Section 2, we introduce a new method for optimizing representations that we call the feature engineering automation tool (FEAT) 1 . The purpose of this method is to optimize an archive of representations that characterize the trade-off between conciseness and accuracy among representations. Algorithmically, two aspects of the method distinguish FEAT from previous work. First, it represents the internal structure of each NN as a set of syntax trees, with the goal of improving the transparency of the resultant representations. Second, it uses weights learned via gradient descent to provide feedback to the variation process at a more granular level. We compare several multi-objective variants of this approach using EC and non-EC methods with different sets of objectives.We discuss related work in more detail in Section 3. In section 4 and 5, we describe and conduct an experiment that benchmarks FEAT against state-of-the-art ML methods on 100 open-source regression problems. Future work based on this analysis is discussed in Section 6, and additional detailed results are provided in the Appendix. | LEARNING CONCISE REPRESENTATIONS FOR REGRES- SION BY EVOLVING NETWORKS OF TREES |
d246442139 | There is a fundamental limitation in the prediction performance that a machine learning model can achieve due to the inevitable uncertainty of the prediction target. In classification problems, this can be characterized by the Bayes error, which is the best achievable error with any classifier. The Bayes error can be used as a criterion to evaluate classifiers with state-of-the-art performance and can be used to detect test set overfitting. We propose a simple and direct Bayes error estimator, where we just take the mean of the labels that show uncertainty of the class assignments. Our flexible approach enables us to perform Bayes error estimation even for weakly supervised data. In contrast to others, our method is model-free and even instancefree. Moreover, it has no hyperparameters and gives a more accurate estimate of the Bayes error than several baselines empirically. Experiments using our method suggest that recently proposed deep networks such as the Vision Transformer may have reached, or is about to reach, the Bayes error for benchmark datasets. Finally, we discuss how we can study the inherent difficulty of the acceptance/rejection decision for scientific articles, by estimating the Bayes error of the ICLR papers from 2017 to 2023.Published as a conference paper at ICLR 2023ICLR et al., 2021 Theisen et al., 2021). To the best of our knowledge, all previous papers have proposed ways to estimate the Bayes error from a dataset consisting of pairs of instances and their hard labels. When instances and hard labels are available, one can also train a supervised classifier, which is known to approach the Bayes classifier (that achieves the Bayes error) with sufficient training data provided that the model is correctly specified. This is an interesting research problem from the point of view of Vapnik's principle 2 (Vapnik, 2000) since we can derive the Bayes error from the Bayes classifier (and the underlying distribution) while we cannot recover the Bayes classifier from the knowledge of the Bayes error, which is just a scalar. How can we take full advantage of this property? While the Bayes error is usually defined as the best achievable expected error with any measurable function, it is known to be equivalent to the expectation of the minimum of class-posteriors with respect to classes for binary classification. Inspired by Vapnik's principle, our main idea is to skip the intermediate step of learning a function model, and we directly approximate the minimum of the class-posteriors by using soft labels (corresponding to the class probability) or uncertainty labels (corresponding to the class uncertainty). 3Our proposed method has two benefits. Firstly, our method is model-free. Since we do not learn a model, we can escape the curse of dimensionality, while dealing with high-dimensional instances would cause issues such as overfitting if we were to train a model. High dimensionality may cause performance deterioration for other Bayes error estimation methods(Berisha et al., 2016;Noshad et al., 2019) due to divergence estimation. We experimentally show how our method can more accurately estimate the Bayes error than baselines that utilize instances and soft labels. Our modelfree method is also extremely fast since we do not have any hyperparameters to tune nor a function model to train.The second benefit is a more practical one: our method is completely instance-free. Suppose our final goal is to estimate the Bayes error instead of training a classifier. In that case, we do not need to collect instance-label pairs, and it may be less costly to collect soft/uncertainty labels without instances. Dealing with instances can cause privacy issues, and it can be expensive due to data storage costs especially when they are high-dimensional or can come in large quantities. It may lead to security costs to protect instances from a data breach. As an example of an instance-free scenario, we can consider doctors who are diagnosing patients by inspecting symptoms and asking questions, without explicitly collecting or storing the patients' data in the database. In this scenario, the hospital will only have the decisions and confidence of doctors, which can be used as soft labels.The contributions of the paper is as follows. We first propose a direct way to estimate the Bayes error from soft (or uncertainty) labels without a model nor instances. We show that our estimator is unbiased and consistent. In practice, collecting soft/uncertainty labels can be difficult since the labelling process can become noisy. We propose a modified estimator that is still unbiased and consistent even when the soft labels are contaminated with zero-mean noise. We also show that our approach can be applied to other classification problems, such as weakly supervised learning (Sugiyama et al., 2022). Finally, we show the proposed methods' behavior through various experiments. Our results suggest that recently proposed deep networks such as the Vision Transformer (Dosovitskiy et al., 2021) has reached or is about to reach the Bayes error for benchmark datasets, such as CIFAR-10H (Peterson et al., 2019) and Fashion-MNIST-H (which is a new dataset we present; explained in Sec. 5.3). We also demonstrate how our proposed method can be used to estimate the Bayes error for academic conferences such as ICLR, by regarding them as an accept/reject binary classification problem. | IS THE PERFORMANCE OF MY DEEP NETWORK TOO GOOD TO BE TRUE? A DIRECT APPROACH TO ESTIMAT- ING THE BAYES ERROR IN BINARY CLASSIFICATION |
d18424928 | The human visual system has a hierarchical structure consisting of layers of processing, such as the retina, V1, V2, etc. Understanding the functional roles of these visual processing layers would help to integrate the psychophysiological and neurophysiological models into a consistent theory of human vision, and would also provide insights to computer vision research. One classical theory of the early visual pathway hypothesizes that it serves to capture the statistical structure of the visual inputs by efficiently coding the visual information in its outputs. Until recently, most computational models following this theory have focused upon explaining the receptive field properties of one or two visual layers. Recent work in deep networks has eliminated this concern, however, there is till the retinal layer to consider. Here we improve on a previouslydescribed hierarchical model Recursive ICA (RICA) [1] which starts with PCA, followed by a layer of sparse coding or ICA, followed by a component-wise nonlinearity derived from considerations of the variable distributions expected by ICA. This process is then repeated. In this work, we improve on this model by using a new version of sparse PCA (sPCA), which results in biologicallyplausible receptive fields for both the sPCA and ICA/sparse coding. When applied to natural image patches, our model learns visual features exhibiting the receptive field properties of retinal ganglion cells/lateral geniculate nucleus (LGN) cells, V1 simple cells, V1 complex cells, and V2 cells. Our work provides predictions for experimental neuroscience studies. For example, our result suggests that a previous neurophysiological study improperly discarded some of their recorded neurons; we predict that their discarded neurons capture the shape contour of objects. | Efficient Visual Coding: From Retina To V2 |
d59600012 | Generalization error (also known as the out-of-sample error) measures how well the hypothesis learned from training data generalizes to previously unseen data. Proving tight generalization error bounds is a central question in statistical learning theory. In this paper, we obtain generalization error bounds for learning general non-convex objectives, which has attracted significant attention in recent years. We develop a new framework, termed Bayes-Stability, for proving algorithm-dependent generalization error bounds. The new framework combines ideas from both the PAC-Bayesian theory and the notion of algorithmic stability. Applying the Bayes-Stability method, we obtain new data-dependent generalization bounds for stochastic gradient Langevin dynamics (SGLD) and several other noisy gradient methods (e.g., with momentum, mini-batch and acceleration, Entropy-SGD). Our result recovers (and is typically tighter than) a recent result in Mou et al. (2018) and improves upon the results in Pensia et al. (2018). Our experiments demonstrate that our data-dependent bounds can distinguish randomly labelled data from normal data, which provides an explanation to the intriguing phenomena observed in Zhang et al. (2017a). We also study the setting where the total loss is the sum of a bounded loss and an additional 2 regularization term. We obtain new generalization bounds for the continuous Langevin dynamic in this setting by developing a new Log-Sobolev inequality for the parameter distribution at any time. Our new bounds are more desirable when the noise level of the process is not very small, and do not become vacuous even when T tends to infinity. * Published as a conference paper at ICLR 2020Extensions. We remark that our technique allows for an arguably simpler proof of (Mou et al., 2018, Theorem 1); the original proof is based on SDE and Fokker-Planck equation. More importantly, our technique can be easily extended to handle mini-batches and a variety of general settings as follows.1. Extension to other gradient-based methods. Our results naturally extends to other noisy stochastic gradient methods including momentum due to Polyak (1964) (Theorem 26), Nes-1. Batch size b ≤ n/2. | Published as a conference paper at ICLR 2020 ON GENERALIZATION ERROR BOUNDS OF NOISY GRADIENT METHODS FOR NON-CONVEX LEARNING |
d253107541 | Overparameterization in deep learning typically refers to settings where a trained neural network (NN) has representational capacity to fit the training data in many ways, some of which generalize well, while others do not. In the case of Recurrent Neural Networks (RNNs), there exists an additional layer of overparameterization, in the sense that a model may exhibit many solutions that generalize well for sequence lengths seen in training, some of which extrapolate to longer sequences, while others do not. Numerous works have studied the tendency of Gradient Descent (GD) to fit overparameterized NNs with solutions that generalize well. On the other hand, its tendency to fit overparameterized RNNs with solutions that extrapolate has been discovered only recently and is far less understood.In this paper, we analyze the extrapolation properties of GD when applied to overparameterized linear RNNs. In contrast to recent arguments suggesting an implicit bias towards short-term memory, we provide theoretical evidence for learning low-dimensional state spaces, which can also model long-term memory. Our result relies on a dynamical characterization which shows that GD (with small step size and near-zero initialization) strives to maintain a certain form of balancedness, as well as on tools developed in the context of the moment problem from statistics (recovery of a probability distribution from its moments). Experiments corroborate our theory, demonstrating extrapolation via learning low-dimensional state spaces with both linear and non-linear RNNs.Published as a conference paper at ICLR 2023 GD has been argued to exhibit an implicit bias towards short-term memory. While such results are informative, their generality remains in question, particularly since infinitely wide NNs are known to substantially differ from their finite-width counterparts, and since a memoryless teacher essentially neglects the main characteristic of RNNs (memory).In this paper, we theoretically investigate the implicit extrapolation of GD when applied to overparameterized finite-width linear RNNs learning from a teacher with memory. We consider models with symmetric transition matrices, in the case where a student (learned model) with state space dimension d is trained on sequences of length k generated by a teacher with state space dimensiond. Our interest lies in the overparameterized regime, where d is greater than both k andd, meaning that the student has state space dimensions large enough to fully agree with the teacher on sequences of length k, while potentially disagreeing with it on longer sequences. As a necessary assumption on initialization, we follow prior work and focus on a certain balancedness condition, which is known (see experiments in Cohen-Karlik et al. (2022), as well as our theoretical analysis) to capture near-zero initialization as commonly employed in practice.Our main theoretical result states that GD originating from a balanced initialization leads the student to extrapolate, irrespective of how large its state space dimension is. Key to the result is a surprising connection to a moment matching theorem from Cohen & Yeredor(2011), whose proof relies on ideas from compressed sensing (Elad, 2010; Eldar & Kutyniok, 2012) and neighborly polytopes (Gale, 1963). This connection may be of independent interest, and in particular may prove useful in deriving other results concerning the implicit properties of GD. We corroborate our theory with experiments, which demonstrate extrapolation via learning low-dimensional state spaces in both the analyzed setting and ones involving non-linear RNNs.The implicit extrapolation of GD is an emerging and exciting area of inquiry. Our results suggest that short-term memory is not enough for explaining it as previously believed. We hope the techniques developed in this paper will contribute to a further understanding of this phenomenon.If the above holds for all q P N then the student is said to -extrapolate w.r.t the teacher, and if it holds for all q P N with " 0 then the student is simply said to extrapolate w.r.t the teacher. | Published as a conference paper at ICLR 2023 LEARNING LOW DIMENSIONAL STATE SPACES WITH OVERPARAMETERIZED RECURRENT NEURAL NETS |
d256827133 | Finding the best way to schedule operations in a computation graph is a classical NP-hard problem which is central to compiler optimization. However, evaluating the goodness of a schedule on the target hardware can be very time-consuming. Traditional approaches as well as previous machine learning ones typically optimize proxy metrics, which are fast to evaluate but can lead to bad schedules when tested on the target hardware. In this work, we propose a new approach to scheduling by sampling proportionally to the proxy metric using a novel GFlowNet method. We introduce a technique to control the trade-off between diversity and goodness of the proposed schedules at inference time and demonstrate empirically that the pure optimization baselines can lead to subpar performance with respect to our approach when tested on a target model. Furthermore, we show that conditioning the GFlowNet on the computation graph enables generalization to unseen scheduling problems for both synthetic and real-world compiler datasets. * Work completed during internship at Qualcomm Technologies Netherlands B.V.; Qualcomm AI Research is an initiative of Qualcomm Technologies, Inc. | Published as a conference paper at ICLR 2023 ROBUST SCHEDULING WITH GFLOWNETS |
d225075866 | Graph Neural Networks (GNNs) are the predominant technique for learning over graphs. However, there is relatively little understanding of why GNNs are successful in practice and whether they are necessary for good performance. Here, we show that for many standard transductive node classification benchmarks, we can exceed or match the performance of state-of-the-art GNNs by combining shallow models that ignore the graph structure with two simple post-processing steps that exploit correlation in the label structure: (i) an "error correlation" that spreads residual errors in training data to correct errors in test data and (ii) a "prediction correlation" that smooths the predictions on the test data. We call this overall procedure Correct and Smooth (C&S), and the post-processing steps are implemented via simple modifications to standard label propagation techniques from early graph-based semi-supervised learning methods. Our approach exceeds or nearly matches the performance of state-of-the-art GNNs on a wide variety of benchmarks, with just a small fraction of the parameters and orders of magnitude faster runtime. For instance, we exceed the best known GNN performance on the OGB-Products dataset with 137 times fewer parameters and greater than 100 times less training time. The performance of our methods highlights how directly incorporating label information into the learning algorithm (as was done in traditional techniques) yields easy and substantial performance gains. We can also incorporate our techniques into big GNN models, providing modest gains. Our code for the OGB results is at https://github.com/CUAI/CorrectAndSmooth. | COMBINING LABEL PROPAGATION AND SIMPLE MOD- ELS OUT-PERFORMS GRAPH NEURAL NETWORKS |
d231839638 | Convolutional neural networks (CNNs) constructed natively on the sphere have been developed recently and shown to be highly effective for the analysis of spherical data. While an efficient framework has been formulated, spherical CNNs are nevertheless highly computationally demanding; typically they cannot scale beyond spherical signals of thousands of pixels. We develop scattering networks constructed natively on the sphere that provide a powerful representational space for spherical data. Spherical scattering networks are computationally scalable and exhibit rotational equivariance, while their representational space is invariant to isometries and provides efficient and stable signal representations. By integrating scattering networks as an additional type of layer in the generalized spherical CNN framework, we show how they can be leveraged to scale spherical CNNs to the high-resolution data typical of many practical applications, with spherical signals of many tens of megapixels and beyond. | Published as a conference paper at ICLR 2022 SCATTERING NETWORKS ON THE SPHERE FOR SCALABLE AND ROTATIONALLY EQUIVARIANT SPHERICAL CNNS |
d250334642 | We introduce Joint Multidimensional Scaling, a novel approach for unsupervised manifold alignment, which maps datasets from two different domains, without any known correspondences between data instances across the datasets, to a common low-dimensional Euclidean space. Our approach integrates Multidimensional Scaling (MDS) and Wasserstein Procrustes analysis into a joint optimization problem to simultaneously generate isometric embeddings of data and learn correspondences between instances from two different datasets, while only requiring intra-dataset pairwise dissimilarities as input. This unique characteristic makes our approach applicable to datasets without access to the input features, such as solving the inexact graph matching problem. We propose an alternating optimization scheme to solve the problem that can fully benefit from the optimization techniques for MDS and Wasserstein Procrustes. We demonstrate the effectiveness of our approach in several applications, including joint visualization of two datasets, unsupervised heterogeneous domain adaptation, graph matching, and protein structure alignment.Published as a conference paper at ICLR 2023 One well-known method for aligning data instances from different spaces is Procrustes analysis. When used together with dimensionality reduction, it results in a manifold alignment method(Wang & Mahadevan, 2008;Kohli et al., 2021;Lin et al., 2021). However, these approaches require prior knowledge about the correspondences between data instances across the domains, which limits their applicability in many real-world applications where this information is hard or expensive to obtain. Unsupervised manifold alignment approaches(Wang & Mahadevan, 2009;Cui et al., 2014)have been proposed to overcome this limitation by aligning the underlying manifold structures of two datasets with unknown correspondences while projecting data onto a common low-dimensional space.In this work, we propose to combine MDS with the idea of unsupervised manifold alignment to simultaneously embed data instances from two domains without known correspondences to a common low-dimensional space, while only requiring intra-dataset dissimilarities. We formulate the problem as a joint optimization problem, where we integrate the stress functions for each dataset that measure the distance deviations and adopt the idea of Wasserstein Procrustes analysis (Alvarez-Melis et al., 2019) to align the embedded data instances from two datasets in a fully unsupervised manner. We propose to solve the resulting optimization problem through an alternating optimization strategy, resulting in an algorithm that can benefit from the optimization techniques for solving each individual sub-problem. Our approach, named Joint MDS, allows recovering the correspondences between instances across domains while also producing aligned low-dimensional embeddings for data from both domains, which is the main advantage compared to Gromov-Wasserstein (GW) optimal transport(Mémoli, 2011;Yan et al., 2018)for only correspondence finding. We show the effectiveness of joint MDS in several machine learning applications, including joint visualization of two datasets, unsupervised heterogeneous domain adaptation, graph matching, and protein structure alignment.RELATED WORKWe present here the work most related to ours, namely MDS, unsupervised manifold alignment and optimal transport (OT) for correspondence finding.Multidimensional scaling and extensions MDS is one of the most commonly used dimensionality reduction methods that only require pairwise (dis)similarities between data instances as input. Classical MDS(Torgerson, 1965)was introduced under the assumption that the dissimilarity is an Euclidean distance, which has an analytic solution via SVD. As an extension of classic MDS, metric MDS consists in learning low-dimensional embeddings that preserve any dissimilarity by minimizing a stress function. Several extensions of MDS have also been proposed for various practical reasons, such as non-metric MDS(Agarwal et al., 2007), Isomap(Tenenbaum et al., 2000), local MDS (Chen & Buja, 2009) and so on. MDS has also been used for graph drawing(Gansner et al., 2004)by producing node embeddings using shortest path distances on the graph. Our approach can be seen as an important extension of MDS to work with multiple datasets. | Published as a conference paper at ICLR 2023 UNSUPERVISED MANIFOLD ALIGNMENT WITH JOINT MULTIDIMENSIONAL SCALING |
d252683179 | Predicting the responses of a cell under perturbations may bring important benefits to drug discovery and personalized therapeutics. In this work, we propose a novel graph variational Bayesian causal inference framework to predict a cell's gene expressions under counterfactual perturbations (perturbations that this cell did not factually receive), leveraging information representing biological knowledge in the form of gene regulatory networks (GRNs) to aid individualized cellular response predictions. Aiming at a data-adaptive GRN, we also developed an adjacency matrix updating technique for graph convolutional networks and used it to refine GRNs during pre-training, which generated more insights on gene relations and enhanced model performance. Additionally, we propose a robust estimator within our framework for the asymptotically efficient estimation of marginal perturbation effect, which is yet to be carried out in previous works. With extensive experiments, we exhibited the advantage of our approach over state-of-the-art deep learning models for individual response prediction. | Published as a conference paper at ICLR 2023 PREDICTING CELLULAR RESPONSES WITH VARIATIONAL CAUSAL INFERENCE AND REFINED RELATIONAL INFORMATION |
d249605546 | Estimating the distance of objects is a safety-critical task for autonomous driving. Focusing on short-range objects, existing methods and datasets neglect the equally important long-range objects. In this paper, we introduce a challenging and underexplored task, which we refer to as Long-Range Distance Estimation, as well as two datasets to validate new methods developed for this task. We then propose R4D, the first framework to accurately estimate the distance of long-range objects by using references with known distances in the scene. Drawing inspiration from human perception, R4D builds a graph by connecting a target object to all references. An edge in the graph encodes the relative distance information between a pair of target and reference objects. An attention module is then used to weigh the importance of reference objects and combine them into one target object distance prediction. Experiments on the two proposed datasets demonstrate the effectiveness and robustness of R4D by showing significant improvements compared to existing baselines. We're looking to make the proposed dataset, Waymo Open Dataset -Long-Range Labels, available publicly, at waymo.com/open/download. * Work done while at Waymo LLC. †Equal contribution. | R4D: UTILIZING REFERENCE OBJECTS FOR LONG- RANGE DISTANCE ESTIMATION |
d231603232 | Deep ensembles perform better than a single network thanks to the diversity among their members. Recent approaches regularize predictions to increase diversity; however, they also drastically decrease individual members' performances. In this paper, we argue that learning strategies for deep ensembles need to tackle the trade-off between ensemble diversity and individual accuracies. Motivated by arguments from information theory and leveraging recent advances in neural estimation of conditional mutual information, we introduce a novel training criterion called DICE: it increases diversity by reducing spurious correlations among features. The main idea is that features extracted from pairs of members should only share information useful for target class prediction without being conditionally redundant. Therefore, besides the classification loss with information bottleneck, we adversarially prevent features from being conditionally predictable from each other. We manage to reduce simultaneous errors while protecting class information. We obtain state-of-the-art accuracy results on CIFAR-10/100: for example, an ensemble of 5 networks trained with DICE matches an ensemble of 7 networks trained independently. We further analyze the consequences on calibration, uncertainty estimation, out-of-distribution detection and online co-distillation. | DICE: DIVERSITY IN DEEP ENSEMBLES VIA CONDI- TIONAL REDUNDANCY ADVERSARIAL ESTIMATION |
d214743496 | Model selection when designing deep learning systems for specific use-cases can be a challenging task as many options exist and it can be difficult to know the trade-off between them. Therefore, we investigate a number of state of the art CNN models for the task of measuring kernel fragmentation in harvested corn silage. The models are evaluated across a number of feature extractors and image sizes in order to determine optimal model design choices based upon the tradeoff between model complexity, accuracy and speed. We show that accuracy improvements can be made with more complex meta-architectures and speed can be optimised by decreasing the image size with only slight losses in accuracy. Additionally, we show improvements in Average Precision at an Intersection over Union of 0.5 of up to 20 percentage points while also decreasing inference time in comparison to previously published work. This result for better model selection enables opportunities for creating systems that can aid farmers in improving their silage quality while harvesting. | EVALUATION OF MODEL SELECTION FOR KERNEL FRAGMENT RECOGNITION IN CORN SILAGE |
d219708387 | While theoretically appealing, the application of the Wasserstein distance to large-scale machine learning problems has been hampered by its prohibitive computational cost. The sliced Wasserstein distance and its variants improve the computational efficiency through the random projection, yet they suffer from low accuracy if the number of projections is not sufficiently large, because the majority of projections result in trivially small values. In this work, we propose a new family of distance metrics, called augmented sliced Wasserstein distances (ASWDs), constructed by first mapping samples to higher-dimensional hypersurfaces parameterized by neural networks. It is derived from a key observation that (random) linear projections of samples residing on these hypersurfaces would translate to much more flexible nonlinear projections in the original sample space, so they can capture complex structures of the data distribution. We show that the hypersurfaces can be optimized by gradient ascent efficiently. We provide the condition under which the ASWD is a valid metric and show that this can be obtained by an injective neural network architecture. Numerical results demonstrate that the ASWD significantly outperforms other Wasserstein variants for both synthetic and real-world problems.An alternative approach is to approximate the Wasserstein distance through slicing, i.e. linearly projecting, the distributions to be compared. The sliced Wasserstein distance (SWD) (Bonneel et al., 2015) is defined as the expected value of Wasserstein distances between one-dimensional random projections of high-dimensional distributions. The SWD shares similar theoretical properties with the Wasserstein distance (Bonnotte, 2013) and is computationally efficient since the Wasserstein distance in one-dimensional space has a closed-form solution based on sorting. (Deshpande et al., 2019) extends the sliced Wasserstein distance to the max-sliced Wasserstein distance (Max-SWD), by finding a single projection direction with the maximal distance between projected samples. The subspace robust Wasserstein distance extends the idea of slicing to projecting distributions on linear subspaces (Paty and 1 arXiv:2006.08812v7 [cs.LG] Figure 1: (a) and (b) are visualizations of projections for the ASWD and the SWD between two 2-dimensional Gaussians. (c) and (d) are distance histograms for the ASWD and the SWD between two 100-dimensional Gaussians.Figure 1(a)shows that the injective neural network embedded in the ASWD learns data patterns (in the X-Y plane) and produces well-separate projected values (Z-axis) between distributions in a random projection direction. The high projection efficiency of the ASWD is evident inFigure 1(c), as almost all random projection directions in a 100-dimensional space lead to significant distances between 1-dimensional projections. In contrast, random linear mappings in the SWD often produce closer 1-d projections (Z-axis)(Figure 1(b)); as a result, a large percentage of random projection directions in the 100-d space result in trivially small distances(Figure 1(d)), leading to a low projection efficiency in high-dimensional spaces. Cuturi, 2019). However, the linear nature of these projections usually leads to low projection efficiency of the resulted metrics in high-dimensional spaces (Deshpande et al., 2019; Kolouri et al., 2019a).Different variants of the SWD have been proposed to improve the projection efficiency of the SWD, either by introducing nonlinear projections or by optimizing the distribution of random projections. Specifically, (Kolouri et al., 2019a) extends the connection between the sliced Wasserstein distance and the Radon transform (Radon, 1917) to introduce generalized sliced Wasserstein distances (GSWDs) by utilizing generalized Radon transforms (GRTs), which are defined by nonlinear defining functions and lead to nonlinear projections. A variant named the GSWD-NN was proposed in (Kolouri et al., 2019a) to generate nonlinear projections directly with neural network outputs, but it does not fit into the theoretical framework of the GSWD and does not guarantee a valid metric. In contrast, the distributional sliced Wasserstein distance (DSWD) and its nonlinear version, the distributional generalized sliced Wasserstein distance (DGSWD), improve their projection efficiency by finding a distribution of projections that maximizes the expected distances over these projections. The GSWD and the DGSWD exhibit higher projection efficiency than the SWD in the experiment evaluation, yet they require the specification of the particular form of defining functions from a limited class of candidates. However, the selection of defining functions is usually a task-dependent problem and requires domain knowledge, and the impact on performance from different defining functions is still unclear.In this paper, we present the augmented sliced Wasserstein distance (ASWD), a distance metric constructed by first mapping samples to hypersurfaces in an augmented space, which enables flexible nonlinear slicing of data distributions for improved projection efficiency (SeeFigure 1). Our main contributions include: (i) We exploit the capacity of nonlinear projections employed in the ASWD by constructing injective mapping with arbitrary neural networks; (ii) We prove that the ASWD is a valid distance metric; (iii) We provide a mechanism in which the hypersurface where high-dimensional distributions are projected onto can be optimized and show that the optimization of hypersurfaces can help improve the projection efficiency of slice-based Wasserstein distances. Hence, the ASWD is data-adaptive, i.e. the hypersurfaces can be learned from data. This implies one does not need to manually design a function from the limited class of candidates; (iv) We demonstrate superior performance of the ASWD in numerical experiments for both synthetic and real-world datasets. | Published as a conference paper at ICLR 2022 AUGMENTED SLICED WASSERSTEIN DISTANCES |
d231662243 | Graph neural networks (GNNs) are powerful models that have been successful in various graph representation learning tasks. Whereas gradient boosted decision trees (GBDT) often outperform other machine learning methods when faced with heterogeneous tabular data. But what approach should be used for graphs with tabular node features? Previous GNN models have mostly focused on networks with homogeneous sparse features and, as we show, are suboptimal in the heterogeneous setting. In this work, we propose a novel architecture that trains GBDT and GNN jointly to get the best of both worlds: the GBDT model deals with heterogeneous features, while GNN accounts for the graph structure. Our model benefits from endto-end optimization by allowing new trees to fit the gradient updates of GNN. With an extensive experimental comparison to the leading GBDT and GNN models, we demonstrate a significant increase in performance on a variety of graphs with tabular features. The code is available: https://github.com/nd7141/bgnn. . Attention-based graph neural network for semi-supervised learning. arXiv preprint arXiv:1803.03735, 2018. | Published as a conference paper at ICLR 2021 BOOST THEN CONVOLVE: GRADIENT BOOSTING MEETS GRAPH NEURAL NETWORKS |
d246634193 | We introduce Corrupted Image Modeling (CIM) for self-supervised visual pretraining. CIM uses an auxiliary generator with a small trainable BEiT (Bao et al., 2021) to corrupt the input image instead of using artificial [MASK] tokens, where some patches are randomly selected and replaced with plausible alternatives sampled from the BEiT output distribution. Given this corrupted image, an enhancer network learns to either recover all the original image pixels, or predict whether each visual token is replaced by a generator sample or not. The generator and the enhancer are simultaneously trained and synergistically updated. After pre-training, the enhancer can be used as a high-capacity visual encoder for downstream tasks. CIM is a general and flexible visual pre-training framework that is suitable for various network architectures. For the first time, CIM demonstrates that both ViT and CNN can learn rich visual representations using a unified, non-Siamese framework. Experimental results show that our approach achieves compelling results in vision benchmarks, such as ImageNet classification and ADE20K semantic segmentation. | Published as a conference paper at ICLR 2023 CORRUPTED IMAGE MODELING FOR SELF-SUPERVISED VISUAL PRE-TRAINING |
d195346934 | We consider the problem of representing a large population's behavior policy that drives the evolution of the population distribution over a discrete state space. A discrete time mean field game (MFG) is motivated as an interpretable model founded on game theory for understanding the aggregate effect of individual actions and predicting the temporal evolution of population distributions. We achieve a synthesis of MFG and Markov decision processes (MDP) by showing that a special MFG is reducible to an MDP. This enables us to broaden the scope of mean field game theory and infer MFG models of large real-world systems via deep inverse reinforcement learning. Our method learns both the reward function and forward dynamics of an MFG from real data, and we report the first empirical test of a mean field game model of a real-world social media population. | Deep Mean Field Games for Learning Optimal Behavior Policy of Large Populations |
d3461154 | Exploration is a fundamental aspect of Reinforcement Learning, typically implemented using stochastic action-selection. Exploration, however, can be more efficient if directed toward gaining new world knowledge. Visit-counters have been proven useful both in practice and in theory for directed exploration. However, a major limitation of counters is their locality. While there are a few model-based solutions to this shortcoming, a model-free approach is still missing. We propose E-values, a generalization of counters that can be used to evaluate the propagating exploratory value over state-action trajectories. We compare our approach to commonly used RL techniques, and show that using E-values improves learning and performance over traditional counters. We also show how our method can be implemented with function approximation to efficiently learn continuous MDPs. We demonstrate this by showing that our approach surpasses state of the art performance in the Freeway Atari 2600 game. | Published as a conference paper at ICLR 2018 DORA THE EXPLORER: DIRECTED OUTREACHING REINFORCEMENT ACTION-SELECTION |
d247519194 | Language models typically need to be trained or finetuned in order to acquire new knowledge, which involves updating their weights. We instead envision language models that can simply read and memorize new data at inference time, thus acquiring new knowledge immediately. In this work, we extend language models with the ability to memorize the internal representations of past inputs. We demonstrate that an approximate kNN lookup into a non-differentiable memory of recent (key, value) pairs improves language modeling across various benchmarks and tasks, including generic webtext (C4), math papers (arXiv), books (PG-19), code (Github), as well as formal theorems (Isabelle). We show that the performance steadily improves when we increase the size of memory up to 262K tokens. On benchmarks including code and mathematics, we find that the model is capable of making use of newly defined functions and theorems during test time. | Published as a conference paper at ICLR 2022 MEMORIZING TRANSFORMERS |
d210932569 | Recent deep generative models are able to provide photo-realistic images as well as visual or textual content embeddings useful to address various tasks of computer vision and natural language processing. Their usefulness is nevertheless often limited by the lack of control over the generative process or the poor understanding of the learned representation. To overcome these major issues, very recent work has shown the interest of studying the semantics of the latent space of generative models. In this paper, we propose to advance on the interpretability of the latent space of generative models by introducing a new method to find meaningful directions in the latent space of any generative model along which we can move to control precisely specific properties of the generated image like the position or scale of the object in the image. Our method does not require human annotations and is particularly well suited for the search of directions encoding simple transformations of the generated image, such as translation, zoom or color variations. We demonstrate the effectiveness of our method qualitatively and quantitatively, both for GANs and variational auto-encoders. Figure 1: Images generated with our approach and a BigGAN model (Brock et al., 2018), showing that the position of the object can be controlled within the image. | Published as a conference paper at ICLR 2020 CONTROLLING GENERATIVE MODELS WITH CONTINU- OUS FACTORS OF VARIATIONS |
d235358191 | While recent work has shown that scores from models trained by the ubiquitous masked language modeling (MLM) objective effectively discriminate probable from improbable sequences, it is still an open question if these MLMs specify a principled probability distribution over the space of possible sequences. In this paper, we interpret MLMs as energy-based sequence models and propose two energy parametrizations derivable from the trained MLMs. In order to draw samples correctly from these models, we develop a tractable sampling scheme based on the Metropolis-Hastings Monte Carlo algorithm. In our approach, samples are proposed from the same masked conditionals used for training the masked language models, and they are accepted or rejected based on their energy values according to the target distribution. We validate the effectiveness of the proposed parametrizations by exploring the quality of samples drawn from these energybased models for both open-ended unconditional generation and a conditional generation task of machine translation. We theoretically and empirically justify our sampling algorithm by showing that the masked conditionals on their own do not yield a Markov chain whose stationary distribution is that of our target distribution, and our approach generates higher quality samples than other recently proposed undirected generation approachesGhazvininejad et al., 2019). | Published as a conference paper at ICLR 2022 EXPOSING THE IMPLICIT ENERGY NETWORKS BEHIND MASKED LANGUAGE MODELS VIA METROPOLIS-HASTINGS |
d253255129 | Today's computer vision models achieve human or near-human level performance across a wide variety of vision tasks. However, their architectures, data, and learning algorithms differ in numerous ways from those that give rise to human vision. In this paper, we investigate the factors that affect the alignment between the representations learned by neural networks and human mental representations inferred from behavioral responses. We find that model scale and architecture have essentially no effect on the alignment with human behavioral responses, whereas the training dataset and objective function both have a much larger impact. These findings are consistent across three datasets of human similarity judgments collected using two different tasks. Linear transformations of neural network representations learned from behavioral responses from one dataset substantially improve alignment with human similarity judgments on the other two datasets. In addition, we find that some human concepts such as food and animals are well-represented by neural networks whereas others such as royal or sports-related objects are not. Overall, although models trained on larger, more diverse datasets achieve better alignment with humans than models trained on ImageNet alone, our results indicate that scaling alone is unlikely to be sufficient to train neural networks with conceptual representations that match those used by humans. | HUMAN ALIGNMENT OF NEURAL NETWORK REPRE- SENTATIONS |
d8606632 | Precisely-labeled data sets with sufficient amount of samples are very important for training deep convolutional neural networks (CNNs). However, many of the available real-world data sets contain erroneously labeled samples and those errors substantially hinder the learning of very accurate CNN models. In this work, we consider the problem of training a deep CNN model for image classification with mislabeled training samples -an issue that is common in real image data sets with tags supplied by amateur users. To solve this problem, we propose an auxiliary image regularization technique, optimized by the stochastic Alternating Direction Method of Multipliers (ADMM) algorithm, that automatically exploits the mutual context information among training images and encourages the model to select reliable images to robustify the learning process. Comprehensive experiments on benchmark data sets clearly demonstrate our proposed regularized CNN model is resistant to label noise in training data. | Published as a conference paper at ICLR 2016 AUXILIARY IMAGE REGULARIZATION FOR DEEP CNNS WITH NOISY LABELS |
d13576081 | Neural networks with low-precision weights and activations offer compelling efficiency advantages over their full-precision equivalents. The two most frequently discussed benefits of quantization are reduced memory consumption, and a faster forward pass when implemented with efficient bitwise operations. We propose a third benefit of very low-precision neural networks: improved robustness against some adversarial attacks, and in the worst case, performance that is on par with full-precision models. We focus on the very low-precision case where weights and activations are both quantized to ±1, and note that stochastically quantizing weights in just one layer can sharply reduce the impact of iterative attacks. We observe that non-scaled binary neural networks exhibit a similar effect to the original defensive distillation procedure that led to gradient masking, and a false notion of security. We address this by conducting both black-box and white-box experiments with binary models that do not artificially mask gradients. 1 | Published as a conference paper at ICLR 2018 ATTACKING BINARIZED NEURAL NETWORKS |
d252383681 | Inspired by humans' exceptional ability to master arithmetic and generalize to new problems, we present a new dataset, Handwritten arithmetic with INTegers (HINT), to examine machines' capability of learning generalizable concepts at three levels: perception, syntax, and semantics. In HINT, machines are tasked with learning how concepts are perceived from raw signals such as images (i.e., perception), how multiple concepts are structurally combined to form a valid expression (i.e., syntax), and how concepts are realized to afford various reasoning tasks (i.e., semantics), all in a weakly supervised manner. Focusing on systematic generalization, we carefully design a five-fold test set to evaluate both the interpolation and the extrapolation of learned concepts w.r.t. the three levels. Further, we design a few-shot learning split to determine whether or not models can rapidly learn new concepts and generalize them to more complex scenarios. To comprehend existing models' limitations, we undertake extensive experiments with various sequence-to-sequence models, including RNNs, Transformers, and GPT-3 (with the chain of thought prompting).The results indicate that current models struggle to extrapolate to long-range syntactic dependency and semantics. Models exhibit a considerable gap toward human-level generalization when evaluated with new concepts in a few-shot setting. Moreover, we discover that it is infeasible to solve HINT by merely scaling up the dataset and the model size; this strategy contributes little to the extrapolation of syntax and semantics. Finally, in zero-shot GPT-3 experiments, the chain of thought prompting exhibits impressive results and significantly boosts the test accuracy. We believe the HINT dataset and the experimental findings are of great interest to the learning community on systematic generalization. | A MINIMALIST DATASET FOR SYSTEMATIC GENERAL- IZATION OF PERCEPTION, SYNTAX, AND SEMANTICS |
d256616076 | Previous studies have shown that leveraging domain index can significantly boost domain adaptation performance (Wang et al., 2020; Xu et al., 2022). However, such domain indices are not always available. To address this challenge, we first provide a formal definition of domain index from the probabilistic perspective, and then propose an adversarial variational Bayesian framework that infers domain indices from multi-domain data, thereby providing additional insight on domain relations and improving domain adaptation performance. Our theoretical analysis shows that our adversarial variational Bayesian framework finds the optimal domain index at equilibrium. Empirical results on both synthetic and real data verify that our model can produce interpretable domain indices which enable us to achieve superior performance compared to state-of-the-art domain adaptation methods. Code is available at https://github.com/Wang-ML-Lab/VDI. . Sentry: Selective entropy optimization via committee consistency for unsupervised domain adaptation. In . Bridging the day and night domain gap for semantic segmentation. In 2019 IEEE Intelligent Vehicles Symposium (IV), pp. 1312-1318. IEEE, 2019.Yossi Rubner, Carlo Tomasi, and Leonidas J Guibas. The earth mover's distance as a metric for image retrieval. source domain adaptation for semantic segmentation of very high resolution satellite images by data standardization. In | DOMAIN-INDEXING VARIATIONAL BAYES: INTER- PRETABLE DOMAIN INDEX FOR DOMAIN ADAPTATION |
d209439505 | Knowledge graph embedding research has overlooked the problem of probability calibration. We show popular embedding models are indeed uncalibrated. That means probability estimates associated to predicted triples are unreliable. We present a novel method to calibrate a model when ground truth negatives are not available, which is the usual case in knowledge graphs. We propose to use Platt scaling and isotonic regression alongside our method. Experiments on three datasets with ground truth negatives show our contribution leads to well calibrated models when compared to the gold standard of using negatives. We get significantly better results than the uncalibrated models from all calibration methods. We show isotonic regression offers the best the performance overall, not without trade-offs. We also show that calibrated models reach state-of-the-art accuracy without the need to define relation-specific decision thresholds.INTRODUCTIONKnowledge graph embedding models are neural architectures that learn vector representations (i.e. embeddings) of nodes and edges of a knowledge graph. Such knowledge graph embeddings have applications in knowledge graph completion, knowledge discovery, entity resolution, and link-based clustering, just to cite a few(Nickel et al., 2016a).Despite burgeoning research, the problem of calibrating such models has been overlooked, and existing knowledge graph embedding models do not offer any guarantee on the probability estimates they assign to predicted facts. Probability calibration is important whenever you need the predictions to make probabilistic sense, i.e., if the model predicts a fact is true with 80% confidence, it should to be correct 80% of the times. Prior art suggests to use a sigmoid layer to turn logits returned by models into probabilities (Nickel et al., 2016a) (also called the expit transform), but we show that this provides poor calibration.Figure 1shows reliability diagrams for off-the-shelf TransE and ComplEx. The identity function represents perfect calibration. Both models are miscalibrated: all TransE combinations inFigure 1aunder-forecast the probabilities (i.e. probabilities are too small), whereas ComplEx under-forecasts or over-forecasts according to which loss is used(Figure1b).Calibration is crucial in high-stakes scenarios such as drug-target discovery from biological networks, where end-users need trustworthy and interpretable decisions. Moreover, since probabilities are not calibrated, when classifying triples (i.e. facts) as true or false, users must define relationspecific thresholds, which can be awkward for graphs with a great number of relation types.To the best of our knowledge, this is the first work to focus on calibration for knowledge embeddings. Our contribution is two-fold: First, we use Platt Scaling and isotonic regression to calibrate knowledge graph embedding models on datasets that include ground truth negatives. One peculiar feature of knowledge graphs is that they usually rely on the open world assumption (facts not present are not necessarily false, they are simply unknown). This makes calibration troublesome because of the lack of ground truth negatives. For this reason, our second and main contribution is a calibration heuristics that combines Platt-scaling or isotonic regression with synthetically generated negatives.Experimental results show that we obtain better-calibrated models and that it is possible to calibrate knowledge graph embedding models even when ground truth negatives are not present. We | Published as a conference paper at ICLR 2020 PROBABILITY CALIBRATION FOR KNOWLEDGE GRAPH EMBEDDING MODELS |
d258823285 | A Markov network characterizes the conditional independence structure, or Markov property, among a set of random variables. Existing work focuses on specific families of distributions (e.g., exponential families) and/or certain structures of graphs, and most of them can only handle variables of a single data type (continuous or discrete). In this work, we characterize the conditional independence structure in general distributions for all data types (i.e., continuous, discrete, and mixed-type) with a Generalized Precision Matrix (GPM). Besides, we also allow general functional relations among variables, thus giving rise to a Markov network structure learning algorithm in one of the most general settings. To deal with the computational challenge of the problem, especially for large graphs, we unify all cases under the same umbrella of a regularized score matching framework. We validate the theoretical results and demonstrate the scalability empirically in various settings.While non-Gaussianity is more common in real-world data generating process, few results are applicable to Markov network structure learning on non-Gaussian data. In the discrete setting, Ravikumar et al. (2010) showed that a binary Ising model can be recovered by neighborhood selection using 1 penalized logistic regression. Loh & Wainwright (2013) encoded extra structural relations in the proposed generalized covariance matrix to model the dependencies for Markov networks with certain structures (e.g., tree structures or graphs with only singleton separator sets) among variables from exponential families. Several approaches allowed estimation for non-Gaussian continuous variables while most of them assumed parametric assumptions such as the exponential families (Yang et al., 2015; Lin et al., 2016; Suggala et al., 2017) or Gaussian copulas (Liu et al., 2009; 2012; Harris & Drton, 2013). These methods illustrate the possibility of reliable Markov network estimations in several non-Gaussian cases, but still, the models are restricted to specific parametric families of distributions and/or structures of conditional independencies. . Towards an integrated proteinprotein interaction network: A relational markov network approach. . Estimation of high-dimensional graphical models using regularized score matching. . Transelliptical graphical models. Advances in neural information processing systems, 25, 2012.Po-LingLoh and Peter Bühlmann. High-dimensional learning of linear causal networks via inverse covariance estimation. Journal of Machine Learning Research, 15(88):3065-3105, 2014. Po-Ling Loh and Martin J Wainwright. Structure estimation for discrete graphical models: Generalized covariance matrices and their inverses. Advances in Neural Information Processing Systems, 25, 2012. Po-Ling Loh and Martin J. Wainwright. Structure estimation for discrete graphical models: Generalized covariance matrices and their inverses. The Annals of Statistics, 41(6):3022-3049, 2013. Po-Ling Loh and Martin J. Wainwright. Support recovery without incoherence: A case for nonconvex regularization. The Annals of Statistics, 45(6):2455-2482, 2017. Siwei Lyu. Interpretation and generalization of score matching. arXiv preprint arXiv:1205.2629, 2012. Nicolai Meinshausen and Peter Bühlmann. High-dimensional graphs and variable selection with the lasso. The annals of statistics, 34(3):1436-1462, 2006.Rebecca Morrison, Ricardo Baptista, and Youssef Marzouk. Beyond normality: Learning sparse probabilistic graphical models in the non-gaussian setting. Advances in neural information processing systems, 30, 2017. . Blind separation of mixture of independent sources through a quasi-maximum likelihood approach. . Model selection in gaussian graphical models: High-dimensional consistency of 1 -regularized mle. . Model selection in Gaussian graphical models: High-dimensional consistency of ell 1 -regularized MLE. -dimensional ising model selection using 1 -regularized logistic regression. | Published as a conference paper at ICLR 2023 GENERALIZED PRECISION MATRIX FOR SCALABLE ESTIMATION OF NONPARAMETRIC MARKOV NET- WORKS |
d159298330 | resolution of past human rights violations through extra-judicial organizations is an advanced step towards resolving the case, whereas a conflict approach can be used to settle the case.The existence of the Human Rights Law provides a new frontier in implementing the principle of restorative justice in the approach of case settlement.It is hoped that such restorative justice can create a political balance between the past and the future. | Human Rights Court and Truth Reconciliation Commission for the Settlement of Human Rights in Indonesia |
d238583582 | Real world tournaments are almost always intransitive. Recent works have noted that parametric models which assume d dimensional node representations can effectively model intransitive tournaments. However, nothing is known about the structure of the class of tournaments that arise out of any fixed d dimensional representations. In this work, we develop a novel theory for understanding parametric tournament representations. Our first contribution is to structurally characterize the class of tournaments that arise out of d dimensional representations. We do this by showing that these tournament classes have forbidden configurations which must necessarily be union of flip classes, a novel way to partition the set of all tournaments. We further characterize rank 2 tournaments completely by showing that the associated forbidden flip class contains just 2 tournaments. Specifically, we show that the rank 2 tournaments are equivalent to locally-transitive tournaments. This insight allows us to show that the minimum feedback arc set problem on this tournament class can be solved using the standard Quicksort procedure. For a general rank d tournament class, we show that the flip class associated with a coned-doubly regular tournament of size O( √ d) must be a forbidden configuration. To answer a dual question, using a celebrated result of [10], we show a lower bound of O( √ n) on the minimum dimension needed to represent all tournaments on n nodes. For any given tournament, we show a novel upper bound on the smallest representation dimension that depends on the least size of the number of unique nodes in any feedback arc set of the flip class associated with a tournament. We show how our results also shed light on upper bound of sign-rank of matrices. | A THEORY OF TOURNAMENT REPRESENTATIONS A PREPRINT |
d90262267 | We propose a neural network for unsupervised anomaly detection with a novel robust subspace recovery layer (RSR layer). This layer seeks to extract the underlying subspace from a latent representation of the given data and removes outliers that lie away from this subspace. It is used within an autoencoder. The encoder maps the data into a latent space, from which the RSR layer extracts the subspace. The decoder then smoothly maps back the underlying subspace to a "manifold" close to the original inliers. Inliers and outliers are distinguished according to the distances between the original and mapped positions (small for inliers and large for outliers). Extensive numerical experiments with both image and document datasets demonstrate state-of-the-art precision and recall. * Equal contribution. | Published as a conference paper at ICLR 2020 ROBUST SUBSPACE RECOVERY LAYER FOR UNSUPERVISED ANOMALY DETECTION |
d48519159 | Many hallmarks of human intelligence, such as generalizing from limited experience, abstract reasoning and planning, analogical reasoning, creative problem solving, and capacity for language require the ability to consolidate experience into concepts, which act as basic building blocks of understanding and reasoning. We present a framework that defines a concept by an energy function over events in the environment, as well as an attention mask over entities participating in the event. Given few demonstration events, our method uses inference-time optimization procedure to generate events involving similar concepts or identify entities involved in the concept. We evaluate our framework on learning visual, quantitative, relational, temporal concepts from demonstration events in an unsupervised manner. Our approach is able to successfully generate and identify concepts in a few-shot setting and resulting learned concepts can be reused across environments. Example videos of our results are available at sites.google.com/site/energyconceptmodels | Concept Learning with Energy-Based Models |
d254070104 | In an era of countless content offerings, recommender systems alleviate information overload by providing users with personalized content suggestions. Due to the scarcity of explicit user feedback, modern recommender systems typically optimize for the same fixed combination of implicit feedback signals across all users. However, this approach disregards a growing body of work highlighting that (i) implicit signals can be used by users in diverse ways, signaling anything from satisfaction to active dislike, and (ii) different users communicate preferences in different ways. We propose applying the recent Interaction Grounded Learning (IGL) paradigm to address the challenge of learning representations of diverse user communication modalities. Rather than requiring a fixed, human-designed reward function, IGL is able to learn personalized reward functions for different users and then optimize directly for the latent user satisfaction. We demonstrate the success of IGL with experiments using simulations as well as with real-world production traces. | Published as a conference paper at ICLR 2023 PERSONALIZED REWARD LEARNING WITH INTERACTION-GROUNDED LEARNING (IGL) |
d17225395 | Hessian-free (HF) optimization has been successfully used for training deep autoencoders and recurrent networks. HF uses the conjugate gradient algorithm to construct update directions through curvature-vector products that can be computed on the same order of time as gradients. In this paper we exploit this property and study stochastic HF with gradient and curvature mini-batches independent of the dataset size. We modify Martens' HF for these settings and integrate dropout, a method for preventing co-adaptation of feature detectors, to guard against overfitting. Stochastic Hessian-free optimization gives an intermediary between SGD and HF that achieves competitive performance on both classification and deep autoencoder experiments. | Training Neural Networks with Stochastic Hessian-Free Optimization |
d8990548 | Recurrent Neural Networks (RNNs) are powerful tools for solving sequence-based problems, but their efficacy and execution time are dependent on the size of the network. Following recent work in simplifying these networks with model pruning and a novel mapping of work onto GPUs, we design an efficient implementation for sparse RNNs. We investigate several optimizations and tradeoffs: Lamport timestamps, wide memory loads, and a bank-aware weight layout. With these optimizations, we achieve speedups of over 6× over the next best algorithm for a hidden layer of size 2304, batch size of 4, and a density of 30%. Further, our technique allows for models of over 5× the size to fit on a GPU for a speedup of 2×, enabling larger networks to help advance the state-of-the-art. We perform case studies on NMT and speech recognition tasks in the appendix, accelerating their recurrent layers by up to 3×. | Published as a conference paper at ICLR 2018 SPARSE PERSISTENT RNNS: SQUEEZING LARGE RECURRENT NETWORKS ON- CHIP |
d257771340 | Molecular representation learning plays a crucial role in AI-assisted drug discovery research. Encoding 3D molecular structures through Euclidean neural networks has become the prevailing method in the geometric deep learning community. However, the equivariance constraints and message passing in Euclidean space may limit the network expressive power. In this work, we propose a Harmonic Molecular Representation learning (HMR) framework, which represents a molecule using the Laplace-Beltrami eigenfunctions of its molecular surface. HMR offers a multi-resolution representation of molecular geometric and chemical features on 2D Riemannian manifold. We also introduce a harmonic message passing method to realize efficient spectral message passing over the surface manifold for better molecular encoding. Our proposed method shows comparable predictive power to current models in small molecule property prediction, and outperforms the state-of-the-art deep learning models for ligand-binding protein pocket classification and the rigid protein docking challenge, demonstrating its versatility in molecular representation learning. | Published as a conference paper at ICLR 2023 LEARNING HARMONIC MOLECULAR REPRESENTA- TIONS ON RIEMANNIAN MANIFOLD |
d250334344 | Interpretable machine learning offers insights into what factors drive a certain prediction of a black-box system. A large number of interpreting methods focus on identifying explanatory input features, which generally fall into two main categories: attribution and selection. A popular attribution-based approach is to exploit local neighborhoods for learning instance-specific explainers in an additive manner. The process is thus inefficient and susceptible to poorly-conditioned samples. Meanwhile, many selection-based methods directly optimize local feature distributions in an instance-wise training framework, thereby being capable of leveraging global information from other inputs. However, they can only interpret single-class predictions and many suffer from inconsistency across different settings, due to a strict reliance on a pre-defined number of features selected. This work exploits the strengths of both methods and proposes a framework for learning local explanations simultaneously for multiple target classes. Our model explainer significantly outperforms additive and instance-wise counterparts on faithfulness with more compact and comprehensible explanations. We also demonstrate the capacity to select stable and important features through extensive experiments on various data sets and black-box model architectures. | Published as a conference paper at ICLR 2023 AN ADDITIVE INSTANCE-WISE APPROACH TO MULTI- CLASS MODEL INTERPRETATION |
d208921206 | Recently, progress has been made in the application of neural networks to the numerical analysis of partial differential equations (PDEs) (cf.[Weinan et al., 2017],[Weinan and Yu, 2018]). In the latter the variational formulation of the Poisson problem is used in order to obtain an objective function a regularised Dirichlet energy that was used for the optimisation of some neural networks. Although this approach showed good visual performance and promising empirical results it is lacking any convergence guarantees. In this notes we use the notion of Γ-convergence to show that ReLU networks of growing architecture that are trained with respect to suitably regularised Dirichlet energies converge to the true solution of the Poisson problem. We discuss how this approach generalises to arbitrary variational problems under certain universality assumptions of neural networks and see that this covers some nonlinear stationary PDEs like the p-Laplace. | DEEP RITZ REVISITED A PREPRINT |
d238419652 | Plug-and-Play (PnP) methods constitute a class of iterative algorithms for imaging problems where regularization is performed by an off-the-shelf denoiser. Although PnP methods can lead to tremendous visual performance for various image problems, the few existing convergence guarantees are based on unrealistic (or suboptimal) hypotheses on the denoiser, or limited to strongly convex data-fidelity terms. We propose a new type of PnP method, based on half-quadratic splitting, for which the denoiser is realized as a gradient descent step on a functional parameterized by a deep neural network. Exploiting convergence results for proximal gradient descent algorithms in the nonconvex setting, we show that the proposed PnP algorithm is a convergent iterative scheme that targets stationary points of an explicit global functional. Besides, experiments show that it is possible to learn such a deep denoiser while not compromising the performance in comparison to other state-of-the-art deep denoisers used in PnP schemes. We apply our proximal gradient algorithm to various ill-posed inverse problems, e.g. deblurring, superresolution and inpainting. For all these applications, numerical results empirically confirm the convergence results. Experiments also show that this new algorithm reaches state-of-the-art performance, both quantitatively and qualitatively. . Beyond a gaussian denoiser: Residual learning of deep cnn for image denoising. . Ffdnet: Toward a fast and flexible solution for cnnbased image denoising. | Published as a conference paper at ICLR 2022 GRADIENT STEP DENOISER FOR CONVERGENT PLUG- AND-PLAY |
d3199842 | We develop a scalable deep non-parametric generative model by augmenting deep Gaussian processes with a recognition model. Inference is performed in a novel scalable variational framework where the variational posterior distributions are reparametrized through a multilayer perceptron. The key aspect of this reformulation is that it prevents the proliferation of variational parameters which otherwise grow linearly in proportion to the sample size. We derive a new formulation of the variational lower bound that allows us to distribute most of the computation in a way that enables to handle datasets of the size of mainstream deep learning tasks. We show the efficacy of the method on a variety of challenges including deep unsupervised learning and deep Bayesian optimization. | VARIATIONAL AUTO-ENCODED DEEP GAUSSIAN PRO- CESSES |
d2135897 | While most approaches to automatically recognizing entailment relations have used classifiers employing hand engineered features derived from complex natural language processing pipelines, in practice their performance has been only slightly better than bag-of-word pair classifiers using only lexical similarity. The only attempt so far to build an end-to-end differentiable neural network for entailment failed to outperform such a simple similarity classifier. In this paper, we propose a neural model that reads two sentences to determine entailment using long short-term memory units. We extend this model with a word-by-word neural attention mechanism that encourages reasoning over entailments of pairs of words and phrases. Furthermore, we present a qualitative analysis of attention weights produced by this model, demonstrating such reasoning capabilities. On a large entailment dataset this model outperforms the previous best neural model and a classifier with engineered features by a substantial margin. It is the first generic end-to-end differentiable system that achieves state-of-the-art accuracy on a textual entailment dataset. | Published as a conference paper at ICLR 2016 REASONING ABOUT ENTAILMENT WITH NEURAL ATTENTION |
d223957176 | Convolutional neural networks often dominate fully-connected counterparts in generalization performance, especially on image classification tasks. This is often explained in terms of "better inductive bias." However, this has not been made mathematically rigorous, and the hurdle is that the sufficiently wide fully-connected net can always simulate the convolutional net. Thus the training algorithm plays a role. The current work describes a natural task on which a provable sample complexity gap can be shown, for standard training algorithms. We construct a single natural distribution on R d × {±1} on which any orthogonal-invariant algorithm (i.e. fully-connected networks trained with most gradient-based methods from gaussian initialization) requires Ω(d 2 ) samples to generalize while O(1) samples suffice for convolutional architectures. Furthermore, we demonstrate a single target function, learning which on all possible distributions leads to an O(1) vs Ω(d 2 /ε) gap. The proof relies on the fact that SGD on fully-connected network is orthogonal equivariant. Similar results are achieved for 2 regression and adaptive training algorithms, e.g. Adam and AdaGrad, which are only permutation equivariant.arXiv:2010.08515v2 [cs.LG] 4 May 2021Published as a conference paper at ICLR 2021 10 2 | Published as a conference paper at ICLR 2021 WHY ARE CONVOLUTIONAL NETS MORE SAMPLE- EFFICIENT THAN FULLY-CONNECTED NETS? |
d253117079 | This paper studies learning on text-attributed graphs (TAGs), where each node is associated with a text description. An ideal solution for such a problem would be integrating both the text and graph structure information with large language models and graph neural networks (GNNs). However, the problem becomes very challenging when graphs are large due to the high computational complexity brought by training large language models and GNNs together. In this paper, we propose an efficient and effective solution to learning on large text-attributed graphs by fusing graph structure and language learning with a variational Expectation-Maximization (EM) framework, called GLEM. Instead of simultaneously training large language models and GNNs on big graphs, GLEM proposes to alternatively update the two modules in the E-step and M-step. Such a procedure allows training the two modules separately while simultaneously allowing the two modules to interact and mutually enhance each other. Extensive experiments on multiple data sets demonstrate the efficiency and effectiveness of the proposed approach 1 . | Published as a conference paper at ICLR 2023 LEARNING ON LARGE-SCALE TEXT-ATTRIBUTED GRAPHS VIA VARIATIONAL INFERENCE |
d247762924 | The interventional nature of recommendation has attracted increasing attention in recent years. It particularly motivates researchers to formulate learning and evaluating recommendation as causal inference and data missing-not-at-random problems. However, few take seriously the consequence of violating the critical assumption of overlapping, which we prove can significantly threaten the validity and interpretation of the outcome. We find a critical piece missing in the current understanding of information retrieval (IR) systems: as interventions, recommendation not only affects the already observed data, but it also interferes with the target domain (distribution) of interest. We then rephrase optimizing recommendation as finding an intervention that best transports the patterns it learns from the observed domain to its intervention domain. Towards this end, we use domain transportation to characterize the learning-intervention mechanism of recommendation. We design a principled transportation-constraint risk minimization objective and convert it to a two-player minimax game. We prove the consistency, generalization, and excessive risk bounds for the proposed objective, and elaborate how they compare to the current results. Finally, we carry out extensive real-data and semi-synthetic experiments to demonstrate the advantage of our approach, and launch online testing with a real-world IR system. | Published as a conference paper at ICLR 2022 FROM INTERVENTION TO DOMAIN TRANSPORTATION: A NOVEL PERSPECTIVE TO OPTIMIZE RECOMMENDA- TION |
d254974436 | Figure 1: Comparison of GOOD with different baselines. Images in the first column are from validation sets of ADE20K (Zhou et al., 2019). From the second to fourth columns we show the detection results of three open-world object detection methods: OLN Kim et al. (2021), GGN Wang et al. (2022), and our Geometry-guided Open-world Object Detector (GOOD). The shown detection results are true-positive proposals from the top 100 proposals of each method. The numbers of true positive proposals or ground truth objects are denoted in parentheses. All models are trained on the RGB images from the PASCAL-VOC classes of the COCO dataset (Lin et al., 2014), which do not include houses, trees, or kitchen furniture. Both OLN and GGN fail to detect many objects not seen during training. GOOD generalizes better to unseen categories by exploiting the geometric cues.ABSTRACTWe address the task of open-world class-agnostic object detection, i.e., detecting every object in an image by learning from a limited number of base object classes. State-of-the-art RGB-based models suffer from overfitting the training classes and often fail at detecting novel-looking objects. This is because RGB-based models primarily rely on appearance similarity to detect novel objects and are also prone to overfitting short-cut cues such as textures and discriminative parts. To address these shortcomings of RGB-based object detectors, we propose incorporating geometric cues such as depth and normals, predicted by general-purpose monocular estimators. Specifically, we use the geometric cues to train an object proposal network for pseudo-labeling unannotated novel objects in the training set. Our resulting Geometry-guided Open-world Object Detector (GOOD) significantly improves detection recall for novel object categories and already performs well with only a few training classes. Using a single "person" class for training on the COCO dataset, GOOD surpasses SOTA methods by 5.0% AR@100, a relative improvement of 24%. | Published as a conference paper at ICLR 2023 GOOD: EXPLORING GEOMETRIC CUES FOR DETECT- ING OBJECTS IN AN OPEN WORLD (a) Ground truth (20) (b) OLN (13) (c) GGN (14) (d) GOOD (18) (e) Ground truth (22) (f) OLN (6) (g) GGN (6) (h) GOOD (15) AR@100 VOC to Non-VOC of COCO Base Novel |
d238531375 | Message Passing Neural Networks (MPNNs) are a common type of Graph Neural Network (GNN), in which each node's representation is computed recursively by aggregating representations ("messages") from its immediate neighbors akin to a star-shaped pattern. MPNNs are appealing for being efficient and scalable, however their expressiveness is upper-bounded by the 1st-order Weisfeiler-Leman isomorphism test (1-WL). In response, prior works propose highly expressive models at the cost of scalability and sometimes generalization performance. Our work stands between these two regimes: we introduce a general framework to uplift any MPNN to be more expressive, with limited scalability overhead and greatly improved practical performance. We achieve this by extending local aggregation in MPNNs from star patterns to general subgraph patterns (e.g., k-egonets): in our framework, each node representation is computed as the encoding of a surrounding induced subgraph rather than encoding of immediate neighbors only (i.e. a star). We choose the subgraph encoder to be a GNN (mainly MPNNs, considering scalability) to design a general framework that serves as a wrapper to uplift any GNN. We call our proposed method GNN-AK (GNN As Kernel), as the framework resembles a convolutional neural network by replacing the kernel with GNNs. Theoretically, we show that our framework is strictly more powerful than 1&2-WL, and is not less powerful than 3-WL. We also design subgraph sampling strategies which greatly reduce memory footprint and improve speed while maintaining performance. Our method sets new state-of-the-art performance by large margins for several well-known graph ML tasks; specifically, 0.08 MAE on ZINC, 74.79% and 86.887% accuracy on CIFAR10 and PATTERN respectively. | Published as a conference paper at ICLR 2022 FROM STARS TO SUBGRAPHS: UPLIFTING ANY GNN WITH LOCAL STRUCTURE AWARENESS |
d246485648 | Convolutional neural networks typically contain several downsampling operators, such as strided convolutions or pooling layers, that progressively reduce the resolution of intermediate representations. This provides some shift-invariance while reducing the computational complexity of the whole architecture. A critical hyperparameter of such layers is their stride: the integer factor of downsampling. As strides are not differentiable, finding the best configuration either requires crossvalidation or discrete optimization (e.g. architecture search), which rapidly become prohibitive as the search space grows exponentially with the number of downsampling layers. Hence, exploring this search space by gradient descent would allow finding better configurations at a lower computational cost. This work introduces DiffStride, the first downsampling layer with learnable strides. Our layer learns the size of a cropping mask in the Fourier domain, that effectively performs resizing in a differentiable way. Experiments on audio and image classification show the generality and effectiveness of our solution: we use DiffStride as a drop-in replacement to standard downsampling layers and outperform them. In particular, we show that introducing our layer into a ResNet-18 architecture allows keeping consistent high performance on CIFAR10, CIFAR100 and ImageNet even when training starts from poor random stride configurations. Moreover, formulating strides as learnable variables allows us to introduce a regularization term that controls the computational complexity of the architecture. We show how this regularization allows trading off accuracy for efficiency on ImageNet. | Published as a conference paper at ICLR 2022 LEARNING STRIDES IN CONVOLUTIONAL NEURAL NETWORKS |
d247996737 | In this paper, we propose a new, simplified high probability analysis of AdaGrad for smooth, non-convex problems. More specifically, we focus on a particular accelerated gradient (AGD) template (Lan, 2020), through which we recover the original AdaGrad and its variant with averaging, and prove a convergence rate of O(1/ √ T ) with high probability without the knowledge of smoothness and variance. We use a particular version of Freedman's concentration bound for martingale difference sequences (Kakade & Tewari, 2008) which enables us to achieve the best-known dependence of log(1/δ) on the probability margin δ. We present our analysis in a modular way and obtain a complementary O(1/T ) convergence rate in the deterministic setting. To the best of our knowledge, this is the first high probability result for AdaGrad with a truly adaptive scheme, i.e., completely oblivious to the knowledge of smoothness and uniform variance bound, which simultaneously has best-known dependence of log(1/δ). We further prove noise adaptation property of AdaGrad under additional noise assumptions. * A Viterbi fellow Universality, adaptive methods and acceleration We call an algorithm universal if it achieves optimal rates under different settings, without any modifications. For convex minimization problems, Levy et al. (2018) showed that AdaGrad attains a rate of O(1/T + σ/ √ T ) by implicitly adapting to smoothness and noise levels; here T is the number of oracle queries and σ is the noise variance. They also proposed an accelerated AdaGrad variant with scalar step-size. The latter result was extended for compactly constrained problems via accelerated Mirror-Prox algorithm (Kavis et al., 2019), and for composite objectives (Joulani et al., 2020). Recently, Ene et al. (2021) have further generalized the latter results by designing a novel adaptive, accelerated algorithm with per-coordinate step-sizes. Convergence properties of such algorithms under smooth, non-convex losses are unknown to date.Adaptive methods for nonconvex optimization Following the popularity of neural networks, adaptive methods have attracted massive attention due to their favorable performance in training and their ease of tuning. The literature is quite vast, which is impossible to cover exhaustively | HIGH PROBABILITY BOUNDS FOR A CLASS OF NON- CONVEX ALGORITHMS WITH ADAGRAD STEPSIZE |
d53216389 | We consider the problem of aligning continuous word representations, learned in multiple languages, to a common space. It was recently shown that, in the case of two languages, it is possible to learn such a mapping without supervision. This paper extends this line of work to the problem of aligning multiple languages to a common space. A solution is to independently map all languages to a pivot language. Unfortunately, this degrades the quality of indirect word translation. We thus propose a novel formulation that ensures composable mappings, leading to better alignments. We evaluate our method by jointly aligning word vectors in eleven languages, showing consistent improvement with indirect mappings while maintaining competitive performance on direct word translation. | UNSUPERVISED HYPERALIGNMENT FOR MULTILIN- GUAL WORD EMBEDDINGS |
d258108311 | We propose the gradient-weighted Object Detector Activation Maps (ODAM), a visualized explanation technique for interpreting the predictions of object detectors. Utilizing the gradients of detector targets flowing into the intermediate feature maps, ODAM produces heat maps that show the influence of regions on the detector's decision for each predicted attribute. Compared to previous works classification activation maps (CAM), ODAM generates instance-specific explanations rather than class-specific ones. We show that ODAM is applicable to both one-stage detectors and two-stage detectors with different types of detector backbones and heads, and produces higher-quality visual explanations than the state-of-the-art both effectively and efficiently. We next propose a training scheme, Odam-Train, to improve the explanation ability on object discrimination of the detector through encouraging consistency between explanations for detections on the same object, and distinct explanations for detections on different objects. Based on the heat maps produced by ODAM with Odam-Train, we propose Odam-NMS, which considers the information of the model's explanation for each prediction to distinguish the duplicate detected objects. We present a detailed analysis of the visualized explanations of detectors and carry out extensive experiments to validate the effectiveness of the proposed ODAM.arXiv:2304.06354v1 [cs.CV] 13 Apr 2023Published as a conference paper at ICLR 2023 (a) Grad-CAM (b) D-RISE (5000 masks with 8x8) (b) D-RISE (5000 masks with resolution 16x16) (c) ODAM w/o Odam-Train (ours) (d) ODAM w/ Odam-Train (ours) Figure 1: Comparison of heat maps from Grad-CAM (Selvaraju et al., 2017), D-RISE (Petsiuk et al., 2021) and our ODAM. The white box shows the corresponding detected object. (a) Grad-CAM highlights all objects of the same category (person) instead of the specific object instance. (b) D-RISE maps have noisy backgrounds and its effectiveness depends on the mask size; the 16x16 mask is better for smaller objects (baseball bat) than larger objects (person). (c) ODAM generates instance-specific heat maps with less noise and is robust to object size. (d) With Odam-Train, the heat map is better localized over the object and separated from other objects.However, Grad-CAM provides class-specific explanations, and thus produces heat maps that highlight all objects of in a category instead of explaining a single detection (e.g., seeFig. 1a). For object detection, the explanations should be instance-specific rather than class-specific, so as to discriminate each individual object. Exploring the spatial importance of different objects can help interpret the models' decision and show the important area in the feature maps for each prediction.Considering that direct application of existing CAM methods to object detectors is infeasible and the drawbacks of the current state-of-the-art D-RISE, we propose gradient-weighted Object Detector Activation Maps (ODAM). ODAM adopts a similar assumption as Grad-CAM that feature maps correlate with some concept for making the final outputs. Thus ODAM uses the gradients w.r.t. each pixel in the feature map to obtain the explanation heat map for each attribute of the object prediction. Compared with the perturbation-based D-RISE, ODAM is more efficient and generates less noisy heat maps (seeFig. 1c), while also explaining each attribute separately.We also explore a unique explanation task for object detectors, object discrimination, which aims to explain which object was detected. This is different from the traditional explanation task of what features are important for class prediction (i.e., object specfication). We propose a training scheme, Odam-Train, to improve the explanation ability for object discrimination by introducing consistency and separation losses. The training encourages the model to produce consistent heat maps for the same object, and distinctive heat maps for different objects (seeFig. 1d). We further propose Odam-NMS, which uses the instance-level heat maps from ODAM to aid the non-maximum suppression (NMS) process of removing duplicate predictions of the same object.The contributions of our paper are summarized as follows: 1. We propose ODAM, a gradient-based visual explanation approach to produce instance-specific heat maps for explaining prediction attributes of object detectors, which is more efficient and robust compared with the current state-of-the-art. 2. We demonstrate the generalizability of ODAM by exhibiting explanations on one-and twostage, and transformer-based detectors with different types of backbones and detector heads. 3. We explore a unique explanation task for detector, object discrimination, for explaining which object was detected, and propose Odam-Train to obtain model with better object discrimination ability. 4. We propose Odam-NMS, which uses the instance-level heat maps generated by ODAM with Odam-Train to remove duplicate predictions during NMS, and its effectiveness verifies the object discrimination ability of ODAM with Odam-Train. Zeming Lin, Alban Desmaison, Luca Antiga, and Adam Lerer. Automatic differentiation in pytorch. 2017.Vitali Petsiuk, Abir Das, and Kate Saenko. Rise: Randomized input sampling for explanation of black-box models. arXiv preprint arXiv:1806.07421, 2018. | Published as a conference paper at ICLR 2023 ODAM: GRADIENT-BASED INSTANCE-SPECIFIC VI- SUAL EXPLANATIONS FOR OBJECT DETECTION |
d231632854 | The recent research in semi-supervised learning (SSL) is mostly dominated by consistency regularization based methods which achieve strong performance. However, they heavily rely on domain-specific data augmentations, which are not easy to generate for all data modalities. Pseudo-labeling (PL) is a general SSL approach that does not have this constraint but performs relatively poorly in its original formulation. We argue that PL underperforms due to the erroneous high confidence predictions from poorly calibrated models; these predictions generate many incorrect pseudo-labels, leading to noisy training. We propose an uncertainty-aware pseudo-label selection (UPS) framework which improves pseudo labeling accuracy by drastically reducing the amount of noise encountered in the training process. Furthermore, UPS generalizes the pseudo-labeling process, allowing for the creation of negative pseudo-labels; these negative pseudo-labels can be used for multi-label classification as well as negative learning to improve the single-label classification. We achieve strong performance when compared to recent SSL methods on the CIFAR-10 and CIFAR-100 datasets. Also, we demonstrate the versatility of our method on the video dataset UCF-101 and the multi-label dataset Pascal VOC. | IN DEFENSE OF PSEUDO-LABELING: AN UNCERTAINTY-AWARE PSEUDO-LABEL SELEC- TION FRAMEWORK FOR SEMI-SUPERVISED LEARNING |
d233241158 | A Recent work has highlighted several advantages of enforcing orthogonality in the weight layers of deep networks, such as maintaining the stability of activations, preserving gradient norms, and enhancing adversarial robustness by enforcing low Lipschitz constants. Although numerous methods exist for enforcing the orthogonality of fully-connected layers, those for convolutional layers are more heuristic in nature, often focusing on penalty methods or limited classes of convolutions. In this work, we propose and evaluate an alternative approach to directly parameterize convolutional layers that are constrained to be orthogonal. Specifically, we propose to apply the Cayley transform to a skew-symmetric convolution in the Fourier domain, so that the inverse convolution needed by the Cayley transform can be computed efficiently. We compare our method to previous Lipschitz-constrained and orthogonal convolutional layers and show that it indeed preserves orthogonality to a high degree even for large convolutions. Applied to the problem of certified adversarial robustness, we show that networks incorporating the layer outperform existing deterministic methods for certified defense against 2 -norm-bounded adversaries, while scaling to larger architectures than previously investigated. Code is available at httpsPublished as a conference paper at ICLR 2021 non-square weight matrices. The transform requires efficiently computing the inverse of a particular convolution in the Fourier domain, which we show works well in practice.We demonstrate that our Cayley layer is indeed orthogonal in practice when implemented in 32-bit precision, irrespective of the number of channels. Further, we compare it to alternative convolutional and Lipschitz-constrained layers: we include them in several architectures and evaluate their deterministic certifiable robustness against an 2 -norm-bounded adversary. Our layer provides stateof-the-art results on this task. We also demonstrate that the layers empirically endow a considerable degree of robustness without adversarial training. Our layer generally outperforms the alternatives, particularly for larger architectures. | |
d13995862 | We propose a deep learning framework for modeling complex high-dimensional densities via Nonlinear Independent Component Estimation (NICE). It is based on the idea that a good representation is one in which the data has a distribution that is easy to model. For this purpose, a non-linear deterministic transformation of the data is learned that maps it to a latent space so as to make the transformed data conform to a factorized distribution, i.e., resulting in independent latent variables. We parametrize this transformation so that computing the determinant of the Jacobian and inverse Jacobian is trivial, yet we maintain the ability to learn complex non-linear transformations, via a composition of simple building blocks, each based on a deep neural network. The training criterion is simply the exact log-likelihood, which is tractable, and unbiased ancestral sampling is also easy. We show that this approach yields good generative models on four image datasets and can be used for inpainting. * Yoshua Bengio is a CIFAR Senior Fellow. | NICE: Non-linear Independent Components Estimation |
d252815535 | Successful and effective communication between humans and AI relies on a shared experience of the world. By training solely on written text, current language models (LMs) miss the grounded experience of humans in the real-world-their failure to relate language to the physical world causes knowledge to be misrepresented and obvious mistakes in their reasoning. We present Mind's Eye, a paradigm to ground language model reasoning in the physical world. Given a physical reasoning question, we use a computational physics engine (DeepMind's MuJoCo) to simulate the possible outcomes, and then use the simulation results as part of the input, which enables language models to perform reasoning. Experiments on 39 tasks in a physics alignment benchmark demonstrate that Mind's Eye can improve reasoning ability by a large margin (27.9% zero-shot, and 46.0% few-shot absolute accuracy improvement on average). Smaller language models armed with Mind's Eye can obtain similar performance to models that are 100× larger. Finally, we confirm the robustness of Mind's Eye through ablation studies. | MIND'S EYE: GROUNDED LANGUAGE MODEL REA- SONING THROUGH SIMULATION |
d233405190 | The learning rate (LR) schedule is one of the most important hyper-parameters needing careful tuning in training DNNs. However, it is also one of the least automated parts of machine learning systems and usually costs significant manual effort and computing. Though there are pre-defined LR schedules and optimizers with adaptive LR, they introduce new hyperparameters that need to be tuned separately for different tasks/datasets. In this paper, we consider the question: Can we automatically tune the LR over the course of training without human involvement? We propose an efficient method, AutoLRS, which automatically optimizes the LR for each training stage by modeling training dynamics. AutoLRS aims to find an LR applied to every τ steps that minimizes the resulted validation loss. We solve this black-box optimization on the fly by Bayesian optimization (BO). However, collecting training instances for BO requires a system to evaluate each LR queried by BO's acquisition function for τ steps, which is prohibitively expensive in practice. Instead, we apply each candidate LR for only τ τ steps and train an exponential model to predict the validation loss after τ steps. This mutual-training process between BO and the loss-prediction model allows us to limit the training steps invested in the BO search. We demonstrate the advantages and the generality of AutoLRS through extensive experiments of training DNNs for tasks from diverse domains using different optimizers. The LR schedules auto-generated by AutoLRS lead to a speedup of 1.22×, 1.43×, and 1.5× when training ResNet-50, Transformer, and BERT, respectively, compared to the LR schedules in their original papers, and an average speedup of 1.31× over state-of-the-art heavily-tuned LR schedules. A novel bandit-based approach to hyperparameter optimization. efficient online hyperparameter optimization with population-based bandits. Advances in Neural Information Processing Systems, 2020.Pranav Rajpurkar, Jian Zhang, Konstantin Lopyrev, and Percy Liang. SQuAD: 100,000+ questions for machine comprehension of text. In Leslie N Smith. A disciplined approach to neural network hyper-parameters: Part 1-learning rate, batch size, momentum, and weight decay. arXiv preprint arXiv:1803.09820, 2018. | Published as a conference paper at ICLR 2021 AUTOLRS: AUTOMATIC LEARNING-RATE SCHEDULE BY BAYESIAN OPTIMIZATION ON THE FLY |
d24029589 | Machine learning models with very low test error have been shown to be consistently vulnerable to small, adversarially chosen perturbations of the input. We hypothesize that this counterintuitive behavior is a result of the high-dimensional geometry of the data manifold, and explore this hypothesis on a simple highdimensional dataset. For this dataset we show a fundamental bound relating the classification error rate to the average distance to the nearest misclassification, which is independent of the model. We train different neural network architectures on this dataset and show their error sets approach this theoretical bound. As a result of the theory, the vulnerability of machine learning models to small adversarial perturbations is a logical consequence of the amount of test error observed. We hope that our theoretical analysis of this foundational synthetic case will point a way forward to explore how the geometry of complex real-world data sets leads to adversarial examples. | The Relationship Between High-Dimensional Geometry and Adversarial Examples |
d3532489 | Neural program embeddings have shown much promise recently for a variety of program analysis tasks, including program synthesis, program repair, fault localization, etc. However, most existing program embeddings are based on syntactic features of programs, such as raw token sequences or abstract syntax trees. Unlike images and text, a program has an unambiguous semantic meaning that can be difficult to capture by only considering its syntax (i.e. syntactically similar programs can exhibit vastly different run-time behavior), which makes syntax-based program embeddings fundamentally limited. This paper proposes a novel semantic program embedding that is learned from program execution traces. Our key insight is that program states expressed as sequential tuples of live variable values not only captures program semantics more precisely, but also offer a more natural fit for Recurrent Neural Networks to model. We evaluate different syntactic and semantic program embeddings on classifying the types of errors that students make in their submissions to an introductory programming class and two exercises on the CodeHunt education platform. Evaluation results show that our new semantic program embedding significantly outperforms the syntactic program embeddings based on token sequences and abstract syntax trees. In addition, we augment a search-based program repair system with the predictions obtained from our semantic embedding, and show that search efficiency is also significantly improved. | Published as a conference paper at ICLR 2018 DYNAMIC NEURAL PROGRAM EMBEDDINGS FOR PRO- GRAM REPAIR |
d256598360 | There is a recent trend of applying multi-agent reinforcement learning (MARL) to train an agent that can cooperate with humans in a zero-shot fashion without using any human data. The typical workflow is to first repeatedly run self-play (SP) to build a policy pool and then train the final adaptive policy against this pool. A crucial limitation of this framework is that every policy in the pool is optimized w.r.t. the environment reward function, which implicitly assumes that the testing partners of the adaptive policy will be precisely optimizing the same reward function as well. However, human objectives are often substantially biased according to their own preferences, which can differ greatly from the environment reward. We propose a more general framework, Hidden-Utility Self-Play (HSP), which explicitly models human biases as hidden reward functions in the self-play objective. By approximating the reward space as linear functions, HSP adopts an effective technique to generate an augmented policy pool with biased policies. We evaluate HSP on the Overcooked benchmark. Empirical results show that our HSP method produces higher rewards than baselines when cooperating with learned human models, manually scripted policies, and real humans. The HSP policy is also rated as the most assistive policy based on human feedback.Recently, multi-agent reinforcement learning (MARL) has become a promising approach for many challenging decision-making problems. Particularly in competitive settings, AIs developed by MARL algorithms based on self-play (SP) defeated human professionals in a variety of domains (Silver et al., 2018; Vinyals et al., 2019; Berner et al., 2019). This empirical evidence suggests a new direction of developing strong AIs that can directly cooperate with humans in a similar "model-free" fashion, i.e., via self-play.Different from zero-sum games, where simply adopting a Nash equilibrium strategy is sufficient, an obvious issue when training cooperative agents by self-play is convention overfitting. Due to the existence of a large number of possible optimal strategies in a cooperative game, SP-trained agents can easily converge to a particular optimum and make decisions solely based on a specific behavior pattern, i.e., convention (Lowe et al., 2019; Hu et al., 2020), of its co-trainers, leading to poor generalization ability to unseen partners. To tackle this problem, recent works proposed a two-staged framework by first developing a diverse policy pool consisting of multiple SP-trained policies, which possibly cover different conventions, and then further training an adaptive policy against this policy pool (Lupu et al., 2021; Strouse et al., 2021;Zhao et al., 2021).Despite the empirical success of this two-staged framework, a fundamental drawback exists. Even though the policy pool prevents convention overfitting, each SP-trained policy in the pool remains a solution, which is either optimal or sub-optimal, to a fixed reward function specified by the underlying cooperative game. This implies a crucial generalization assumption that any test-time partner * Equal contribution † Corresponding Author prospects of the human-robot collaboration. Autonomous Robots, 42(5):957-975, 2018.Bowen Baker. Emergent reciprocity and team formation from randomized uncertain social preferences. Advances in Neural Information Processing Systems, 33:15786-15799, 2020.Nicholas C Barberis. Thirty years of prospect theory in economics: A review and assessment. On the utility of learning about humans for human-AI coordination. Advances in neural information processing systems, 32, 2019.Rujikorn Charakorn, Poramate Manoonpong, and Nat Dilokthanakul. Learning to cooperate with unseen agents through meta-reinforcement learning. In | LEARNING ZERO-SHOT COOPERATION WITH HU- MANS, ASSUMING HUMANS ARE BIASED |
d257663689 | Although deep reinforcement learning (DRL) has many success stories, the largescale deployment of policies learned through these advanced techniques in safetycritical scenarios is hindered by their lack of formal guarantees. Variational Markov Decision Processes (VAE-MDPs) are discrete latent space models that provide a reliable framework for distilling formally verifiable controllers from any RL policy. While the related guarantees address relevant practical aspects such as the satisfaction of performance and safety properties, the VAE approach suffers from several learning flaws (posterior collapse, slow learning speed, poor dynamics estimates), primarily due to the absence of abstraction and representation guarantees to support latent optimization. We introduce the Wasserstein auto-encoded MDP (WAE-MDP), a latent space model that fixes those issues by minimizing a penalized form of the optimal transport between the behaviors of the agent executing the original policy and the distilled policy, for which the formal guarantees apply. Our approach yields bisimulation guarantees while learning the distilled policy, allowing concrete optimization of the abstraction and representation model quality. Our experiments show that, besides distilling policies up to 10 times faster, the latent model quality is indeed better in general. Moreover, we present experiments from a simple time-to-failure verification algorithm on the latent space. The fact that our approach enables such simple verification techniques highlights its applicability. | WASSERSTEIN AUTO-ENCODED MDPS FORMAL VERIFICATION OF EFFICIENTLY DISTILLED RL POLICIES WITH MANY- SIDED GUARANTEES |
d252735292 | Federated Learning (FL) is a setting for training machine learning models in distributed environments where the clients do not share their raw data but instead send model updates to a server. However, model updates can be subject to attacks and leak private information. Differential Privacy (DP) is a leading mitigation strategy which involves adding noise to clipped model updates, trading off performance for strong theoretical privacy guarantees. Previous work has shown that the threat model of DP is conservative and that the obtained guarantees may be vacuous or may overestimate information leakage in practice. In this paper, we aim to achieve a tighter measurement of the model exposure by considering a realistic threat model. We propose a novel method, CANIFE, that uses canaries-carefully crafted samples by a strong adversary to evaluate the empirical privacy of a training round. We apply this attack to vision models trained on CIFAR-10 and CelebA and to language models trained on Sent140 and Shakespeare. In particular, in realistic FL scenarios, we demonstrate that the empirical per-round epsilon obtained with CANIFE is 4 -5× lower than the theoretical bound. * Work done during an internship at Meta. | CANIFE: CRAFTING CANARIES FOR EMPIRICAL PRI- VACY MEASUREMENT IN FEDERATED LEARNING |
d250334377 | Gradient-based multilevel optimization (MLO) has gained attention as a framework for studying numerous problems, ranging from hyperparameter optimization and meta-learning to neural architecture search and reinforcement learning. However, gradients in MLO, which are obtained by composing best-response Jacobians via the chain rule, are notoriously difficult to implement and memory/compute intensive. We take an initial step towards closing this gap by introducing BETTY, a software library for large-scale MLO. At its core, we devise a novel dataflow graph for MLO, which allows us to (1) develop efficient automatic differentiation for MLO that reduces the computational complexity from O(d 3 ) to O(d 2 ), (2) incorporate systems support such as mixed-precision and data-parallel training for scalability, and (3) facilitate implementation of MLO programs of arbitrary complexity while allowing a modular interface for diverse algorithmic and systems design choices. We empirically demonstrate that BETTY can be used to implement an array of MLO programs, while also observing up to 11% increase in test accuracy, 14% decrease in GPU memory usage, and 20% decrease in training wall time over existing implementations on multiple benchmarks. We also showcase that BETTY enables scaling MLO to models with hundreds of millions of parameters. We open-source the code at https://github.com/leopard-ai/betty. | Published as a conference paper at ICLR 2023 BETTY: AN AUTOMATIC DIFFERENTIATION LIBRARY FOR MULTILEVEL OPTIMIZATION |
d195345577 | Sample inefficiency is a long-lasting problem in reinforcement learning (RL). The state-ofthe-art uses value function to derive policy while it usually requires an extensive search over the state-action space, which is one reason for the inefficiency. Towards the sample-efficient RL, we propose ranking policy gradient (RPG), a policy gradient method that learns the optimal ranking of a set of discrete actions. To accelerate the learning of policy gradient methods, we describe a novel off-policy learning framework and establish the equivalence between maximizing the lower bound of return and imitating a near-optimal policy without accessing any oracles. These results lead to a general sample-efficient off-policy learning framework, which accelerates learning and reduces variance. Furthermore, the sample complexity of RPG does not depend on the dimension of state space, which enables RPG for large-scale problems. We conduct extensive experiments showing that when consolidating with the off-policy learning framework, RPG substantially reduces the sample complexity, comparing to the state-of-the-art. Schapire. Contextual decision processes with low bellman rank are pac-learnable. In Sham Machandranath Kakade et al. On the sample complexity of reinforcement learning. PhD thesis, , et al. Human-level control through deep reinforcement learning. Nature, 518(7540):529, 2015. | Ranking Policy Gradient Ranking Policy Gradient |
d207880633 | Federated learning improves data privacy and efficiency in machine learning performed over networks of distributed devices, such as mobile phones, IoT and wearable devices, etc. Yet models trained with federated learning can still fail to generalize to new devices due to the problem of domain shift. Domain shift occurs when the labeled data collected by source nodes statistically differs from the target node's unlabeled data. In this work, we present a principled approach to the problem of federated domain adaptation, which aims to align the representations learned among the different nodes with the data distribution of the target node. Our approach extends adversarial adaptation techniques to the constraints of the federated setting. In addition, we devise a dynamic attention mechanism and leverage feature disentanglement to enhance knowledge transfer. Empirically, we perform extensive experiments on several image and text classification tasks and show promising results under unsupervised federated domain adaptation setting. | Published as a conference paper at ICLR 2020 FEDERATED ADVERSARIAL DOMAIN ADAPTATION |
d30293072 | Several recent works have discussed tree structured sparse coding[8,10,7,3], where N data points in R d written as the d × N matrix X are approximately decomposed into the product of matrices W Z. Here W is a d×K dictionary matrix, and Z is a K ×N matrix of coefficients. In tree structured sparse coding, the rows of Z correspond to nodes on a tree, and the columns of Z are encouraged to be nonzero on only a few branches of the tree; or alternatively, the columns are constrained to lie on at most a specified number of branches of the tree.When viewed from a geometric perspective, this kind of decomposition is a "wavelet analysis" of the data points in X [9, 6, 11, 1]. As each row in Z is associated to a column of W , the columns of W also take a tree structure. The decomposition corresponds to a multiscale clustering of the data, where the scale of the clustering is given by the depth in the tree, and cluster membership corresponds to activation of a row in Z. The root node rows of Z corresponds to the whole data set, and the root node columns of W are a best fit linear representation of X. The set of rows of Z corresponding to each node specify a cluster-a data point x is in that cluster if it has active responses in those rows. The set of columns of W corresponding to a node specify a linear correction to the best fit subspace defined by the nodes ancestors; the correction is valid on the corresponding cluster.Here we discuss the analagous construction on the binary cube {−1, 1} d . Linear best fit is replaced by best fit subcubes. | Tree structured sparse coding on cubes |
d249192278 | We study the Neural Optimal Transport (NOT) algorithm which uses the general optimal transport formulation and learns stochastic transport plans. We show that NOT with the weak quadratic cost may learn fake plans which are not optimal. To resolve this issue, we introduce kernel weak quadratic costs. We show that they provide improved theoretical guarantees and practical performance. We test NOT with kernel costs on the unpaired image-to-image translation task. (a) Celeba (female) → anime, 128 × 128. (b) Outdoor → church, 128 × 128.Figure 1: Unpaired image-to-image translation (one-to-many) by Kernel Neural Optimal Transport.1 . Score-based generative neural networks for large-scale optimal transport. | Published as a conference paper at ICLR 2023 KERNEL NEURAL OPTIMAL TRANSPORT |
d232146022 | Shapley values have become one of the most popular feature attribution explanation methods. However, most prior work has focused on post-hoc Shapley explanations, which can be computationally demanding due to its exponential time complexity and preclude model regularization based on Shapley explanations during training. Thus, we propose to incorporate Shapley values themselves as latent representations in deep models-thereby making Shapley explanations first-class citizens in the modeling paradigm. This intrinsic explanation approach enables layer-wise explanations, explanation regularization of the model during training, and fast explanation computation at test time. We define the Shapley transform that transforms the input into a Shapley representation given a specific function. We operationalize the Shapley transform as a neural network module and construct both shallow and deep networks, called SHAPNETs, by composing Shapley modules. We prove that our Shallow SHAPNETs compute the exact Shapley values and our Deep SHAPNETs maintain the missingness and accuracy properties of Shapley values. We demonstrate on synthetic and real-world datasets that our SHAPNETs enable layer-wise Shapley explanations, novel Shapley regularizations during training, and fast computation while maintaining reasonable performance. Code is available at https: //github.com/inouye-lab/ShapleyExplanationNetworks. | Published as a conference paper at ICLR 2021 SHAPLEY EXPLANATION NETWORKS |
d238419578 | Figure 1: Transform2Act learns a transform-and-control policy that first applies transform actions to design an agent and then controls the designed agent to interact with the environment. The giraffelike agent obtained by Transform2Act can run extremely fast and remain stable (see video).ABSTRACTAn agent's functionality is largely determined by its design, i.e., skeletal structure and joint attributes (e.g., length, size, strength). However, finding the optimal agent design for a given function is extremely challenging since the problem is inherently combinatorial and the design space is prohibitively large. Additionally, it can be costly to evaluate each candidate design which requires solving for its optimal controller. To tackle these problems, our key idea is to incorporate the design procedure of an agent into its decision-making process. Specifically, we learn a conditional policy that, in an episode, first applies a sequence of transform actions to modify an agent's skeletal structure and joint attributes, and then applies control actions under the new design. To handle a variable number of joints across designs, we use a graph-based policy where each graph node represents a joint and uses message passing with its neighbors to output joint-specific actions. Using policy gradient methods, our approach enables joint optimization of agent design and control as well as experience sharing across different designs, which improves sample efficiency substantially. Experiments show that our approach, Transform2Act, outperforms prior methods significantly in terms of convergence speed and final performance. Notably, Transform2Act can automatically discover plausible designs similar to giraffes, squids, and spiders. Code and videos are available at https://sites.google.com/view/transform2act. arXiv:2110.03659v3 [cs.LG] 9 Apr 2022Published as a conference paper at ICLR 2022 thermore, zeroth-order optimization methods such as ES are known to be sample-inefficient when the optimization search space (i.e., design space) is high-dimensional(Vemula et al., 2019). | TRANSFORM2ACT: LEARNING A TRANSFORM-AND- CONTROL POLICY FOR EFFICIENT AGENT DESIGN |
d258179227 | Decision Transformers (DT) have demonstrated strong performances in offline reinforcement learning settings, but quickly adapting to unseen novel tasks remains challenging. To address this challenge, we propose a new framework, called Hyper-Decision Transformer (HDT), that can generalize to novel tasks from a handful of demonstrations in a data-and parameter-efficient manner. To achieve such a goal, we propose to augment the base DT with an adaptation module, whose parameters are initialized by a hyper-network. When encountering unseen tasks, the hyper-network takes a handful of demonstrations as inputs and initializes the adaptation module accordingly. This initialization enables HDT to efficiently adapt to novel tasks by only fine-tuning the adaptation module. We validate HDT's generalization capability on object manipulation tasks. We find that with a single expert demonstration and fine-tuning only 0.5% of DT parameters, HDT adapts faster to unseen tasks than fine-tuning the whole DT model. Finally, we explore a more challenging setting where expert actions are not available, and we show that HDT outperforms state-of-the-art baselines in terms of task success rates by a large margin. Demos are available on our project page. 1 | HYPER-DECISION TRANSFORMER FOR EFFICIENT ON- LINE POLICY ADAPTATION |
d246276212 | We introduce a new approach for speech pre-training named SPIRAL which works by learning denoising representation of perturbed data in a teacher-student framework. Specifically, given a speech utterance, we first feed the utterance to a teacher network to obtain corresponding representation. Then the same utterance is perturbed and fed to a student network. The student network is trained to output representation resembling that of the teacher. At the same time, the teacher network is updated as moving average of student's weights over training steps. In order to prevent representation collapse, we apply an in-utterance contrastive loss as pre-training objective and impose position randomization on the input to the teacher. SPIRAL achieves competitive or better results compared to state-of-theart speech pre-training method wav2vec 2.0, with significant reduction of training cost (80% for BASE model, 65% for LARGE model). Furthermore, we address the problem of noise-robustness that is critical to real-world speech applications. We propose multi-condition pre-training by perturbing the student's input with various types of additive noise. We demonstrate that multi-condition pre-trained SPIRAL models are more robust to noisy speech (9.0% -13.3% relative word error rate reduction on real noisy test data), compared to applying multi-condition training solely in the fine-tuning stage. Source code is available 1 . | SPIRAL: SELF-SUPERVISED PERTURBATION- INVARIANT REPRESENTATION LEARNING FOR SPEECH PRE-TRAINING |
d204512179 | Reinforcement learning encounters major challenges in multi-agent settings, such as scalability and non-stationarity. Recently, value function factorization learning emerges as a promising way to address these challenges in collaborative multi-agent systems. However, existing methods have been focusing on learning fully decentralized value function, which are not efficient for tasks requiring communication. To address this limitation, this paper presents a novel framework for learning nearly decomposable value functions with communication, with which agents act on their own most of the time but occasionally send messages to other agents in order for effective coordination. This framework hybridizes value function factorization learning and communication learning by introducing two information-theoretic regularizers. These regularizers are maximizing mutual information between decentralized Q functions and communication messages while minimizing the entropy of messages between agents. We show how to optimize these regularizers in a way that is easily integrated with existing value function factorization methods such as QMIX. Finally, we demonstrate that, on the StarCraft unit micromanagement benchmark, our framework significantly outperforms baseline methods and allows to cut off more than 80% communication without sacrificing the performance. The video of our experiments is available at | LEARNING NEARLY DECOMPOSABLE VALUE FUNC- TIONS VIA COMMUNICATION MINIMIZATION |
d260446846 | We apply recurrent neural networks (RNN) on a new domain, namely recommender systems. Real-life recommender systems often face the problem of having to base recommendations only on short session-based data (e.g. a small sportsware website) instead of long user histories (as in the case of Netflix). In this situation the frequently praised matrix factorization approaches are not accurate. This problem is usually overcome in practice by resorting to item-to-item recommendations, i.e. recommending similar items. We argue that by modeling the whole session, more accurate recommendations can be provided. We therefore propose an RNNbased approach for session-based recommendations. Our approach also considers practical aspects of the task and introduces several modifications to classic RNNs such as a ranking loss function that make it more viable for this specific problem. Experimental results on two data-sets show marked improvements over widely used approaches. * The author spent 3 months at Telefonica Research during the research of this topic. † This work was done while the author was a member | SESSION-BASED RECOMMENDATIONS WITH RECURRENT NEURAL NETWORKS |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.