title
stringlengths 5
246
| categories
stringlengths 5
94
⌀ | abstract
stringlengths 54
5.03k
| authors
stringlengths 0
6.72k
| doi
stringlengths 12
54
⌀ | id
stringlengths 6
10
⌀ | year
float64 2.02k
2.02k
⌀ | venue
stringclasses 13
values |
|---|---|---|---|---|---|---|---|
When AUC meets DRO: Optimizing Partial AUC for Deep Learning with Non-Convex Convergence Guarantee | null | In this paper, we propose systematic and efficient gradient-based methods for both one-way and two-way partial AUC (pAUC) maximization that are applicable to deep learning. We propose new formulations of pAUC surrogate objectives by using the distributionally robust optimization (DRO) to define the loss for each individual positive data. We consider two formulations of DRO, one of which is based on conditional-value-at-risk (CVaR) that yields a non-smooth but exact estimator for pAUC, and another one is based on a KL divergence regularized DRO that yields an inexact but smooth (soft) estimator for pAUC. For both one-way and two-way pAUC maximization, we propose two algorithms and prove their convergence for optimizing their two formulations, respectively. Experiments demonstrate the effectiveness of the proposed algorithms for pAUC maximization for deep learning on various datasets. | Dixian Zhu, Gang Li, Bokun Wang, Xiaodong Wu, Tianbao Yang | null | null | 2,022 | icml |
Topology-aware Generalization of Decentralized SGD | null | This paper studies the algorithmic stability and generalizability of decentralized stochastic gradient descent (D-SGD). We prove that the consensus model learned by D-SGD is $\mathcal{O}{(m/N\unaryplus1/m\unaryplus\lambda^2)}$-stable in expectation in the non-convex non-smooth setting, where $N$ is the total sample size of the whole system, $m$ is the worker number, and $1\unaryminus\lambda$ is the spectral gap that measures the connectivity of the communication topology. These results then deliver an $\mathcal{O}{(1/N\unaryplus{({(m^{-1}\lambda^2)}^{\frac{\alpha}{2}}\unaryplus m^{\unaryminus\alpha})}/{N^{1\unaryminus\frac{\alpha}{2}}})}$ in-average generalization bound, which is non-vacuous even when $\lambda$ is closed to $1$, in contrast to vacuous as suggested by existing literature on the projected version of D-SGD. Our theory indicates that the generalizability of D-SGD has a positive correlation with the spectral gap, and can explain why consensus control in initial training phase can ensure better generalization. Experiments of VGG-11 and ResNet-18 on CIFAR-10, CIFAR-100 and Tiny-ImageNet justify our theory. To our best knowledge, this is the first work on the topology-aware generalization of vanilla D-SGD. Code is available at \url{https://github.com/Raiden-Zhu/Generalization-of-DSGD}. | Tongtian Zhu, Fengxiang He, Lan Zhang, Zhengyang Niu, Mingli Song, Dacheng Tao | null | null | 2,022 | icml |
Contextual Bandits with Smooth Regret: Efficient Learning in Continuous Action Spaces | null | Designing efficient general-purpose contextual bandit algorithms that work with large—or even infinite—action spaces would facilitate application to important scenarios such as information retrieval, recommendation systems, and continuous control. While obtaining standard regret guarantees can be hopeless, alternative regret notions have been proposed to tackle the large action setting. We propose a smooth regret notion for contextual bandits, which dominates previously proposed alternatives. We design a statistically and computationally efficient algorithm—for the proposed smooth regret—that works with general function approximation under standard supervised oracles. We also present an adaptive algorithm that automatically adapts to any smoothness level. Our algorithms can be used to recover the previous minimax/Pareto optimal guarantees under the standard regret, e.g., in bandit problems with multiple best arms and Lipschitz/H{ö}lder bandits. We conduct large-scale empirical evaluations demonstrating the efficacy of our proposed algorithms. | Yinglun Zhu, Paul Mineiro | null | null | 2,022 | icml |
Contextual Bandits with Large Action Spaces: Made Practical | null | A central problem in sequential decision making is to develop algorithms that are practical and computationally efficient, yet support the use of flexible, general-purpose models. Focusing on the contextual bandit problem, recent progress provides provably efficient algorithms with strong empirical performance when the number of possible alternatives (“actions”) is small, but guarantees for decision making in large, continuous action spaces have remained elusive, leading to a significant gap between theory and practice. We present the first efficient, general-purpose algorithm for contextual bandits with continuous, linearly structured action spaces. Our algorithm makes use of computational oracles for (i) supervised learning, and (ii) optimization over the action space, and achieves sample complexity, runtime, and memory independent of the size of the action space. In addition, it is simple and practical. We perform a large-scale empirical evaluation, and show that our approach typically enjoys superior performance and efficiency compared to standard baselines. | Yinglun Zhu, Dylan J Foster, John Langford, Paul Mineiro | null | null | 2,022 | icml |
Region-Based Semantic Factorization in GANs | null | Despite the rapid advancement of semantic discovery in the latent space of Generative Adversarial Networks (GANs), existing approaches either are limited to finding global attributes or rely on a number of segmentation masks to identify local attributes. In this work, we present a highly efficient algorithm to factorize the latent semantics learned by GANs concerning an arbitrary image region. Concretely, we revisit the task of local manipulation with pre-trained GANs and formulate region-based semantic discovery as a dual optimization problem. Through an appropriately defined generalized Rayleigh quotient, we manage to solve such a problem without any annotations or training. Experimental results on various state-of-the-art GAN models demonstrate the effectiveness of our approach, as well as its superiority over prior arts regarding precise control, region robustness, speed of implementation, and simplicity of use. | Jiapeng Zhu, Yujun Shen, Yinghao Xu, Deli Zhao, Qifeng Chen | null | null | 2,022 | icml |
Neural-Symbolic Models for Logical Queries on Knowledge Graphs | null | Answering complex first-order logic (FOL) queries on knowledge graphs is a fundamental task for multi-hop reasoning. Traditional symbolic methods traverse a complete knowledge graph to extract the answers, which provides good interpretation for each step. Recent neural methods learn geometric embeddings for complex queries. These methods can generalize to incomplete knowledge graphs, but their reasoning process is hard to interpret. In this paper, we propose Graph Neural Network Query Executor (GNN-QE), a neural-symbolic model that enjoys the advantages of both worlds. GNN-QE decomposes a complex FOL query into relation projections and logical operations over fuzzy sets, which provides interpretability for intermediate variables. To reason about the missing links, GNN-QE adapts a graph neural network from knowledge graph completion to execute the relation projections, and models the logical operations with product fuzzy logic. Experiments on 3 datasets show that GNN-QE significantly improves over previous state-of-the-art models in answering FOL queries. Meanwhile, GNN-QE can predict the number of answers without explicit supervision, and provide visualizations for intermediate variables. | Zhaocheng Zhu, Mikhail Galkin, Zuobai Zhang, Jian Tang | null | null | 2,022 | icml |
VLUE: A Multi-Task Multi-Dimension Benchmark for Evaluating Vision-Language Pre-training | null | Recent advances in vision-language pre-training (VLP) have demonstrated impressive performance in a range of vision-language (VL) tasks. However, there exist several challenges for measuring the community’s progress in building general multi-modal intelligence. First, most of the downstream VL datasets are annotated using raw images that are already seen during pre-training, which may result in an overestimation of current VLP models’ generalization ability. Second, recent VLP work mainly focuses on absolute performance but overlooks the efficiency-performance trade-off, which is also an important indicator for measuring progress. To this end, we introduce the Vision-Language Understanding Evaluation (VLUE) benchmark, a multi-task multi-dimension benchmark for evaluating the generalization capabilities and the efficiency-performance trade-off (“Pareto SOTA”) of VLP models. We demonstrate that there is a sizable generalization gap for all VLP models when testing on out-of-distribution test sets annotated on images from a more diverse distribution that spreads across cultures. Moreover, we find that measuring the efficiency-performance trade-off of VLP models leads to complementary insights for several design choices of VLP. We release the VLUE benchmark to promote research on building vision-language models that generalize well to images unseen during pre-training and are practical in terms of efficiency-performance trade-off. | Wangchunshu Zhou, Yan Zeng, Shizhe Diao, Xinsong Zhang | null | null | 2,022 | icml |
Towards Uniformly Superhuman Autonomy via Subdominance Minimization | null | Prevalent imitation learning methods seek to produce behavior that matches or exceeds average human performance. This often prevents achieving expert-level or superhuman performance when identifying the better demonstrations to imitate is difficult. We instead assume demonstrations are of varying quality and seek to induce behavior that is unambiguously better (i.e., Pareto dominant or minimally subdominant) than all human demonstrations. Our minimum subdominance inverse optimal control training objective is primarily defined by high quality demonstrations; lower quality demonstrations, which are more easily dominated, are effectively ignored instead of degrading imitation. With increasing probability, our approach produces superhuman behavior incurring lower cost than demonstrations on the demonstrator’s unknown cost function{—}even if that cost function differs for each demonstration. We apply our approach on a computer cursor pointing task, producing behavior that is 78% superhuman, while minimizing demonstration suboptimality provides 50% superhuman behavior{—}and only 72% even after selective data cleaning. | Brian Ziebart, Sanjiban Choudhury, Xinyan Yan, Paul Vernaza | null | null | 2,022 | icml |
Residual-Based Sampling for Online Outlier-Robust PCA | null | Outlier-robust principal component analysis (ORPCA) has been broadly applied in scientific discovery in the last decades. In this paper, we study online ORPCA, an important variant that addresses the practical challenge that the data points arrive in a sequential manner and the goal is to recover the underlying subspace of the clean data with one pass of the data. Our main contribution is the first provable algorithm that enjoys comparable recovery guarantee to the best known batch algorithm, while significantly improving upon the state-of-the-art online ORPCA algorithms. The core technique is a robust version of the residual norm which, informally speaking, leverages not only the importance of a data point, but also how likely it behaves as an outlier. | Tianhao Zhu, Jie Shen | null | null | 2,022 | icml |
SpaceMAP: Visualizing High-Dimensional Data by Space Expansion | null | Dimensionality reduction (DR) of high-dimensional data is of theoretical and practical interest in machine learning. However, there exist intriguing, non-intuitive discrepancies between the geometry of high- and low-dimensional space. We look into such discrepancies and propose a novel visualization method called Space-based Manifold Approximation and Projection (SpaceMAP). Our method establishes an analytical transformation on distance metrics between spaces to address the “crowding problem" in DR. With the proposed equivalent extended distance (EED), we are able to match the capacity of high- and low-dimensional space in a principled manner. To handle complex data with different manifold properties, we propose hierarchical manifold approximation to model the similarity function in a data-specific manner. We evaluated SpaceMAP on a range of synthetic and real datasets with varying manifold properties, and demonstrated its excellent performance in comparison with classical and state-of-the-art DR methods. In particular, the concept of space expansion provides a generic framework for understanding nonlinear DR methods including the popular t-distributed Stochastic Neighbor Embedding (t-SNE) and Uniform Manifold Approximation and Projection | Xinrui Zu, Qian Tao | null | null | 2,022 | icml |
Inductive Matrix Completion: No Bad Local Minima and a Fast Algorithm | null | The inductive matrix completion (IMC) problem is to recover a low rank matrix from few observed entries while incorporating prior knowledge about its row and column subspaces. In this work, we make three contributions to the IMC problem: (i) we prove that under suitable conditions, the IMC optimization landscape has no bad local minima; (ii) we derive a simple scheme with theoretical guarantees to estimate the rank of the unknown matrix; and (iii) we propose GNIMC, a simple Gauss-Newton based method to solve the IMC problem, analyze its runtime and derive for it strong recovery guarantees. The guarantees for GNIMC are sharper in several aspects than those available for other methods, including a quadratic convergence rate, fewer required observed entries and stability to errors or deviations from low-rank. Empirically, given entries observed uniformly at random, GNIMC recovers the underlying matrix substantially faster than several competing methods. | Pini Zilber, Boaz Nadler | null | null | 2,022 | icml |
Counterfactual Prediction for Outcome-Oriented Treatments | null | Large amounts of efforts have been devoted into learning counterfactual treatment outcome under various settings, including binary/continuous/multiple treatments. Most of these literature aims to minimize the estimation error of counterfactual outcome for the whole treatment space. However, in most scenarios when the counterfactual prediction model is utilized to assist decision-making, people are only concerned with the small fraction of treatments that can potentially induce superior outcome (i.e. outcome-oriented treatments). This gap of objective is even more severe when the number of possible treatments is large, for example under the continuous treatment setting. To overcome it, we establish a new objective of optimizing counterfactual prediction on outcome-oriented treatments, propose a novel Outcome-Oriented Sample Re-weighting (OOSR) method to make the predictive model concentrate more on outcome-oriented treatments, and theoretically analyze that our method can improve treatment selection towards the optimal one. Extensive experimental results on both synthetic datasets and semi-synthetic datasets demonstrate the effectiveness of our method. | Hao Zou, Bo Li, Jiangang Han, Shuiping Chen, Xuetao Ding, Peng Cui | null | null | 2,022 | icml |
Wiring Up Vision: Minimizing Supervised Synaptic Updates Needed to Produce a Primate Ventral Stream | null | After training on large datasets, certain deep neural networks are surprisingly good models of the neural mechanisms of adult primate visual object recognition. Nevertheless, these models are considered poor models of the development of the visual system because they posit millions of sequential, precisely coordinated synaptic updates, each based on a labeled image. While ongoing research is pursuing the use of unsupervised proxies for labels, we here explore a complementary strategy of reducing the required number of supervised synaptic updates to produce an adult-like ventral visual stream (as judged by the match to V1, V2, V4, IT, and behavior). Such models might require less precise machinery and energy expenditure to coordinate these updates and would thus move us closer to viable neuroscientific hypotheses about how the visual system wires itself up. Relative to standard model training on labeled images in ImageNet, we here demonstrate that the total number of supervised weight updates can be substantially reduced using three complementary strategies: First, we find that only 2% of supervised updates (epochs and images) are needed to achieve 80% of the match to adult ventral stream. Specifically, training benefits predictions of higher visual cortex the most whereas early visual cortex predictions only improve marginally over the course of training. Second, by improving the random distribution of synaptic connectivity, we find that 54% of the brain match can already be achieved “at birth" (i.e. no training at all). Third, we find that, by training only 5% of model synapses, we can still achieve nearly 80% of the match to the ventral stream. This approach further improves on ImageNet performance over previous attempts in computer vision of minimizing trained components without substantially increasing the relative number of trained parameters. These results reflect first steps in modeling not just primate adult visual processing during inference, but also how the ventral visual stream might be "wired up" by evolution (a model's "birth" state) and by developmental learning (a model's updates based on visual experience). | Franziska Geiger, Martin Schrimpf, Tiago Marques, James J. DiCarlo | null | null | 2,022 | iclr |
DeSKO: Stability-Assured Robust Control with a Deep Stochastic Koopman Operator | null | The Koopman operator theory linearly describes nonlinear dynamical systems in a high-dimensional functional space and it allows to apply linear control methods to highly nonlinear systems. However, the Koopman operator does not account for any uncertainty in dynamical systems, causing it to perform poorly in real-world applications.
Therefore, we propose a deep stochastic Koopman operator (DeSKO) model in a robust learning control framework to guarantee stability of nonlinear stochastic systems. The DeSKO model captures a dynamical system's uncertainty by inferring a distribution of observables. We use the inferred distribution to design a robust, stabilizing closed-loop controller for a dynamical system. Modeling and control experiments on several advanced control benchmarks show that our framework is more robust and scalable than state-of-the-art deep Koopman operators and reinforcement learning methods. Tested control benchmarks include a soft robotic arm, a legged robot, and a biological gene regulatory network. We also demonstrate that this robust control method resists previously unseen uncertainties, such as external disturbances, with a magnitude of up to five times the maximum control input. Our approach opens up new possibilities in learning control for high-dimensional nonlinear systems while robustly managing internal or external uncertainty. | Minghao Han, Jacob Euler-Rolle, Robert K. Katzschmann | null | null | 2,022 | iclr |
Gesture2Vec: Clustering Gestures using Representation Learning Methods for Co-speech Gesture Generation | null | Co-speech gestures are a principal component in conveying messages and enhancing interaction experiences between humans. Similarly, the co-speech gesture is a key ingredient in human-agent interaction including both virtual agents and robots. Existing machine learning approaches have yielded only marginal success in learning speech-to-motion at the frame level. Current methods generate repetitive gesture sequences that lack appropriateness with respect to the speech context. In this paper, we propose a Gesture2Vec model using representation learning methods to learn the relationship between semantic features and corresponding gestures. We propose a vector-quantized variational autoencoder structure as well as training techniques to learn a rigorous representation of gesture sequences. Furthermore, we use a machine translation model that takes input text and translates it into a discrete sequence of associated gesture chunks in the learned gesture space. Ultimately, we use translated quantized gestures from the input text as an input to the autoencoder’s decoder to produce gesture sequences. The resulting gestures can be applied to both virtual agents and humanoid robots. Subjective and objective evaluations confirm the success of our approach in terms of appropriateness, human-likeness, and diversity. | Payam Jome Yazdian, Mo Chen, Angelica Lim | null | null | 2,022 | iclr |
Learning Hierarchical Structures with Differentiable Nondeterministic Stacks | null | Learning hierarchical structures in sequential data -- from simple algorithmic patterns to natural language -- in a reliable, generalizable way remains a challenging problem for neural language models. Past work has shown that recurrent neural networks (RNNs) struggle to generalize on held-out algorithmic or syntactic patterns without supervision or some inductive bias. To remedy this, many papers have explored augmenting RNNs with various differentiable stacks, by analogy with finite automata and pushdown automata (PDAs). In this paper, we improve the performance of our recently proposed Nondeterministic Stack RNN (NS-RNN), which uses a differentiable data structure that simulates a nondeterministic PDA, with two important changes. First, the model now assigns unnormalized positive weights instead of probabilities to stack actions, and we provide an analysis of why this improves training. Second, the model can directly observe the state of the underlying PDA. Our model achieves lower cross-entropy than all previous stack RNNs on five context-free language modeling tasks (within 0.05 nats of the information-theoretic lower bound), including a task on which the NS-RNN previously failed to outperform a deterministic stack RNN baseline. Finally, we propose a restricted version of the NS-RNN that incrementally processes infinitely long sequences, and we present language modeling results on the Penn Treebank. | Brian DuSell, David Chiang | null | null | 2,022 | iclr |
Strongly Self-Normalizing Neural Networks with Applications to Implicit Representation Learning | null | Recent studies have show that wide neural networks with orthogonal linear layers and Gaussian Poincaré normalized activation functions avoid vanishing and exploding gradients for input vectors with the correct magnitude. This paper introduces a strengthening of the condition that the activation function must be Gaussian Poincaré normalized which creates robustness to deviations from standard normal distribution in the pre-activations, thereby reducing the dependence on the requirement that the network is wide and that the input vector has the correct magnitude. In implicit representation learning this allows the training of deep networks of this type where the linear layers are no longer constrained to be orthogonal linear transformations. Networks of this type can be fitted to a reference image to 1/10th the mean square error achievable with previous methods. Herein is also given an improved positional encoding for implicit representation learning of two-dimensional images and a small-batch training procedure for fitting of neural networks to images which allows fitting in fewer epochs, leading to substantial improvement in training time. | Marcus Lång | null | null | 2,022 | iclr |
Fairness in Representation for Multilingual NLP: Insights from Controlled Experiments on Conditional Language Modeling | null | We perform systematically and fairly controlled experiments with the 6-layer Transformer to investigate the hardness in conditional-language-modeling languages which have been traditionally considered morphologically rich (AR and RU) and poor (ZH). We evaluate through statistical comparisons across 30 possible language directions from the 6 languages of the United Nations Parallel Corpus across 5 data sizes on 3 representation levels --- character, byte, and word. Results show that performance is relative to the representation granularity of each of the languages, not to the language as a whole. On the character and byte levels, we are able to eliminate statistically significant performance disparity, hence demonstrating that a language cannot be intrinsically hard. The disparity that mirrors the morphological complexity hierarchy is shown to be a byproduct of word segmentation. Evidence from data statistics, along with the fact that word segmentation is qualitatively indeterminate, renders a decades-long debate on morphological complexity (unless it is being intentionally modeled in a word-based, meaning-driven context) irrelevant in the context of computing. The intent of our work is to help effect more objectivity and adequacy in evaluation as well as fairness and inclusivity in experimental setup in the area of language and computing so to uphold diversity in Machine Learning and Artificial Intelligence research. Multilinguality is real and relevant in computing not due to canonical, structural linguistic concepts such as morphology or "words" in our minds, but rather standards related to internationalization and localization, such as character encoding --- something which has thus far been sorely overlooked in our discourse and curricula. | Ada Wan | null | null | 2,022 | iclr |
MaGNET: Uniform Sampling from Deep Generative Network Manifolds Without Retraining | null | Deep Generative Networks (DGNs) are extensively employed in Generative Adversarial Networks (GANs), Variational Autoencoders (VAEs), and their variants to approximate the data manifold, and data distribution on that manifold. However, training samples are often obtained based on preferences, costs, or convenience producing artifacts in the empirical data distribution e.g. the large fraction of smiling faces in the CelebA dataset or the large fraction of dark-haired individuals in FFHQ). {\em These inconsistencies will be reproduced when sampling from the trained DGN, which has far-reaching potential implications for fairness, data augmentation, anomaly detection, domain adaptation, and beyond.} In response, we develop a differential geometry based sampler -coined MaGNET- that, given any trained DGN, produces samples that are uniformly distributed on the learned manifold. We prove theoretically and empirically that our technique produces a uniform distribution on the manifold regardless of the training set distribution. We perform a range of experiments on various datasets and DGNs. One of them considers the state-of-the-art StyleGAN2 trained on FFHQ dataset, where uniform sampling via MaGNET increases distribution precision \& recall by 4.12\% \& 3.01\% and decreases gender bias by 41.2\%, without requiring labels or retraining. | Ahmed Imtiaz Humayun, Randall Balestriero, Richard Baraniuk | null | null | 2,022 | iclr |
iLQR-VAE : control-based learning of input-driven dynamics with applications to neural data | null | Understanding how neural dynamics give rise to behaviour is one of the most fundamental questions in systems neuroscience. To achieve this, a common approach is to record neural populations in behaving animals, and model these data as emanating from a latent dynamical system whose state trajectories can then be related back to behavioural observations via some form of decoding. As recordings are typically performed in localized circuits that form only a part of the wider implicated network, it is important to simultaneously learn the local dynamics and infer any unobserved external input that might drive them. Here, we introduce iLQR-VAE, a novel control-based approach to variational inference in nonlinear dynamical systems, capable of learning both latent dynamics, initial conditions, and ongoing external inputs. As in recent deep learning approaches, our method is based on an input-driven sequential variational autoencoder (VAE). The main novelty lies in the use of the powerful iterative linear quadratic regulator algorithm (iLQR) in the recognition model. Optimization of the standard evidence lower-bound requires differentiating through iLQR solutions, which is made possible by recent advances in differentiable control. Importantly, having the recognition model be implicitly defined by the generative model greatly reduces the number of free parameters and allows for flexible, high-quality inference. This makes it possible for instance to evaluate the model on a single long trial after training on smaller chunks. We demonstrate the effectiveness of iLQR-VAE on a range of synthetic systems, with autonomous as well as input-driven dynamics. We further apply it to neural and behavioural recordings in non-human primates performing two different reaching tasks, and show that iLQR-VAE yields high-quality kinematic reconstructions from the neural data. | Marine Schimel, Ta-Chu Kao, Kristopher T Jensen, Guillaume Hennequin | null | null | 2,022 | iclr |
DNBP: Differentiable Nonparametric Belief Propagation | null | We present a differentiable approach to learn the probabilistic factors used for inference by a nonparametric belief propagation algorithm. Existing nonparametric belief propagation methods rely on domain-specific features encoded in the probabilistic factors of a graphical model. In this work, we replace each crafted factor with a differentiable neural network enabling the factors to be learned using an efficient optimization routine from labeled data. By combining differentiable neural networks with an efficient belief propagation algorithm, our method learns to maintain a set of marginal posterior samples using end-to-end training. We evaluate our differentiable nonparametric belief propagation (DNBP) method on a set of articulated pose tracking tasks and compare performance with learned baselines. Results from these experiments demonstrate the effectiveness of using learned factors for tracking and suggest the practical advantage over hand-crafted approaches. The project webpage is available at: https://sites.google.com/view/diff-nbp | Anthony Opipari, Jana Pavlasek, Chao Chen, Shoutian Wang, Karthik Desingh, Odest Jenkins | null | null | 2,022 | iclr |
It Takes Four to Tango: Multiagent Self Play for Automatic Curriculum Generation | null | We are interested in training general-purpose reinforcement learning agents that can solve a wide variety of goals. Training such agents efficiently requires automatic generation of a goal curriculum. This is challenging as it requires (a) exploring goals of increasing difficulty, while ensuring that the agent (b) is exposed to a diverse set of goals in a sample efficient manner and (c) does not catastrophically forget previously solved goals. We propose Curriculum Self Play (CuSP), an automated goal generation framework that seeks to satisfy these desiderata by virtue of a multi-player game with 4 agents. We extend the asymmetric curricula learning in PAIRED (Dennis et al., 2020) to a symmetrized game that carefully balances cooperation and competition between two off-policy student learners and two regret-maximizing teachers. CuSP additionally introduces entropic goal coverage and accounts for the non-stationary nature of the students, allowing us to automatically induce a curriculum that balances progressive exploration with anti-catastrophic exploitation. We demonstrate that our method succeeds at generating an effective curricula of goals for a range of control tasks, outperforming other methods at zero-shot test-time generalization to novel out-of-distribution goals. | Yuqing Du, Pieter Abbeel, Aditya Grover | null | null | 2,022 | iclr |
Towards Understanding Catastrophic Overfitting in Fast Adversarial Training | null | After adversarial training was proposed, a series of works focus on improving the compunational efficiency of adversarial training for deep neural networks (DNNs). Recently, FGSM based single-step adversarial training has been found to be able to train a robust model with the robustness comparable to the one trained by multi-step PGD, but it is an order of magnitude faster. However, there exists a failure mode called Catastrophic Overfitting (CO) where the network loses its robustness against PGD attack suddenly and can be hardly recovered by itself during the training process. In this paper, we identify that CO is closely related to the high-order terms in Taylor expansion after rethinking and decomposing the min-max problem in adversarial training. The negative high-order terms lead to a phenomenon called Perturbation Loss Distortion, which is the underlying cause of CO. Based on the observations, we propose a simple but effective regularization method named Fast Linear Adversarial Training (FLAT) to avoid CO in the single-step adversarial training by making the loss surface flat. | Renjie Chen, Yuan Luo, Yisen Wang | null | null | 2,022 | iclr |
Domino: Discovering Systematic Errors with Cross-Modal Embeddings | null | Machine learning models that achieve high overall accuracy often make systematic errors on important subsets (or slices) of data. Identifying underperforming slices is particularly challenging when working with high-dimensional inputs (e.g. images, audio), where important slices are often unlabeled. In order to address this issue, recent studies have proposed automated slice discovery methods (SDMs), which leverage learned model representations to mine input data for slices on which a model performs poorly. To be useful to a practitioner, these methods must identify slices that are both underperforming and coherent (i.e. united by a human-understandable concept). However, no quantitative evaluation framework currently exists for rigorously assessing SDMs with respect to these criteria. Additionally, prior qualitative evaluations have shown that SDMs often identify slices that are incoherent. In this work, we address these challenges by first designing a principled evaluation framework that enables a quantitative comparison of SDMs across 1,235 slice discovery settings in three input domains (natural images, medical images, and time-series data).
Then, motivated by the recent development of powerful cross-modal representation learning approaches, we present Domino, an SDM that leverages cross-modal embeddings and a novel error-aware mixture model to discover and describe coherent slices. We find that Domino accurately identifies 36% of the 1,235 slices in our framework -- a 12 percentage point improvement over prior methods. Further, Domino is the first SDM that can provide natural language descriptions of identified slices, correctly generating the exact name of the slice in 35% of settings. | Sabri Eyuboglu, Maya Varma, Khaled Kamal Saab, Jean-Benoit Delbrouck, Christopher Lee-Messer, Jared Dunnmon, James Zou, Christopher Re | null | null | 2,022 | iclr |
Shaping latent representations using Self-Organizing Maps with Relevance Learning | null | Recent work indicates that Deep Clustering (DC) methods are a viable option for unsupervised representations learning of visual features. By combining representation learning and clustering, traditional approaches have been shown to build latent representations that capture essential features of the data while preserving topological characteristics. In this sense, models based on Self-Organizing Maps models with relevance learning (SOMRL) were considered as they perform well in clustering besides being able to create a map that learns the relevance of each input dimension for each cluster, preserving the original relations and topology of the data. We hypothesize that this type of model can produce a more intuitive and disentangled representation in the latent space by promoting smoother transitions between cluster points over time. This work proposes a representation learning framework that combines a new gradient-based SOMRL model and autoencoders. The SOMRL learns the relevance weights for each input dimension of each cluster. It creates a tendency to separate the information into subspaces. To achieve this, we designed a new loss function term that weighs these learned relevances and provides an estimated unsupervised error to be used in combination with a reconstruction loss. The model is evaluated in terms of clustering performance and quality of the learned representations and then compared with start-of-the-art models, showing competitive results. | Pedro Braga, Heitor Medeiros, Hansenclever Bassani | null | null | 2,022 | iclr |
On Improving Adversarial Transferability of Vision Transformers | null | Vision transformers (ViTs) process input images as sequences of patches via self-attention; a radically different architecture than convolutional neural networks (CNNs). This makes it interesting to study the adversarial feature space of ViT models and their transferability. In particular, we observe that adversarial patterns found via conventional adversarial attacks show very \emph{low} black-box transferability even for large ViT models. We show that this phenomenon is only due to the sub-optimal attack procedures that do not leverage the true representation potential of ViTs. A deep ViT is composed of multiple blocks, with a consistent architecture comprising of self-attention and feed-forward layers, where each block is capable of independently producing a class token. Formulating an attack using only the last class token (conventional approach) does not directly leverage the discriminative information stored in the earlier tokens, leading to poor adversarial transferability of ViTs.Using the compositional nature of ViT models, we enhance transferability of existing attacks by introducing two novel strategies specific to the architecture of ViT models. \emph{(i) Self-Ensemble:} We propose a method to find multiple discriminative pathways by dissecting a single ViT model into an ensemble of networks. This allows explicitly utilizing class-specific information at each ViT block. \emph{(ii) Token Refinement:} We then propose to refine the tokens to further enhance the discriminative capacity at each block of ViT.Our token refinement systematically combines the class tokens with structural information preserved within the patch tokens. An adversarial attack when applied to such refined tokens within the ensemble of classifiers found in a single vision transformer has significantly higher transferability and thereby brings out the true generalization potential of the ViT's adversarial space. Code: https://t.ly/hBbW. | Muzammal Naseer, Kanchana Ranasinghe, Salman Khan, Fahad Khan, Fatih Porikli | null | null | 2,022 | iclr |
Generative Modeling for Multitask Visual Learning | null | Generative modeling has recently shown great promise in computer vision, but it has mostly focused on synthesizing visually realistic images. In this paper, motivated by multi-task learning of shareable feature representations, we consider a novel problem of learning a shared generative model that is useful across various visual perception tasks. Correspondingly, we propose a general multi-task oriented generative modeling (MGM) framework, by coupling a discriminative multi-task network with a generative network. While it is challenging to synthesize both RGB images and pixel-level annotations in multi-task scenarios, our framework enables us to use synthesized images paired with only weak annotations (i.e., image-level scene labels) to facilitate multiple visual tasks. Experimental evaluation on challenging multi-task benchmarks, including NYUv2 and Taskonomy, demonstrates that our MGM framework improves the performance of all the tasks by large margins, especially in the low-data regimes, and our model consistently outperforms state-of-the-art multi-task approaches. | Zhipeng Bao, Yu-Xiong Wang, Martial Hebert | null | null | 2,022 | iclr |
TAMP-S2GCNets: Coupling Time-Aware Multipersistence Knowledge Representation with Spatio-Supra Graph Convolutional Networks for Time-Series Forecasting | null | Graph Neural Networks (GNNs) are proven to be a powerful machinery for learning complex dependencies in multivariate spatio-temporal processes. However, most existing GNNs have inherently static architectures, and as a result, do not explicitly account for time dependencies of the encoded knowledge and are limited in their ability to simultaneously infer latent time-conditioned relations among entities. We postulate that such hidden time-conditioned properties may be captured by the tools of multipersistence, i.e, a emerging machinery in topological data analysis which allows us to quantify dynamics of the data shape along multiple geometric dimensions.
We make the first step toward integrating the two rising research directions, that is, time-aware deep learning and multipersistence, and propose a new model, Time-Aware Multipersistence Spatio-Supra Graph Convolutional Network (TAMP-S2GCNets). We summarize inherent time-conditioned topological properties of the data as time-aware multipersistence Euler-Poincar\'e surface and prove its stability. We then construct a supragraph convolution module which simultaneously accounts for the extracted intra- and inter- spatio-temporal dependencies in the data. Our extensive experiments on highway traffic flow, Ethereum token prices, and COVID-19 hospitalizations demonstrate that TAMP-S2GCNets outperforms the state-of-the-art tools in multivariate time series forecasting tasks. | Yuzhou Chen, Ignacio Segovia-Dominguez, Baris Coskunuzer, Yulia Gel | null | null | 2,022 | iclr |
The MultiBERTs: BERT Reproductions for Robustness Analysis | null | Experiments with pre-trained models such as BERT are often based on a single checkpoint. While the conclusions drawn apply to the artifact tested in the experiment (i.e., the particular instance of the model), it is not always clear whether they hold for the more general procedure which includes the architecture, training data, initialization scheme, and loss function. Recent work has shown that repeating the pre-training process can lead to substantially different performance, suggesting that an alternative strategy is needed to make principled statements about procedures. To enable researchers to draw more robust conclusions, we introduce MultiBERTs, a set of 25 BERT-Base checkpoints, trained with similar hyper-parameters as the original BERT model but differing in random weight initialization and shuffling of training data. We also define the Multi-Bootstrap, a non-parametric bootstrap method for statistical inference designed for settings where there are multiple pre-trained models and limited test data. To illustrate our approach, we present a case study of gender bias in coreference resolution, in which the Multi-Bootstrap lets us measure effects that may not be detected with a single checkpoint. The models and statistical library are available online, along with an additional set of 140 intermediate checkpoints captured during pre-training to facilitate research on learning dynamics. | Thibault Sellam, Steve Yadlowsky, Ian Tenney, Jason Wei, Naomi Saphra, Alexander D'Amour, Tal Linzen, Jasmijn Bastings, Iulia Raluca Turc, Jacob Eisenstein, Dipanjan Das, Ellie Pavlick | null | null | 2,022 | iclr |
Independent SE(3)-Equivariant Models for End-to-End Rigid Protein Docking | null | Protein complex formation is a central problem in biology, being involved in most of the cell's processes, and essential for applications, e.g. drug design or protein engineering. We tackle rigid body protein-protein docking, i.e., computationally predicting the 3D structure of a protein-protein complex from the individual unbound structures, assuming no conformational change within the proteins happens during binding. We design a novel pairwise-independent SE(3)-equivariant graph matching network to predict the rotation and translation to place one of the proteins at the right docked position relative to the second protein. We mathematically guarantee a basic principle: the predicted complex is always identical regardless of the initial locations and orientations of the two structures. Our model, named EquiDock, approximates the binding pockets and predicts the docking poses using keypoint matching and alignment, achieved through optimal transport and a differentiable Kabsch algorithm. Empirically, we achieve significant running time improvements and often outperform existing docking software despite not relying on heavy candidate sampling, structure refinement, or templates. | Octavian-Eugen Ganea, Xinyuan Huang, Charlotte Bunne, Yatao Bian, Regina Barzilay, Tommi S. Jaakkola, Andreas Krause | null | null | 2,022 | iclr |
Out-of-distribution Generalization in the Presence of Nuisance-Induced Spurious Correlations | null | In many prediction problems, spurious correlations are induced by a changing relationship between the label and a nuisance variable that is also correlated with the covariates. For example, in classifying animals in natural images, the background, which is a nuisance, can predict the type of animal. This nuisance-label relationship does not always hold, and the performance of a model trained under one such relationship may be poor on data with a different nuisance-label relationship. To build predictive models that perform well regardless of the nuisance-label relationship, we develop Nuisance-Randomized Distillation (NURD). We introduce the nuisance-randomized distribution, a distribution where the nuisance and the label are independent. Under this distribution, we define the set of representations such that conditioning on any member, the nuisance and the label remain independent. We prove that the representations in this set always perform better than chance, while representations outside of this set may not. NURD finds a representation from this set that is most informative of the label under the nuisance-randomized distribution, and we prove that this representation achieves the highest performance regardless of the nuisance-label relationship. We evaluate NURD on several tasks including chest X-ray classification where, using non-lung patches as the nuisance, NURD produces models that predict pneumonia under strong spurious correlations. | Aahlad Manas Puli, Lily H Zhang, Eric Karl Oermann, Rajesh Ranganath | null | null | 2,022 | iclr |
Pareto Frontier Approximation Network (PA-Net) Applied to Multi-objective TSP | null | Multi-objective optimization is used in various areas of robotics like control, planning etc. Their solutions are dependent on multiple objective functions, which can be conflicting in nature. In such cases, the optimality is defined in terms of Pareto optimality. A set of these Pareto Optimal solutions in the objective space form a Pareto front (or frontier). Each solution has its own trade off. For instance, the travelling salesman problem (TSP) is used in robotics for task/resource allocation. Often this allocation is influenced by multiple objective functions and is solved using Multi-objective travelling salesman problem (MOTSP). In this work, we present PA-Net, a network that generates good approximations of the Pareto front for the multi-objective optimization problems. Our training framework is applicable to other multi-objective optimization problems; however, in this work, we focus on solving MOTSP. Firstly, MOTSP is converted into a constrained optimization problem. We then train our network to solve this constrained problem using the Lagrangian relaxation and policy gradient. With PA-Net we are able to generate better quality Pareto fronts with fast inference times as compared to other learning based and classical methods. Finally, we present the application of PA-Net to find optimal visiting order in coverage planning. | Ishaan Mehta, Sajad Saeedi | null | null | 2,022 | iclr |
Learning Strides in Convolutional Neural Networks | null | 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 cross-validation 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. | Rachid Riad, Olivier Teboul, David Grangier, Neil Zeghidour | null | null | 2,022 | iclr |
GNN-LM: Language Modeling based on Global Contexts via GNN | null | Inspired by the notion that "it to copy is easier than to memorize", in this work, we introduce GNN-LM, which extends vanilla neural language model (LM) by allowing to reference similar contexts in the entire training corpus. We build a directed heterogeneous graph between an input context and its semantically related neighbors selected from the training corpus, where nodes are tokens in the input context and retrieved neighbor contexts, and edges represent connections between nodes. Graph neural networks (GNNs) are constructed upon the graph to aggregate information from similar contexts to decode the token. This learning paradigm provides direct access to the reference contexts and helps improve a model's generalization ability. We conduct comprehensive experiments to validate the effectiveness of the GNN-LM: GNN-LM achieves a new state-of-the-art perplexity of 14.8 on WikiText-103 (a 3.9 point improvement over its counterpart of the vanilla LM model), and shows substantial improvement on One Billion Word and Enwiki8 datasets against strong baselines. In-depth ablation studies are performed to understand the mechanics of GNN-LM. The code can be found at https://github.com/ShannonAI/GNN-LM. | Yuxian Meng, Shi Zong, Xiaoya Li, Xiaofei Sun, Tianwei Zhang, Fei Wu, Jiwei Li | null | null | 2,022 | iclr |
Few-Shot Multi-task Learning via Implicit regularization | null | Modern machine learning is highly data-intensive. Few-shot learning (FSL) aims to resolve this sample efficiency problem by learning from multiple tasks and quickly adapt to new tasks containing only a few samples. However, FSL problems proves to be significantly more challenging and require more compute expensive process to optimize. In this work, we consider multi-task linear regression (MTLR) as a canonical problem for few-shot learning, and investigate the source of challenge of FSL. We find that the MTLR exhibits local minimum problems that are not present in single-task problem, and thus making the learning much more challenging. We also show that the problem can be resolved by overparameterizing the model by increasing both the width and depth of the linear network and initializing the weights with small values, exploiting the implicit regularization bias of gradient descent-based learning. | Dongsung Huh | null | null | 2,022 | iclr |
Complete Verification via Multi-Neuron Relaxation Guided Branch-and-Bound | null | State-of-the-art neural network verifiers are fundamentally based on one of two paradigms: either encoding the whole verification problem via tight multi-neuron convex relaxations or applying a Branch-and-Bound (BaB) procedure leveraging imprecise but fast bounding methods on a large number of easier subproblems. The former can capture complex multi-neuron dependencies but sacrifices completeness due to the inherent limitations of convex relaxations. The latter enables complete verification but becomes increasingly ineffective on larger and more challenging networks. In this work, we present a novel complete verifier which combines the strengths of both paradigms: it leverages multi-neuron relaxations to drastically reduce the number of subproblems generated during the BaB process and an efficient GPU-based dual optimizer to solve the remaining ones. An extensive evaluation demonstrates that our verifier achieves a new state-of-the-art on both established benchmarks as well as networks with significantly higher accuracy than previously considered. The latter result (up to 28% certification gains) indicates meaningful progress towards creating verifiers that can handle practically relevant networks. | Claudio Ferrari, Mark Niklas Mueller, Nikola Jovanović, Martin Vechev | null | null | 2,022 | iclr |
Spatiotemporal Representation Learning on Time Series with Dynamic Graph ODEs | null | Spatiotemporal representation learning on multivariate time series has received tremendous attention in forecasting traffic and energy data. Recent works either rely on complicated discrete neural architectures or graph priors, hindering their effectiveness and applications in the real world. In this paper, inspired by neural ordinary differential equations and graph structure learning, we propose a fully continuous model named Dynamic Graph ODE (DyG-ODE) to capture both long-range spatial and temporal dependencies to learn expressive representations on arbitrary multivariate time series data without being restricted by rigid preconditions (e.g., graph priors). For modeling the continuous dynamics of spatiotemporal clues, we design a simple yet powerful dynamic graph ODE by coupling the proposed spatial and temporal ODEs, which not only allows the model to obtain infinite spatial and temporal receptive fields but also reduces numerical errors and model complexity significantly. Our empirical evaluations demonstrate the superior effectiveness and efficiency of DyG-ODE on a number of benchmark datasets. | Ming Jin, Yuan-Fang Li, Yu Zheng, Bin Yang, Shirui Pan | null | null | 2,022 | iclr |
On Bridging Generic and Personalized Federated Learning for Image Classification | null | Federated learning is promising for its capability to collaboratively train models with multiple clients without accessing their data, but vulnerable when clients' data distributions diverge from each other. This divergence further leads to a dilemma: "Should we prioritize the learned model's generic performance (for future use at the server) or its personalized performance (for each client)?" These two, seemingly competing goals have divided the community to focus on one or the other, yet in this paper we show that it is possible to approach both at the same time. Concretely, we propose a novel federated learning framework that explicitly decouples a model's dual duties with two prediction tasks. On the one hand, we introduce a family of losses that are robust to non-identical class distributions, enabling clients to train a generic predictor with a consistent objective across them. On the other hand, we formulate the personalized predictor as a lightweight adaptive module that is learned to minimize each client's empirical risk on top of the generic predictor. With this two-loss, two-predictor framework which we name Federated Robust Decoupling (Fed-RoD), the learned model can simultaneously achieve state-of-the-art generic and personalized performance, essentially bridging the two tasks. | Hong-You Chen, Wei-Lun Chao | null | null | 2,022 | iclr |
Stochastic Projective Splitting: Solving Saddle-Point Problems with Multiple Regularizers | null | We present a new, stochastic variant of the projective splitting (PS) family of algorithms for monotone inclusion problems. It can solve min-max and noncooperative game formulations arising in applications such as robust ML without the convergence issues associated with gradient descent-ascent, the current de facto standard approach in ML applications. Our proposal is the first version of PS able to use stochastic gradient oracles. It can solve min-max games while handling multiple constraints and nonsmooth regularizers via projection and proximal operators. Unlike other stochastic splitting methods that can solve such problems, our method does not rely on a product-space reformulation of the original problem. We prove almost-sure convergence of the iterates to the solution and a convergence rate for the expected residual. By working with monotone inclusions rather than variational inequalities, our analysis avoids the drawbacks of measuring convergence through the restricted gap function. We close with numerical experiments on a distributionally robust sparse logistic regression problem. | Patrick R. Johnstone, Jonathan Eckstein, Thomas Flynn, Shinjae Yoo | null | null | 2,022 | iclr |
Value Function Spaces: Skill-Centric State Abstractions for Long-Horizon Reasoning | null | Reinforcement learning can train policies that effectively perform complex tasks. However for long-horizon tasks, the performance of these methods degrades with horizon, often necessitating reasoning over and chaining lower-level skills. Hierarchical reinforcement learning aims to enable this by providing a bank of low-level skills as action abstractions. Hierarchies can further improve on this by abstracting the space states as well. We posit that a suitable state abstraction should depend on the capabilities of the available lower-level policies. We propose Value Function Spaces: a simple approach that produces such a representation by using the value functions corresponding to each lower-level skill. These value functions capture the affordances of the scene, thus forming a representation that compactly abstracts task relevant information and robustly ignores distractors. Empirical evaluations for maze-solving and robotic manipulation tasks demonstrate that our approach improves long-horizon performance and enables better zero-shot generalization than alternative model-free and model-based methods. | Dhruv Shah, Peng Xu, Yao Lu, Ted Xiao, Alexander T Toshev, Sergey Levine, brian ichter | null | null | 2,022 | iclr |
DR3: Value-Based Deep Reinforcement Learning Requires Explicit Regularization | null | Despite overparameterization, deep networks trained via supervised learning are surprisingly easy to optimize and exhibit excellent generalization. One hypothesis to explain this is that overparameterized deep networks enjoy the benefits of implicit regularization induced by stochastic gradient descent, which favors parsimonious solutions that generalize well on test inputs. It is reasonable to surmise that deep reinforcement learning (RL) methods could also benefit from this effect. In this paper, we discuss how the implicit regularization effect of SGD seen in supervised learning could in fact be harmful in the offline deep RL setting, leading to poor generalization and degenerate feature representations. Our theoretical analysis shows that when existing models of implicit regularization are applied to temporal difference learning, the resulting derived regularizer favors degenerate solutions with excessive aliasing, in stark contrast to the supervised learning case. We back up these findings empirically, showing that feature representations learned by a deep network value function trained via bootstrapping can indeed become degenerate, aliasing the representations for state-action pairs that appear on either side of the Bellman backup. To address this issue, we derive the form of this implicit regularizer and, inspired by this derivation, propose a simple and effective explicit regularizer, called DR3, that counteracts the undesirable effects of this implicit regularizer. When combined with existing offline RL methods, DR3 substantially improves performance and stability, alleviating unlearning in Atari 2600 games, D4RL domains and robotic manipulation from images. | Aviral Kumar, Rishabh Agarwal, Tengyu Ma, Aaron Courville, George Tucker, Sergey Levine | null | null | 2,022 | iclr |
AEVA: Black-box Backdoor Detection Using Adversarial Extreme Value Analysis | null | Deep neural networks (DNNs) are proved to be vulnerable against backdoor attacks. A backdoor could be embedded in the target DNNs through injecting a backdoor trigger into the training examples, which can cause the target DNNs misclassify an input attached with the backdoor trigger. Recent backdoor detection methods often require the access to the original poisoned training data, the parameters of the target DNNs, or the predictive confidence for each given input, which are impractical in many real-world applications, e.g., on-device de-ployed DNNs. We address the black-box hard-label backdoor detection problem where the DNN is a fully black-box and only its final output label is accessible. We approach this problem from the optimization perspective and show that the objective of backdoor detection is bounded by an adversarial objective. Further theoretical and empirical studies reveal that this adversarial objective leads to a solution with highly skewed distribution; a singularity is often observed in the adversarial map of a backdoor-infected example, which we call the adversarial singularity phenomenon. Based on this observation, we propose the adversarial extreme value analysis(AEVA) algorithm to detect backdoors in black-box neural networks. The AEVA algorithm is based on an extreme value analysis on the adversarial map, computed from the monte-carlo gradient estimation due to the black-box hard-label constraint. Evidenced by extensive experiments across three popular tasks and backdoor attacks, our approach is shown effective in detecting backdoor attacks under the black-box hard-label scenarios | Junfeng Guo, Ang Li, Cong Liu | null | null | 2,022 | iclr |
Metric Learning on Temporal Graphs via Few-Shot Examples | null | Graph metric learning methods aim to learn the distance metric over graphs such that similar graphs are closer and dissimilar graphs are farther apart. This is of critical importance in many graph classification applications such as drug discovery and epidemics categorization. In many real-world applications, the graphs are typically evolving over time; labeling graph data is usually expensive and also requires background knowledge. However, state-of-the-art graph metric learning techniques consider the input graph as static, and largely ignore the intrinsic dynamics of temporal graphs; Furthermore, most of these techniques require abundant labeled examples for training in the representation learning process. To address the two aforementioned problems, we wish to learn a distance metric only over fewer temporal graphs, which metric could not only help accurately categorize seen temporal graphs but also be adapted smoothly to unseen temporal graphs. In this paper, we first propose the streaming-snapshot model to describe temporal graphs on different time scales. Then we propose the MetaTag framework: 1) to learn the metric over a limited number of streaming-snapshot modeled temporal graphs, 2) and adapt the learned metric to unseen temporal graphs via a few examples. Finally, we demonstrate the performance of MetaTag in comparison with state-of-the-art algorithms for temporal graph classification problems. | Dongqi Fu, Liri Fang, Ross Maciejewski, Vetle I Torvik, Jingrui He | null | null | 2,022 | iclr |
Enhanced neural network regularization with macro-block dropout | null | This paper proposes a new regularization algorithm referred to as macro-block dropout. The overfitting issue has been a difficult problem in training large network models. The dropout technique has proven to be simple yet very effective for regularization by preventing complex co-adaptations on training data. In this work, we observe that in the hidden outputs, the correlations between geometrically close elements are usually stronger than those between distant elements. Motivated by this observation, we define a macro-block that contains multiple elements of the hidden output layer in order to reduce co-adaptations more effectively. Rather than applying dropout to each element, we apply random dropout to each macro-block. In our experiments with image classification tasks on the MNIST and the ImageNet datasets as well as a speech recognition task on the LibriSpeech set, this simple algorithm has shown a quite significant improvement over the conventional dropout approach | Chanwoo Kim | null | null | 2,022 | iclr |
Expressiveness and Approximation Properties of Graph Neural Networks | null | Characterizing the separation power of graph neural networks (GNNs) provides an understanding of their limitations for graph learning tasks. Results regarding separation power are, however, usually geared at specific GNNs architectures, and tools for understanding arbitrary GNN architectures are generally lacking. We provide an elegant way to easily obtain bounds on the separation power of GNNs in terms of the Weisfeiler-Leman (WL) tests, which have become the yardstick to measure the separation power of GNNs. The crux is to view GNNs as expressions in a procedural tensor language describing the computations in the layers of the GNNs. Then, by a simple analysis of the obtained expressions, in terms of the number of indexes used and the nesting depth of summations, bounds on the separation power in terms of the WL-tests readily follow. We use tensor language to define Higher-Order Message-Passing Neural Networks (or k-MPNNs), a natural extension of MPNNs. Furthermore, the tensor language point of view allows for the derivation of universality results for classes of GNNs in a natural way. Our approach provides a toolbox with which GNN architecture designers can analyze the separation power of their GNNs, without needing to know the intricacies of the WL-tests. We also provide insights in what is needed to boost the separation power of GNNs. | Floris Geerts, Juan L Reutter | null | null | 2,022 | iclr |
Provably Filtering Exogenous Distractors using Multistep Inverse Dynamics | null | Many real-world applications of reinforcement learning (RL) require the agent to deal with high-dimensional observations such as those generated from a megapixel camera. Prior work has addressed such problems with representation learning, through which the agent can provably extract endogenous, latent state information from raw observations and subsequently plan efficiently. However, such approaches can fail in the presence of temporally correlated noise in the observations, a phenomenon that is common in practice. We initiate the formal study of latent state discovery in the presence of such exogenous noise sources by proposing a new model, the Exogenous Block MDP (EX-BMDP), for rich observation RL. We start by establishing several negative results, by highlighting failure cases of prior representation learning based approaches. Then, we introduce the Predictive Path Elimination (PPE) algorithm, that learns a generalization of inverse dynamics and is provably sample and computationally efficient in EX-BMDPs when the endogenous state dynamics are near deterministic. The sample complexity of PPE depends polynomially on the size of the latent endogenous state space while not directly depending on the size of the observation space, nor the exogenous state space. We provide experiments on challenging exploration problems which show that our approach works empirically. | Yonathan Efroni, Dipendra Misra, Akshay Krishnamurthy, Alekh Agarwal, John Langford | null | null | 2,022 | iclr |
Scalable multimodal variational autoencoders with surrogate joint posterior | null | To obtain a joint representation from multimodal data in variational autoencoders (VAEs), it is important to infer the representation from arbitrary subsets of modalities after learning. A scalable way to achieve this is to aggregate the inferences of each modality as experts. A state-of-the-art approach to learning this aggregation of experts is to encourage all modalities to be reconstructed and cross-generated from arbitrary subsets. However, this learning may be insufficient if cross-generation is difficult. Furthermore, to evaluate its objective function, exponential generation paths concerning the number of modalities are required. To alleviate these problems, we propose to explicitly minimize the divergence between inferences from arbitrary subsets and the surrogate joint posterior that approximates the true joint posterior. We also proposed using a gradient origin network, a deep generative model that learns inferences without using an inference network, thereby reducing the need for additional parameters by introducing the surrogate posterior. We demonstrate that our method performs better than existing scalable multimodal VAEs in inference and generation.
| Masahiro Suzuki, Yutaka Matsuo | null | null | 2,022 | iclr |
Dynamics-Aware Comparison of Learned Reward Functions | null | The ability to learn reward functions plays an important role in enabling the deployment of intelligent agents in the real world. However, $\textit{comparing}$ reward functions, for example as a means of evaluating reward learning methods, presents a challenge. Reward functions are typically compared by considering the behavior of optimized policies, but this approach conflates deficiencies in the reward function with those of the policy search algorithm used to optimize it. To address this challenge, Gleave et al. (2020) propose the Equivalent-Policy Invariant Comparison (EPIC) distance. EPIC avoids policy optimization, but in doing so requires computing reward values at transitions that may be impossible under the system dynamics. This is problematic for learned reward functions because it entails evaluating them outside of their training distribution, resulting in inaccurate reward values that we show can render EPIC ineffective at comparing rewards. To address this problem, we propose the Dynamics-Aware Reward Distance (DARD), a new reward pseudometric. DARD uses an approximate transition model of the environment to transform reward functions into a form that allows for comparisons that are invariant to reward shaping while only evaluating reward functions on transitions close to their training distribution. Experiments in simulated physical domains demonstrate that DARD enables reliable reward comparisons without policy optimization and is significantly more predictive than baseline methods of downstream policy performance when dealing with learned reward functions. | Blake Wulfe, Logan Michael Ellis, Jean Mercat, Rowan Thomas McAllister, Adrien Gaidon | null | null | 2,022 | iclr |
Fine-Tuning can Distort Pretrained Features and Underperform Out-of-Distribution | null | When transferring a pretrained model to a downstream task, two popular methods are full fine-tuning (updating all the model parameters) and linear probing (updating only the last linear layer---the "head"). It is well known that fine-tuning leads to better accuracy in-distribution (ID). However, in this paper, we find that fine-tuning can achieve worse accuracy than linear probing out-of-distribution (OOD) when the pretrained features are good and the distribution shift is large. On 10 distribution shift datasets (BREEDS-Living17, BREEDS-Entity30, DomainNet, CIFAR $\to$ STL, CIFAR-10.1, FMoW, ImageNetV2, ImageNet-R, ImageNet-A, ImageNet-Sketch), fine-tuning obtains on average 2% higher accuracy ID but 7% lower accuracy OOD than linear probing. We show theoretically that this tradeoff between ID and OOD accuracy arises even in a simple setting: fine-tuning overparameterized two-layer linear networks. We prove that the OOD error of fine-tuning is high when we initialize with a fixed or random head---this is because while fine-tuning learns the head, the lower layers of the neural network change simultaneously and distort the pretrained features. Our analysis suggests that the easy two-step strategy of linear probing then full fine-tuning (LP-FT), sometimes used as a fine-tuning heuristic, combines the benefits of both fine-tuning and linear probing. Empirically, LP-FT outperforms both fine-tuning and linear probing on the above datasets (1% better ID, 10% better OOD than full fine-tuning). | Ananya Kumar, Aditi Raghunathan, Robbie Matthew Jones, Tengyu Ma, Percy Liang | null | null | 2,022 | iclr |
PI3NN: Out-of-distribution-aware Prediction Intervals from Three Neural Networks | null | We propose a novel prediction interval (PI) method for uncertainty quantification, which addresses three major issues with the state-of-the-art PI methods. First, existing PI methods require retraining of neural networks (NNs) for every given confidence level and suffer from the crossing issue in calculating multiple PIs. Second, they usually rely on customized loss functions with extra sensitive hyperparameters for which fine tuning is required to achieve a well-calibrated PI. Third, they usually underestimate uncertainties of out-of-distribution (OOD) samples leading to over-confident PIs. Our PI3NN method calculates PIs from linear combinations of three NNs, each of which is independently trained using the standard mean squared error loss. The coefficients of the linear combinations are computed using root-finding algorithms to ensure tight PIs for a given confidence level. We theoretically prove that PI3NN can calculate PIs for a series of confidence levels without retraining NNs and it completely avoids the crossing issue. Additionally, PI3NN does not introduce any unusual hyperparameters resulting in a stable performance. Furthermore, we address OOD identification challenge by introducing an initialization scheme which provides reasonably larger PIs of the OOD samples than those of the in-distribution samples. Benchmark and real-world experiments show that our method outperforms several state-of-the-art approaches with respect to predictive uncertainty quality, robustness, and OOD samples identification. | Siyan Liu, Pei Zhang, Dan Lu, Guannan Zhang | null | null | 2,022 | iclr |
Measuring the Effectiveness of Self-Supervised Learning using Calibrated Learning Curves | null | Self-supervised learning has witnessed remarkable progress in recent years, in particular with the introduction of augmentation-based contrastive methods. While a number of large-scale empirical studies on the performance of self-supervised pre-training have been conducted, there isn't yet an agreed upon set of control baselines, evaluation practices, and metrics to report. We identify this as an important angle of investigation and propose an evaluation standard that aims to quantify and communicate transfer learning performance in an informative yet accessible setup. This is done by baking in a number of key control baselines in the evaluation method, particularly the blind guess (quantifying the dataset bias), the scratch model (quantifying the architectural contribution), and the gold standard (quantifying the upper-bound). We further provide a number of experiments to demonstrate how the proposed evaluation can be employed in empirical studies of basic questions -- for example, whether the effectiveness of existing self-supervised learning methods is skewed towards image classification versus other tasks, such as dense pixel-wise predictions.
| Andrei Atanov, Shijian Xu, Onur Beker, Andrey Filatov, Amir Zamir | null | null | 2,022 | iclr |
CareGraph: A Graph-based Recommender System for Diabetes Self-Care | null | In this work, we build a knowledge graph that captures key attributes of content and notifications in a digital health platform for diabetes management. We propose a Deep Neural Network-based recommender that uses the knowledge graph embeddings to recommend health nudges for maximizing engagement by combating the cold-start and sparsity problems. We use a leave-one-out approach to evaluate the model. We compare the proposed model performance with a text similarity and Deep-and-Cross Network-based approach as the baseline. The overall improvement in Click-Through-Rate prediction AUC for the Knowledge-Graph-based model was 11%. We also observe that our model improved the average AUC by 5% in cold-start situations. | Sirinart Tangruamsub, Karthik Kappaganthu, John O'Donovan, Anmol Madan | null | null | 2,022 | iclr |
Sound Adversarial Audio-Visual Navigation | null | Audio-visual navigation task requires an agent to find a sound source in a realistic, unmapped 3D environment by utilizing egocentric audio-visual observations. Existing audio-visual navigation works assume a clean environment that solely contains the target sound, which, however, would not be suitable in most real-world applications due to the unexpected sound noise or intentional interference. In this work, we design an acoustically complex environment in which, besides the target sound, there exists a sound attacker playing a zero-sum game with the agent. More specifically, the attacker can move and change the volume and category of the sound to make the agent suffer from finding the sounding object while the agent tries to dodge the attack and navigate to the goal under the intervention. Under certain constraints to the attacker, we can improve the robustness of the agent towards unexpected sound attacks in audio-visual navigation. For better convergence, we develop a joint training mechanism by employing the property of a centralized critic with decentralized actors. Experiments on two real-world 3D scan datasets, Replica, and Matterport3D, verify the effectiveness and the robustness of the agent trained under our designed environment when transferred to the clean environment or the one containing sound attackers with random policy. Project: https://yyf17.github.io/SAAVN . | Yinfeng Yu, Wenbing Huang, Fuchun Sun, Changan Chen, Yikai Wang, Xiaohong Liu | null | null | 2,022 | iclr |
Error-based or target-based? A unifying framework for learning in recurrent spiking networks | null | Learning in biological or artificial networks means changing the laws governing the network dynamics in order to better behave in a specific situation. In the field of supervised learning, two complementary approaches stand out: error-based and target-based learning. However, there exists no consensus on which is better suited for which task, and what is the most biologically plausible. Here we propose a comprehensive theoretical framework that includes these two frameworks as special cases. This novel theoretical formulation offers major insights into the differences between the two approaches. In particular, we show how target-based naturally emerges from error-based when the number of constraints on the target dynamics, and as a consequence on the internal network dynamics, is comparable to the degrees of freedom of the network. Moreover, given the experimental evidences on the relevance that spikes have in biological networks, we investigate the role of coding with specific patterns of spikes by introducing a parameter that defines the tolerance to precise spike timing during learning. Our approach naturally lends itself to Imitation Learning (and Behavioral Cloning in particular) and we apply it to solve relevant closed-loop tasks such as the button-and-food task, and the 2D Bipedal Walker. We show that a high dimensionality feedback structure is extremely important when it is necessary to solve a task that requires retaining memory for a long time (button-and-food). On the other hand, we find that coding with specific patterns of spikes enables optimal performances in a motor task (the 2D Bipedal Walker). Finally, we show that our theoretical formulation suggests protocols to deduce the structure of learning feedback in biological networks. | Cristiano Capone, Paolo Muratore, Pier Stanislao Paolucci | null | null | 2,022 | iclr |
Optimization and Adaptive Generalization of Three layer Neural Networks | null | While there has been substantial recent work studying generalization of neural networks,
the ability of deep nets in automating the process of feature extraction still evades a thorough mathematical understanding.
As a step toward this goal, we analyze learning and generalization of a three-layer neural network with ReLU activations in a regime that goes beyond the linear approximation of the network, and is hence not captured by the common Neural Tangent Kernel. We show that despite nonconvexity of the empirical loss, a variant of SGD converges in polynomially many iterations to a good solution that generalizes. In particular, our generalization bounds are adaptive: they automatically optimize over a family of kernels that includes the Neural Tangent Kernel, to provide the tightest bound. | Khashayar Gatmiry, Stefanie Jegelka, Jonathan Kelner | null | null | 2,022 | iclr |
Value Gradient weighted Model-Based Reinforcement Learning | null | Model-based reinforcement learning (MBRL) is a sample efficient technique to obtain control policies, yet unavoidable modeling errors often lead performance deterioration. The model in MBRL is often solely fitted to reconstruct dynamics, state observations in particular, while the impact of model error on the policy is not captured by the training objective. This leads to a mismatch between the intended goal of MBRL, enabling good policy and value learning, and the target of the loss function employed in practice, future state prediction. Naive intuition would suggest that value-aware model learning would fix this problem and, indeed, several solutions to this objective mismatch problem have been proposed based on theoretical analysis. However, they tend to be inferior in practice to commonly used maximum likelihood (MLE) based approaches. In this paper we propose the Value-gradient weighted Model Learning (VaGraM), a novel method for value-aware model learning which improves the performance of MBRL in challenging settings, such as small model capacity and the presence of distracting state dimensions. We analyze both MLE and value-aware approaches and demonstrate how they fail to account for exploration and the behavior of function approximation when learning value-aware models and highlight the additional goals that must be met to stabilize optimization in the deep learning setting. We verify our analysis by showing that our loss function is able to achieve high returns on the Mujoco benchmark suite while being more robust than maximum likelihood based approaches.
| Claas A Voelcker, Victor Liao, Animesh Garg, Amir-massoud Farahmand | null | null | 2,022 | iclr |
MIDI-DDSP: Detailed Control of Musical Performance via Hierarchical Modeling | null | Musical expression requires control of both what notes that are played, and how they are performed. Conventional audio synthesizers provide detailed expressive controls, but at the cost of realism. Black-box neural audio synthesis and concatenative samplers can produce realistic audio, but have few mechanisms for control. In this work, we introduce MIDI-DDSP a hierarchical model of musical instruments that enables both realistic neural audio synthesis and detailed user control. Starting from interpretable Differentiable Digital Signal Processing (DDSP) synthesis parameters, we infer musical notes and high-level properties of their expressive performance (such as timbre, vibrato, dynamics, and articulation). This creates a 3-level hierarchy (notes, performance, synthesis) that affords individuals the option to intervene at each level, or utilize trained priors (performance given notes, synthesis given performance) for creative assistance. Through quantitative experiments and listening tests, we demonstrate that this hierarchy can reconstruct high-fidelity audio, accurately predict performance attributes for a note sequence, independently manipulate the attributes of a given performance, and as a complete system, generate realistic audio from a novel note sequence. By utilizing an interpretable hierarchy, with multiple levels of granularity, MIDI-DDSP opens the door to assistive tools to empower individuals across a diverse range of musical experience. | Yusong Wu, Ethan Manilow, Yi Deng, Rigel Swavely, Kyle Kastner, Tim Cooijmans, Aaron Courville, Cheng-Zhi Anna Huang, Jesse Engel | null | null | 2,022 | iclr |
On the Convergence of Certified Robust Training with Interval Bound Propagation | null | Interval Bound Propagation (IBP) is so far the base of state-of-the-art methods for training neural networks with certifiable robustness guarantees when potential adversarial perturbations present, while the convergence of IBP training remains unknown in existing literature. In this paper, we present a theoretical analysis on the convergence of IBP training. With an overparameterized assumption, we analyze the convergence of IBP robust training. We show that when using IBP training to train a randomly initialized two-layer ReLU neural network with logistic loss, gradient descent can linearly converge to zero robust training error with a high probability if we have sufficiently small perturbation radius and large network width. | Yihan Wang, Zhouxing Shi, Quanquan Gu, Cho-Jui Hsieh | null | null | 2,022 | iclr |
Deep Dynamic Attention Model with Gate Mechanism for Solving Time-dependent Vehicle Routing Problems | null | Vehicle routing problems (VRPs) are a type of classical combinatorial optimization problems widely existing in logistics and transportation operations. There has been an increasing interest to use deep reinforcement learning (DRL) techniques to tackle VRPs, and previous DRL-based studies assumed time-independent travel times between customers. However, travel times in real-world road networks are time-varying, which need to be considered in practical VRPs. We thus propose a Deep Dynamic Attention Models with Gate Mechanisms (DDAM-GM) to learn heuristics for time-dependent VRPs (TDVRPs) in real-world road networks. It extracts the information of node location, node demand, and time-varying travel times between nodes to obtain enhanced node embeddings through a dimension-reducing MHA layer and a synchronous encoder. In addition, we use a gate mechanism to obtain better context embedding. On the basis of a 110-day travel time dataset with 240 time periods per day from an urban road network with 408 nodes and 1250 directed links, we conduct a series of experiments to validate the effectiveness of the proposed model on TDVRPs without and with consideration of time windows, respectively. Experimental results show that our model outperforms significantly two state-of-the-art DRL-based models. | Feng Guo, Qu Wei, Miao Wang, Zhaoxia Guo | null | null | 2,022 | iclr |
Back2Future: Leveraging Backfill Dynamics for Improving Real-time Predictions in Future | null | For real-time forecasting in domains like public health and macroeconomics, data collection is a non-trivial and demanding task. Often after being initially released, it undergoes several revisions later (maybe due to human or technical constraints) - as a result, it may take weeks until the data reaches a stable value. This so-called ‘backfill’ phenomenon and its effect on model performance have been barely addressed in the prior literature. In this paper, we introduce the multi-variate backfill problem using COVID-19 as the motivating example.
We construct a detailed dataset composed of relevant signals over the past year of the pandemic.
We then systematically characterize several patterns in backfill dynamics and leverage our observations for formulating a novel problem and neural framework, Back2Future, that aims to refines a given model's predictions in real-time. Our extensive experiments demonstrate that our method refines the performance of the diverse set of top models for COVID-19 forecasting and GDP growth forecasting. Specifically, we show that Back2Future refined top COVID-19 models by 6.65% to 11.24% and yield an 18% improvement over non-trivial baselines. In addition, we show that our model improves model evaluation too; hence policy-makers can better understand the true accuracy of forecasting models in real-time. | Harshavardhan Kamarthi, Alexander Rodríguez, B. Aditya Prakash | null | null | 2,022 | iclr |
Continual Learning with Filter Atom Swapping | null | Continual learning has been widely studied in recent years to resolve the catastrophic forgetting of deep neural networks. In this paper, we first enforce a low-rank filter subspace by decomposing convolutional filters within each network layer over a small set of filter atoms. Then, we perform continual learning with filter atom swapping. In other words, we learn for each task a new filter subspace for each convolutional layer, i.e., hundreds of parameters as filter atoms, but keep subspace coefficients shared across tasks. By maintaining a small footprint memory of filter atoms, we can easily archive models for past tasks to avoid forgetting. The effectiveness of this simple scheme for continual learning is illustrated both empirically and theoretically. The proposed atom swapping framework further enables flexible and efficient model ensemble with members selected within a task or across tasks to improve the performance in different continual learning settings. Being validated on multiple benchmark datasets with different convolutional network structures, the proposed method outperforms the state-of-the-art methods in both accuracy and scalability. | Zichen Miao, Ze Wang, Wei Chen, Qiang Qiu | null | null | 2,022 | iclr |
Why so pessimistic? Estimating uncertainties for offline RL through ensembles, and why their independence matters. | null | In order to achieve strong performance in offline reinforcement learning (RL), it is necessary to act conservatively with respect to confident lower-bounds on anticipated values of actions. Thus, a valuable approach would be to obtain high quality uncertainty estimates on action values. In current supervised learning literature, state-of-the-art approaches to uncertainty estimation and calibration rely on ensembling methods. In this work, we aim to transfer the success of ensembles from supervised learning to the setting of batch RL. We propose, MSG, a model-free dynamic programming based offline RL method that trains an ensemble of independent Q-functions, and updates a policy to act conservatively with respect to the uncertainties derived from the ensemble. Theoretically, by referring to the literature on infinite-width neural networks, we demonstrate the crucial dependence of the quality of uncertainty on the manner in which ensembling is performed, a phenomenon that arises due to the dynamic programming nature of RL and overlooked by existing offline RL methods. Our theoretical predictions are corroborated by pedagogical examples on toy MDPs, as well as empirical comparisons in benchmark continuous control domains. In the more challenging domains of the D4RL offline RL benchmark, MSG significantly surpasses highly well-tuned state-of-the-art methods in batch RL. Motivated by the success of MSG, we investigate whether efficient approximations to ensembles can be as effective. We demonstrate that while efficient variants outperform current state-of-the-art, they do not match MSG with deep ensembles. We hope our work engenders increased focus into deep network uncertainty estimation techniques directed for reinforcement learning. | Seyed Kamyar Seyed Ghasemipour, Shixiang Shane Gu, Ofir Nachum | null | null | 2,022 | iclr |
Robot Intent Recognition Method Based on State Grid Business Office | null | Artificial intelligence is currently in an era of change, not only changing the artificial intelligence technology itself, but also changing human society. It has become more and more common to use artificial intelligence as the core human-computer interaction technology to replace manpower. Intention recognition is an important part of the human-machine dialogue system, and deep learning technology is gradually being applied to the task of intent recognition. However, intent recognition based on deep learning often has problems such as low recognition accuracy and slow recognition speed. In response to these problems, this paper designs a BERT fine-tuning to improve the network structure based on the pre-training model and proposes new continuous pre-training goals. To improve the accuracy of intent recognition, a method based on multi-teacher model compression is proposed to compress the pre-training model, which reduces the time consumption of model inference. | Lanfang Dong, Zhao Pu Hu, Hanchao Liu | null | null | 2,022 | iclr |
Einops: Clear and Reliable Tensor Manipulations with Einstein-like Notation | null | Tensor computations underlie modern scientific computing and deep learning.
A number of tensor frameworks emerged varying in execution model, hardware support, memory management, model definition, etc.
However, tensor operations in all frameworks follow the same paradigm.
Recent neural network architectures demonstrate demand for higher expressiveness of tensor operations.
The current paradigm is not suited to write readable, reliable, or easy-to-modify code for multidimensional tensor manipulations.
Moreover, some commonly used operations do not provide sufficient checks and can break a tensor structure.
These mistakes are elusive as no tools or tests can detect them.
Independently, API discrepancies complicate code transfer between frameworks.
We propose einops notation: a uniform and generic way to manipulate tensor structure, that significantly improves code readability and flexibility by focusing on the structure of input and output tensors.
We implement einops notation in a Python package that efficiently supports multiple widely used frameworks and provides framework-independent minimalist API for tensor manipulations. | Alex Rogozhnikov | null | null | 2,022 | iclr |
Neural Extensions: Training Neural Networks with Set Functions | null | Integrating discrete computational steps into deep learning architectures is an important consideration when learning to reason over discrete items. However, many tasks that involve discrete choices are defined via (combinatorial) set functions, and thereby pose challenges for end-to-end training. In this work, we explore a general framework to construct continuous extensions of such discrete functions that enables training via gradient methods. Our framework includes well-known extensions such as the Lovasz extension of submodular set functions and facilitates the design of novel continuous extensions based on problem-specific considerations, including constraints. We demonstrate the versatility of our framework on tasks ranging from combinatorial optimization to image classification. | Nikolaos Karalias, Joshua David Robinson, Andreas Loukas, Stefanie Jegelka | null | null | 2,022 | iclr |
A Large Batch Optimizer Reality Check: Traditional, Generic Optimizers Suffice Across Batch Sizes | null | Recently the LARS and LAMB optimizers have been proposed for training neural networks faster using large batch sizes. LARS and LAMB add layer-wise normalization to the update rules of Heavy-ball momentum and Adam, respectively, and have become popular in prominent benchmarks and deep learning libraries. However, without fair comparisons to standard optimizers, it remains an open question whether LARS and LAMB have any benefit over traditional, generic algorithms. In this work we demonstrate that standard optimization algorithms such as Nesterov momentum and Adam can match or exceed the results of LARS and LAMB at large batch sizes. Our results establish new, stronger baselines for future comparisons at these batch sizes and shed light on the difficulties of comparing optimizers for neural network training more generally. | Zachary Nado, Justin Gilmer, Christopher J Shallue, Rohan Anil, George Edward Dahl | null | null | 2,022 | iclr |
New Definitions and Evaluations for Saliency Methods: Staying Intrinsic and Sound | null | Saliency methods seek to provide human-interpretable explanations for the output of machine learning model on a given input. A plethora of saliency methods exist, as well as an extensive literature on their justifications/criticisms/evaluations. This paper focuses on heat maps based saliency methods that often provide explanations that look best to humans. It tries to introduce methods and evaluations for masked-based saliency methods that are {\em intrinsic} --- use just the training dataset and the trained net, and do not use separately trained nets, distractor distributions, human evaluations or annotations. Since a mask can be seen as a "certificate" justifying the net's answer, we introduce notions of {\em completeness} and {\em soundness} (the latter being the new contribution) motivated by logical proof systems. These notions allow a new evaluation of saliency methods, that experimentally provides a novel and stronger justification for several heuristic tricks in the field (T.V. regularization, upscaling). | Arushi Gupta, Nikunj Saunshi, Dingli Yu, Kaifeng Lyu, Sanjeev Arora | null | null | 2,022 | iclr |
Human imperceptible attacks and applications to improve fairness | null | Modern neural networks are able to perform at least as well as humans in numerous tasks involving object classification and image generation. However, there is also evidence that perturbations which are imperceptible to humans may significantly degrade the performance of well-trained deep neural networks. We provide a Distributionally Robust Optimization (DRO) framework which integrates human-based image quality assessment methods to design optimal attacks that are imperceptible to humans but significantly damaging to deep neural networks. Our attack algorithm can generate better-quality (less perceptible to humans) attacks than other state-of-the-art human imperceptible attack methods. We provide an algorithmic implementation of independent interest which can speed up DRO training significantly. Finally, we demonstrate how the use of optimally designed human imperceptible attacks can improve group fairness in image classification while maintaining a similar accuracy.
| Xinru Hua, Huanzhong Xu, Jose Blanchet, Viet Anh Nguyen | null | null | 2,022 | iclr |
SphereFace2: Binary Classification is All You Need for Deep Face Recognition | null | State-of-the-art deep face recognition methods are mostly trained with a softmax-based multi-class classification framework. Despite being popular and effective, these methods still have a few shortcomings that limit empirical performance. In this paper, we start by identifying the discrepancy between training and evaluation in the existing multi-class classification framework and then discuss the potential limitations caused by the "competitive" nature of softmax normalization. Motivated by these limitations, we propose a novel binary classification training framework, termed SphereFace2. In contrast to existing methods, SphereFace2 circumvents the softmax normalization, as well as the corresponding closed-set assumption. This effectively bridges the gap between training and evaluation, enabling the representations to be improved individually by each binary classification task. Besides designing a specific well-performing loss function, we summarize a few general principles for this "one-vs-all" binary classification framework so that it can outperform current competitive methods. Our experiments on popular benchmarks demonstrate that SphereFace2 can consistently outperform state-of-the-art deep face recognition methods. | Yandong Wen, Weiyang Liu, Adrian Weller, Bhiksha Raj, Rita Singh | null | null | 2,022 | iclr |
Learning Surface Parameterization for Document Image Unwarping | null | In this paper, we present a novel approach to learn texture mapping for a 3D surface and apply it to document image unwarping. We propose an efficient method to learn surface parameterization by learning a continuous bijective mapping between 3D surface positions and 2D texture-space coordinates. Our surface parameterization network can be conveniently plugged into a differentiable rendering pipeline and trained using multi-view images and rendering loss. Recent work on differentiable rendering techniques for implicit surfaces has shown high-quality 3D scene reconstruction and view synthesis results. However, these methods typically learn the appearance color as a function of the surface points and lack explicit surface parameterization. Thus they do not allow texture map extraction or texture editing. By introducing explicit surface parameterization and learning with a recent differentiable renderer for implicit surfaces, we demonstrate state-of-the-art document-unwarping via texture extraction. We show that our approach can reconstruct high-frequency textures for arbitrary document shapes in both synthetic and real scenarios. We also demonstrate the usefulness of our system by applying it to document texture editing. | Sagnik Das, Ke Ma, Zhixin Shu, Dimitris Samaras | null | null | 2,022 | iclr |
Non-Parallel Text Style Transfer with Self-Parallel Supervision | null | The performance of existing text style transfer models is severely limited by the non-parallel datasets on which the models are trained. In non-parallel datasets, no direct mapping exists between sentences of the source and target style; the style transfer models thus only receive weak supervision of the target sentences during training, which often leads the model to discard too much style-independent information, or utterly fail to transfer the style.
In this work, we propose LaMer, a novel text style transfer framework based on large-scale language models. LaMer first mines the roughly parallel expressions in the non-parallel datasets with scene graphs, and then employs MLE training, followed by imitation learning refinement, to leverage the intrinsic parallelism within the data. On two benchmark tasks (sentiment & formality transfer) and a newly proposed challenging task (political stance transfer), our model achieves qualitative advances in transfer accuracy, content preservation, and fluency. Further empirical and human evaluations demonstrate that our model not only makes training more efficient, but also generates more readable and diverse expressions than previous models. | Ruibo Liu, Chongyang Gao, Chenyan Jia, Guangxuan Xu, Soroush Vosoughi | null | null | 2,022 | iclr |
Neural Structured Prediction for Inductive Node Classification | null | This paper studies node classification in the inductive setting, i.e., aiming to learn a model on labeled training graphs and generalize it to infer node labels on unlabeled test graphs. This problem has been extensively studied with graph neural networks (GNNs) by learning effective node representations, as well as traditional structured prediction methods for modeling the structured output of node labels, e.g., conditional random fields (CRFs). In this paper, we present a new approach called the Structured Proxy Network (SPN), which combines the advantages of both worlds. SPN defines flexible potential functions of CRFs with GNNs. However, learning such a model is nontrivial as it involves optimizing a maximin game with high-cost inference. Inspired by the underlying connection between joint and marginal distributions defined by Markov networks, we propose to solve an approximate version of the optimization problem as a proxy, which yields a near-optimal solution, making learning more efficient. Extensive experiments on two settings show that our approach outperforms many competitive baselines. | Meng Qu, Huiyu Cai, Jian Tang | null | null | 2,022 | iclr |
Personalized PageRank meets Graph Attention Networks | null | There has been a rising interest in graph neural networks (GNNs) for representation learning over the past few years. GNNs provide a general and efficient framework to learn from graph-structured data. However, GNNs typically only use the information of a very limited neighborhood for each node. A larger neighborhood would be desirable to provide the model with more information. However, increasing the size of the neighborhood is not trivial since neighborhood aggregation over many layers leads to over-smoothing. In this work, we incorporate the limit distribution of Personalized PageRank (PPR) into graph attention networks (GATs) to address this issue. Intuitively, message aggregation based on Personalized PageRank corresponds to infinitely many neighborhood aggregation layers. We show that our models outperform a variety ofbaseline models across all datasets used for our experiments. Our implementation is publicly available online. | Julie Choi | null | null | 2,022 | iclr |
Cluster Tree for Nearest Neighbor Search | null | Tree-based algorithms are an important and widely used class of algorithms for Nearest Neighbor Search (NNS) with random partition (RP) tree being arguably the most well studied. However, in spite of possessing theoretical guarantees and strong practical performance, a major drawback of the RP tree is its lack of adaptability to the input dataset.
Inspired by recent theoretical and practical works for NNS, we attempt to remedy this by introducing ClusterTree, a new tree based algorithm. Our approach utilizes randomness as in RP trees while adapting to the underlying cluster structure of the dataset to create well-balanced and meaningful partitions. Experimental evaluations on real world datasets demonstrate improvements over RP trees and other tree based methods for NNS while maintaining efficient construction time. In addition, we show theoretically and empirically that ClusterTree finds partitions which are superior to those found by RP trees in preserving the cluster structure of the input dataset. | Dan Kushnir, Sandeep Silwal | null | null | 2,022 | iclr |
Filtered-CoPhy: Unsupervised Learning of Counterfactual Physics in Pixel Space | null | Learning causal relationships in high-dimensional data (images, videos) is a hard task, as they are often defined on low dimensional manifolds and must be extracted from complex signals dominated by appearance, lighting, textures and also spurious correlations in the data. We present a method for learning counterfactual reasoning of physical processes in pixel space, which requires the prediction of the impact of interventions on initial conditions. Going beyond the identification of structural relationships, we deal with the challenging problem of forecasting raw video over long horizons. Our method does not require the knowledge or supervision of any ground truth positions or other object or scene properties. Our model learns and acts on a suitable hybrid latent representation based on a combination of dense features, sets of 2D keypoints and an additional latent vector per keypoint. We show that this better captures the dynamics of physical processes than purely dense or sparse representations. We introduce a new challenging and carefully designed counterfactual benchmark for predictions in pixel space and outperform strong baselines in physics-inspired ML and video prediction. | Steeven JANNY, Fabien Baradel, Natalia Neverova, Madiha Nadri, Greg Mori, Christian Wolf | null | null | 2,022 | iclr |
Multi-Trigger-Key: Towards Multi-Task Privacy-Preserving In Deep Learning | null | Deep learning-based Multi-Task Classification (MTC) is widely used in applications like facial attribute and healthcare that warrant strong privacy guarantees. In this work, we aim to protect sensitive information in the inference phase of MTC and propose a novel Multi-Trigger-Key (MTK) framework to achieve the privacy-preserving objective. MTK associates each secured task in the multi-task dataset with a specifically designed trigger-key. The true information can be revealed by adding the trigger-key if the user is authorized. We obtain such an MTK model by training it with a newly generated training set. To address the information leakage malaise resulting from correlations among different tasks, we generalize the training process by incorporating an MTK decoupling process with a controllable trade-off between the protective efficacy and the model performance. Theoretical guarantees and experimental results demonstrate the effectiveness of the privacy protection without appreciable hindering on the model performance. | Ren Wang, Zhe Xu, Alfred Hero | null | null | 2,022 | iclr |
Neocortical cell type classification from electrophysiology recordings using deep neural networks | null | Understanding the neural code requires identifying different functional units involved in the neural circuits. One way to identify these functional units is to solve a neuron type classification problem. For decades, current clamp electrophysiology recordings have provided the means to classify the neurons based on subtle differences in action potential shapes and spiking patterns. However, significant variations in neuronal type definitions, classification pipelines, and variability in the neuronal activities make unambiguous determination of neuron type challenging. Previous solutions to this electrophysiology-based cell type classification problem consisted of dimensionality reduction juxtaposed with clustering using hand-crafted action potential features. Recent discoveries have allowed genetic-based cell-type classifications, which have fewer ambiguities, but they are less practical in vivo and have even lower throughput. Leveraging the unprecedented ground truth data published in the Allen Institute Cell Types Database, which contains anatomical, genetic, and electrophysiology characterizations of neurons in the mouse neocortex, we construct a robust and efficient convolutional neural network (CNN) that successfully classifies neurons according to their genetic label or broad type (excitatory or inhibitory) solely using current-clamp electrophysiology recordings. The CNN is configured as a multiple-input single-output network consisting of three subnetworks that take in the raw time series electrophysiology recording as well as the real and imaginary components of its Fourier coefficients. Our single pipeline method is fast and streamlined while simultaneously outperforming previous methods and achieving more classification classes using only single current-clamp trace as the input. This end-to-end convolutional neural network-based classification method removes the need for hand-crafted features, specific knowledge, or human intervention for quick identification of the cell type with high accuracy, enabling interpretation of the experimental data in a bias-free manner and a much broader scientific context. | Raymond Wang, Sang Min Han, Marta Agnieszka Gajowa, Chunlei Liu | null | null | 2,022 | iclr |
Learning Towards The Largest Margins | null | One of the main challenges for feature representation in deep learning-based classification is the design of appropriate loss functions that exhibit strong discriminative power. The classical softmax loss does not explicitly encourage discriminative learning of features. A popular direction of research is to incorporate margins in well-established losses in order to enforce extra intra-class compactness and inter-class separability, which, however, were developed through heuristic means, as opposed to rigorous mathematical principles. In this work, we attempt to address this limitation by formulating the principled optimization objective as learning towards the largest margins. Specifically, we firstly propose to employ the class margin as the measure of inter-class separability, and the sample margin as the measure of intra-class compactness. Accordingly, to encourage discriminative representation of features, the loss function should promote the largest possible margins for both classes and samples. Furthermore, we derive a generalized margin softmax loss to draw general conclusions for the existing margin-based losses. Not only does this principled framework offer new perspectives to understand and interpret existing margin-based losses, but it also provides new insights that can guide the design of new tools, including \textit{sample margin regularization} and \textit{largest margin softmax loss} for class balanced cases, and \textit{zero centroid regularization} for class imbalanced cases. Experimental results demonstrate the effectiveness of our strategy for multiple tasks including visual classification, imbalanced classification, person re-identification, and face verification. | Xiong Zhou, Xianming Liu, Deming Zhai, Junjun Jiang, Xin Gao, Xiangyang Ji | null | null | 2,022 | iclr |
Open-Set Recognition: A Good Closed-Set Classifier is All You Need | null | The ability to identify whether or not a test sample belongs to one of the semantic classes in a classifier's training set is critical to practical deployment of the model. This task is termed open-set recognition (OSR) and has received significant attention in recent years. In this paper, we first demonstrate that the ability of a classifier to make the 'none-of-above' decision is highly correlated with its accuracy on the closed-set classes. We find that this relationship holds across loss objectives and architectures, and further demonstrate the trend both on the standard OSR benchmarks as well as on a large-scale ImageNet evaluation. Second, we use this correlation to boost the performance of the maximum softmax probability OSR 'baseline' by improving its closed-set accuracy, and with this strong baseline achieve state-of-the-art on a number of OSR benchmarks. Similarly, we boost the performance of the existing state-of-the-art method by improving its closed-set accuracy, but the resulting discrepancy with the strong baseline is marginal. Our third contribution is to present the 'Semantic Shift Benchmark' (SSB), which better respects the task of detecting semantic novelty, as opposed to low-level distributional shifts as tackled by neighbouring machine learning fields. On this new evaluation, we again demonstrate that there is negligible difference between the strong baseline and the existing state-of-the-art. Code available at: https://github.com/sgvaze/osr_closed_set_all_you_need. | Sagar Vaze, Kai Han, Andrea Vedaldi, Andrew Zisserman | null | null | 2,022 | iclr |
Learning Fast Samplers for Diffusion Models by Differentiating Through Sample Quality | null | Diffusion models have emerged as an expressive family of generative models rivaling GANs in sample quality and autoregressive models in likelihood scores. Standard diffusion models typically require hundreds of forward passes through the model to generate a single high-fidelity sample. We introduce Differentiable Diffusion Sampler Search (DDSS): a method that optimizes fast samplers for any pre-trained diffusion model by differentiating through sample quality scores. We also present Generalized Gaussian Diffusion Models (GGDM), a family of flexible non-Markovian samplers for diffusion models. We show that optimizing the degrees of freedom of GGDM samplers by maximizing sample quality scores via gradient descent leads to improved sample quality. Our optimization procedure backpropagates through the sampling process using the reparametrization trick and gradient rematerialization. DDSS achieves strong results on unconditional image generation across various datasets (e.g., FID scores on LSUN church 128x128 of 11.6 with only 10 inference steps, and 4.82 with 20 steps, compared to 51.1 and 14.9 with strongest DDPM/DDIM baselines). Our method is compatible with any pre-trained diffusion model without fine-tuning or re-training required. | Daniel Watson, William Chan, Jonathan Ho, Mohammad Norouzi | null | null | 2,022 | iclr |
Deep Recurrent Neural Network Layers with Layerwise Loss | null | Deep learning techniques have brought significant performance improvement to various areas of machine learning. Especially in the computer vision area, very deep networks such as ResNet have shown notable performance improvement. However, in speech recognition or language processing, such kinds of a very deep network have not been extensively employed. In this paper, we propose a very deep LSTM structure and their training strategy. In our training strategy, we first start training a conventional model with several LSTM layers. One notable difference is that for the top LSTM layer of the initial model, the Connectionist Temporal Classification (CTC) loss is applied both to the input and output of this top LSTM layer. Once this initial model is sufficiently layered, this top layer is copied to construct a very deep LSTM stack. For this newly constructed stack, the CTC loss is applied to every output of the LSTM layer as well as the top of the stack. Experimental results show that this deep LSTM structure shows significantly better results than the conventional model with 5 ~ 6 layers with a comparable number of parameters.
| Chanwoo Kim | null | null | 2,022 | iclr |
On the Connection between Local Attention and Dynamic Depth-wise Convolution | null | Vision Transformer (ViT) attains state-of-the-art performance in visual recognition, and the variant, Local Vision Transformer, makes further improvements. The major component in Local Vision Transformer, local attention, performs the attention separately over small local windows. We rephrase local attention as a channel-wise locally-connected layer and analyze it from two network regularization manners, sparse connectivity and weight sharing, as well as dynamic weight computation. We point out that local attention resembles depth-wise convolution and its dynamic variants in sparse connectivity: there is no connection across channels, and each position is connected to the positions within a small local window. The main differences lie in (i) weight sharing - depth-wise convolution shares connection weights (kernel weights) across spatial positions and attention shares the connection weights across channels, and (ii) dynamic weight computation manners - local attention is based on dot-products between pairwise positions in the local window, and dynamic convolution is based on linear projections conducted on the center representation or the globally pooled representation. The connection between local attention and dynamic depth-wise convolution is empirically verified by the ablation study about weight sharing and dynamic weight computation in Local Vision Transformer and (dynamic) depth-wise convolution. We empirically observe that the models based on depth-wise convolution and the dynamic variants with lower computation complexity perform on-par with or slightly better than Swin Transformer, an instance of Local Vision Transformer, for ImageNet classification, COCO object detection and ADE semantic segmentation. Code is available at https://github.com/Atten4Vis/DemystifyLocalViT. | Qi Han, Zejia Fan, Qi Dai, Lei Sun, Ming-Ming Cheng, Jiaying Liu, Jingdong Wang | null | null | 2,022 | iclr |
ProtoRes: Proto-Residual Network for Pose Authoring via Learned Inverse Kinematics | null | Our work focuses on the development of a learnable neural representation of human pose for advanced AI assisted animation tooling. Specifically, we tackle the problem of constructing a full static human pose based on sparse and variable user inputs (e.g. locations and/or orientations of a subset of body joints). To solve this problem, we propose a novel neural architecture that combines residual connections with prototype encoding of a partially specified pose to create a new complete pose from the learned latent space. We show that our architecture outperforms a baseline based on Transformer, both in terms of accuracy and computational efficiency. Additionally, we develop a user interface to integrate our neural model in Unity, a real-time 3D development platform. Furthermore, we introduce two new datasets representing the static human pose modeling problem, based on high-quality human motion capture data, which will be released publicly along with model code. | Boris N. Oreshkin, Florent Bocquelet, Felix G. Harvey, Bay Raitt, Dominic Laflamme | null | null | 2,022 | iclr |
Revisiting the Lottery Ticket Hypothesis: A Ramanujan Graph Perspective | null | Neural networks often yield to weight pruning resulting in a sparse subnetwork that is adequate for a given task. Retraining these `lottery ticket' subnetworks from their initialization minimizes the computational burden while preserving the test set accuracy of the original network. Based on our knowledge, the existing literature only confirms that pruning is needed and it can be achieved up to certain sparsity. We analyze the pruned network in the context of the properties of Ramanujan expander graphs. We consider the feed-forward network (both multi-layer perceptron and convolutional network) as a series of bipartite graphs which establish the connection from input to output. Now, as the fraction of remaining weights reduce with increasingly aggressive pruning two distinct regimes are observed: initially, no significant decrease in accuracy is demonstrated, and then the accuracy starts dropping rapidly. We empirically show that in the first regime the pruned lottery ticket sub-network remains a Ramanujan graph. Subsequently, with the loss of Ramanujan graph property, accuracy begins to reduce sharply. This characterizes an absence of resilient connectivity in the pruned sub-network. We also propose a new magnitude-based pruning algorithm to preserve the above property. We perform experiments on MNIST and CIFAR10 datasets using different established feed-forward architectures and show that the winning ticket obtained from the proposed algorithm is much more robust. | BITHIKA PAL, Arindam Biswas, Pabitra Mitra, BISWAJIT BASU | null | null | 2,022 | iclr |
Variational Neural Cellular Automata | null | In nature, the process of cellular growth and differentiation has lead to an amazing diversity of organisms --- algae, starfish, giant sequoia, tardigrades, and orcas are all created by the same generative process.
Inspired by the incredible diversity of this biological generative process, we propose a generative model, the Variational Neural Cellular Automata (VNCA), which is loosely inspired by the biological processes of cellular growth and differentiation. Unlike previous related works, the VNCA is a proper probabilistic generative model, and we evaluate it according to best practices. We find that the VNCA learns to reconstruct samples well and that despite its relatively few parameters and simple local-only communication, the VNCA can learn to generate a large variety of output from information encoded in a common vector format. While there is a significant gap to the current state-of-the-art in terms of generative modeling performance, we show that the VNCA can learn a purely self-organizing generative process of data. Additionally, the self-organizing nature bestows the VNCA with some inherent robustness against perturbations in the early stages of growth. | Rasmus Berg Palm, Miguel González Duque, Shyam Sudhakaran, Sebastian Risi | null | null | 2,022 | iclr |
Reinforcement Learning with Sparse Rewards using Guidance from Offline Demonstration | null | A major challenge in real-world reinforcement learning (RL) is the sparsity of reward feedback. Often, what is available is an intuitive but sparse reward function that only indicates whether the task is completed partially or fully. However, the lack of carefully designed, fine grain feedback implies that most existing RL algorithms fail to learn an acceptable policy in a reasonable time frame. This is because of the large number of exploration actions that the policy has to perform before it gets any useful feedback that it can learn from. In this work, we address this challenging problem by developing an algorithm that exploits the offline demonstration data generated by {a sub-optimal behavior policy} for faster and efficient online RL in such sparse reward settings. The proposed algorithm, which we call the Learning Online with Guidance Offline (LOGO) algorithm, merges a policy improvement step with an additional policy guidance step by using the offline demonstration data. The key idea is that by obtaining guidance from - not imitating - the offline {data}, LOGO orients its policy in the manner of the sub-optimal {policy}, while yet being able to learn beyond and approach optimality. We provide a theoretical analysis of our algorithm, and provide a lower bound on the performance improvement in each learning episode. We also extend our algorithm to the even more challenging incomplete observation setting, where the demonstration data contains only a censored version of the true state observation. We demonstrate the superior performance of our algorithm over state-of-the-art approaches on a number of benchmark environments with sparse rewards {and censored state}. Further, we demonstrate the value of our approach via implementing LOGO on a mobile robot for trajectory tracking and obstacle avoidance, where it shows excellent performance. | Desik Rengarajan, Gargi Vaidya, Akshay Sarvesh, Dileep Kalathil, Srinivas Shakkottai | null | null | 2,022 | iclr |
Quantum Alphatron | null | Finding provably efficient algorithms for learning neural networks is a fundamental challenge in the theory of machine learning. The Alphatron of Goel and Klivans is the first provably efficient algorithm for learning neural networks with more than one nonlinear layer. The algorithm succeeds with any distribution on the $n$-dimensional unit ball and without any assumption on the structure of the network. In this work, we refine the original Alphatron by a pre-computing phase for its most time-consuming part, the evaluation of the kernel function. This refined algorithm improves the run time of the original Alphatron, while retaining the same learning guarantee. Based on the refined algorithm, we quantize the pre-computing phase with provable learning guarantee in the fault-tolerant quantum computing model. In a well-defined learning model, this quantum algorithm is able to provide a quadratic speedup in the data dimension $n$. In addition, we discuss the second type of speedup, quantizing the evaluation of the gradient in the stochastic gradient descent procedure. Our work contributes to the study of quantum learning with kernels and from samples. | Siyi Yang, Patrick Rebentrost, Miklos Santha | null | null | 2,022 | iclr |
Boosting the Confidence of Near-Tight Generalization Bounds for Uniformly Stable Randomized Algorithms | null | High probability generalization bounds of uniformly stable learning algorithms have recently been actively studied with a series of near-tight results established by~\citet{feldman2019high,bousquet2020sharper}. However, for randomized algorithms with on-average uniform stability, such as stochastic gradient descent (SGD) with time decaying learning rates, it still remains less well understood if these deviation bounds still hold with high confidence over the internal randomness of algorithm. This paper addresses this open question and makes progress towards answering it inside a classic framework of confidence-boosting. To this end, we first establish an in-expectation first moment generalization error bound for randomized learning algorithm with on-average uniform stability, based on which we then show that a properly designed subbagging process leads to near-tight high probability generalization bounds over the randomness of data and algorithm. We further substantialize these generic results to SGD to derive improved high probability generalization bounds for convex or non-convex optimization with natural time decaying learning rates, which have not been possible to prove with the existing uniform stability results. Specially for deterministic uniformly stable algorithms, our confidence-boosting results improve upon the best known generalization bounds in terms of a logarithmic factor on sample size, which moves a step forward towards resolving an open question raised by~\citet{bousquet2020sharper}. | Xiaotong Yuan, Ping Li | null | null | 2,022 | iclr |
Sequence Approximation using Feedforward Spiking Neural Network for Spatiotemporal Learning: Theory and Optimization Methods | null | A dynamical system of spiking neurons with only feedforward connections can classify spatiotemporal patterns without recurrent connections. However, the theoretical construct of a feedforward spiking neural network (SNN) for approximating a temporal sequence remains unclear, making it challenging to optimize SNN architectures for learning complex spatiotemporal patterns. In this work, we establish a theoretical framework to understand and improve sequence approximation using a feedforward SNN. Our framework shows that a feedforward SNN with one neuron per layer and skip-layer connections can approximate the mapping function between any arbitrary pairs of input and output spike train on a compact domain. Moreover, we prove that heterogeneous neurons with varying dynamics and skip-layer connections improve sequence approximation using feedforward SNN. Consequently, we propose SNN architectures incorporating the preceding constructs that are trained using supervised backpropagation-through-time (BPTT) and unsupervised spiking-timing-dependent plasticity (STDP) algorithms for classification of spatiotemporal data. A dual-search-space Bayesian optimization method is developed to optimize architecture and parameters of the proposed SNN with heterogeneous neuron dynamics and skip-layer connections. | Xueyuan She, Saurabh Dash, Saibal Mukhopadhyay | null | null | 2,022 | iclr |
Fast Regression for Structured Inputs | null | We study the $\ell_p$ regression problem, which requires finding $\mathbf{x}\in\mathbb R^{d}$ that minimizes $\|\mathbf{A}\mathbf{x}-\mathbf{b}\|_p$ for a matrix $\mathbf{A}\in\mathbb R^{n \times d}$ and response vector $\mathbf{b}\in\mathbb R^{n}$. There has been recent interest in developing subsampling methods for this problem that can outperform standard techniques when $n$ is very large. However, all known subsampling approaches have run time that depends exponentially on $p$, typically, $d^{\mathcal{O}(p)}$, which can be prohibitively expensive.
We improve on this work by showing that for a large class of common \emph{structured matrices}, such as combinations of low-rank matrices, sparse matrices, and Vandermonde matrices, there are subsampling based methods for $\ell_p$ regression that depend polynomially on $p$. For example, we give an algorithm for $\ell_p$ regression on Vandermonde matrices that runs in time $\mathcal{O}(n\log^3 n+(dp^2)^{0.5+\omega}\cdot\text{polylog}\,n)$, where $\omega$ is the exponent of matrix multiplication. The polynomial dependence on $p$ crucially allows our algorithms to extend naturally to efficient algorithms for $\ell_\infty$ regression, via approximation of $\ell_\infty$ by $\ell_{\mathcal{O}(\log n)}$. Of practical interest, we also develop a new subsampling algorithm for $\ell_p$ regression for arbitrary matrices, which is simpler than previous approaches for $p \ge 4$. | Raphael A Meyer, Cameron N Musco, Christopher P Musco, David Woodruff, Samson Zhou | null | null | 2,022 | iclr |
Self-Supervise, Refine, Repeat: Improving Unsupervised Anomaly Detection | null | Anomaly detection (AD) - separating anomalies from normal data - has many applications across domains, from manufacturing to healthcare. While most previous works have been shown to be effective for cases with fully or partially labeled data, that setting is in practice less common due to labeling being particularly tedious for this task. In this paper, we focus on fully unsupervised AD, in which the entire training dataset, containing both normal and anomalous samples, is unlabeled. To tackle this problem effectively, we propose to improve the robustness of one-class classification trained on self-supervised representations using a data refinement process. Our proposed data refinement approach is based on an ensemble of one-class classifiers (OCCs), each of which is trained on a disjoint subset of training data. Representations learned by self-supervised learning on the refined data are iteratively updated as the refinement improves. We demonstrate our method on various unsupervised AD tasks with image and tabular data. With a 10% anomaly ratio on CIFAR-10 image data / 2.5% anomaly ratio on Thyroid tabular data, the proposed method outperforms the state-of-the-art one-class classification method by 6.3 AUC and 12.5 average precision / 22.9 F1-score. | Jinsung Yoon, Kihyuk Sohn, Chun-Liang Li, Sercan O Arik, Chen-Yu Lee, Tomas Pfister | null | null | 2,022 | iclr |
Interacting Contour Stochastic Gradient Langevin Dynamics | null | We propose an interacting contour stochastic gradient Langevin dynamics (ICSGLD) sampler, an embarrassingly parallel multiple-chain contour stochastic gradient Langevin dynamics (CSGLD) sampler with efficient interactions. We show that ICSGLD can be theoretically more efficient than a single-chain CSGLD with an equivalent computational budget. We also present a novel random-field function, which facilitates the estimation of self-adapting parameters in big data and obtains free mode explorations. Empirically, we compare the proposed algorithm with popular benchmark methods for posterior sampling. The numerical results show a great potential of ICSGLD for large-scale uncertainty estimation tasks. | Wei Deng, Siqi Liang, Botao Hao, Guang Lin, Faming Liang | null | null | 2,022 | iclr |
Diverse Client Selection for Federated Learning via Submodular Maximization | null | In every communication round of federated learning, a random subset of clients communicate their model updates back to the server which then aggregates them all. The optimal size of this subset is not known and several studies have shown that typically random selection does not perform very well in terms of convergence, learning efficiency and fairness. We, in this paper, propose to select a small diverse subset of clients, namely those carrying representative gradient information, and we transmit only these updates to the server. Our aim is for updating via only a subset to approximate updating via aggregating all client information. We achieve this by choosing a subset that maximizes a submodular facility location function defined over gradient space. We introduce “federated averaging with diverse client selection (DivFL)”. We provide a thorough analysis of its convergence in the heterogeneous setting and apply it both to synthetic and to real datasets. Empirical results show several benefits to our approach including improved learning efficiency, faster convergence and also more uniform (i.e., fair) performance across clients. We further show a communication-efficient version of DivFL that can still outperform baselines on the above metrics. | Ravikumar Balakrishnan, Tian Li, Tianyi Zhou, Nageen Himayat, Virginia Smith, Jeff Bilmes | null | null | 2,022 | iclr |
Data-Efficient Graph Grammar Learning for Molecular Generation | null | The problem of molecular generation has received significant attention recently. Existing methods are typically based on deep neural networks and require training on large datasets with tens of thousands of samples. In practice, however, the size of class-specific chemical datasets is usually limited (e.g., dozens of samples) due to labor-intensive experimentation and data collection. Another major challenge is to generate only physically synthesizable molecules. This is a non-trivial task for neural network-based generative models since the relevant chemical knowledge can only be extracted and generalized from the limited training data. In this work, we propose a data-efficient generative model that can be learned from datasets with orders of magnitude smaller sizes than common benchmarks. At the heart of this method is a learnable graph grammar that generates molecules from a sequence of production rules. Without any human assistance, these production rules are automatically constructed from training data. Furthermore, additional chemical knowledge can be incorporated into the model by further grammar optimization. Our learned graph grammar yields state-of-the-art results on generating high-quality molecules for three monomer datasets that contain only ${\sim}20$ samples each. Our approach also achieves remarkable performance in a challenging polymer generation task with $only$ $117$ training samples and is competitive against existing methods using $81$k data points.
| Minghao Guo, Veronika Thost, Beichen Li, Payel Das, Jie Chen, Wojciech Matusik | null | null | 2,022 | iclr |
Finetuned Language Models are Zero-Shot Learners | null | This paper explores a simple method for improving the zero-shot learning abilities of language models. We show that instruction tuning—finetuning language models on a collection of datasets described via instructions—substantially improves zero-shot performance on unseen tasks. We take a 137B parameter pretrained language model and instruction tune it on over 60 NLP datasets verbalized via natural language instruction templates. We evaluate this instruction-tuned model, which we call FLAN, on unseen task types. FLAN substantially improves the performance of its unmodified counterpart and surpasses zero-shot 175B GPT-3 on 20 of 25 datasets that we evaluate. FLAN even outperforms few-shot GPT-3 by a large margin on ANLI, RTE, BoolQ, AI2-ARC, OpenbookQA, and StoryCloze. Ablation studies reveal that number of finetuning datasets, model scale, and natural language instructions are key to the success of instruction tuning. | Jason Wei, Maarten Bosma, Vincent Zhao, Kelvin Guu, Adams Wei Yu, Brian Lester, Nan Du, Andrew M. Dai, Quoc V Le | null | null | 2,022 | iclr |
Continual Learning Using Task Conditional Neural Networks | null | Conventional deep learning models have limited capacity in learning multiple tasks sequentially. The issue of forgetting the previously learned tasks in continual learning is known as catastrophic forgetting or interference. When the input data or the goal of learning changes, a continual model will learn and adapt to the new status. However, the model will not remember or recognise any revisits to the previous states. This causes performance reduction and re-training curves in dealing with periodic or irregularly reoccurring changes in the data or goals. Dynamic approaches, which assign new neuron resources to the upcoming tasks, are introduced to address this issue. However, most of the dynamic methods need task information about the upcoming tasks during the inference phase to activate the corresponding neurons. To address this issue, we introduce Task Conditional Neural Network which allows the model to identify the task information automatically. The proposed model can continually learn and embed new tasks into the model without losing the information about previously learned tasks. We evaluate the proposed model combined with the mixture of experts approach on the MNIST and CIFAR100 datasets and show how it significantly improves the continual learning process without requiring task information in advance. | Honglin Li, Frieder Ganz, David J. Sharp, Payam M. Barnaghi | null | null | 2,022 | iclr |
Deep banach space kernels | null | The recent success of deep learning has encouraged many researchers to explore the deep/concatenated variants of classical kernel methods. Some of which includes MLMKL, DGP and DKL. Although, These methods have proven to be quite useful in various real-world settings. They still suffer from the limitations of only utilizing kernels from Hilbert spaces. In this paper, we address these shortcomings by introducing a new class of concatenated kernel learning methods that use the kernels from the reproducing kernel Banach spaces(RKBSs). These spaces turned out to be one of the most general spaces where a reproducing Kernel exists. We propose a framework of construction for these Deep RKBS models and then provide a representer theorem for regularized learning problems. We also describe the relationship with its deep RKHS variant as well as standard Deep Gaussian Processes. In the end, we construct and implement a two-layer deep RKBS model and demonstrate it on a range of machine learning tasks. | Mrityunjay Bhardwaj | null | null | 2,022 | iclr |
Where is the bottleneck in long-tailed classification? | null | A commonly held belief in deep-learning based long-tailed classification is that the representations learned from long-tailed data are ”good enough” and the performance bottleneck is the classification head atop the representation learner. We design experiments to investigate this folk wisdom, and find that representations learned from long-tailed data distributions substantially differ from the representations learned from ”normal” data distributions. We show that the long-tailed representations are volatile and brittle with respect to the true data distribution. Compared to the representations learned from the true, balanced distributions, long-tailed representations fail to localize tail classes and display vastly worse inter-class separation and intra-class compactness when unseen samples from the true data distribution are embedded into the feature space. We provide an explanation for why data augmentation helps long-tailed classification despite leaving the dataset imbalance unchanged — it promotes inter-class separation, intra-class compactness, and improves localization of tail classes w.r.t to the true data distribution. | Zaid Khan, Yun Fu | null | null | 2,022 | iclr |
How to Adapt Your Large-Scale Vision-and-Language Model | null | Pre-training large-scale vision and language models (e.g. CLIP) has shown promising results in representation and transfer learning. We investigate the question of how to efficiently adapt these models to downstream tasks. For image classification, linear probes have been the standard for ease of use and efficiency, while for language, other approaches like prompt tuning have emerged. We analyze several fine-tuning methods across a diverse set of image classification tasks across two spectra investigating the amount and similarity of downstream data to that of pretraining one. We find that just tuning LayerNorm parameters is a surprisingly effective baseline across the board. We further demonstrate a simple yet effective strategy that combines LayerNorm-tuning with general fine-tuning methods to improve their performance and benchmark them on few-shot adaption and distribution shift tasks. Finally, we provide an empirical analysis and recommend general recipes for efficient transfer learning of vision and language models. Website at https://sites.google.com/view/adapt-large-scale-models | Konwoo Kim, Michael Laskin, Igor Mordatch, Deepak Pathak | null | null | 2,022 | iclr |
Divisive Feature Normalization Improves Image Recognition Performance in AlexNet | null | Local divisive normalization provides a phenomenological description of many nonlinear response properties of neurons across visual cortical areas. To gain insight into the utility of this operation, we studied the effects on AlexNet of a local divisive normalization between features, with learned parameters. Developing features were arranged in a line topology, with the influence between features determined by an exponential function of the distance between them. We compared an AlexNet model with no normalization or with canonical normalizations (Batch, Group, Layer) to the same models with divisive normalization added. Divisive normalization always improved performance for models with batch or group or no normalization, generally by 1-2 percentage points, on both the CIFAR-100 and ImageNet databases. To gain insight into mechanisms underlying the improved performance, we examined several aspects of network representations. In the early layers both canonical and divisive normalizations reduced manifold capacities and increased average dimension of the individual categorical manifolds. In later layers the capacity was higher and manifold dimension lower for models roughly in order of their performance improvement. Examining the sparsity of activations across a given layer, divisive normalization layers increased sparsity, while the canonical normalization layers decreased it. Nonetheless, in the final layer, the sparseness of activity increased in the order of no normalization, divisive, com- bined, and canonical. We also investigated how the receptive fields (RFs) in the first convolutional layer (where RFs are most interpretable) change with normalization. Divisive normalization enhanced RF Fourier power at low wavelengths, while divisive+canonical enhanced power at mid (batch, group) or low (layer) wavelengths, compared to canonical alone or no normalization. In conclusion, divisive normalization enhances image recognition performance, most strongly when combined with canonical normalization, and in doing so it reduces manifold capacity and sparsity in early layers while increasing them in final layers, and increases low- or mid-wavelength power in the first-layer receptive fields. | Michelle Miller, SueYeon Chung, Kenneth D. Miller | null | null | 2,022 | iclr |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.