abstract
stringlengths 383
2.62k
| TLDR
stringlengths 4
250
|
|---|---|
A central task in computational drug discovery is to construct models from known active molecules to find further promising molecules for subsequent screening. However, typically only very few active molecules are known. Therefore, few-shot learning methods have the potential to improve the effectiveness of this critical phase of the drug discovery process. We introduce a new method for few-shot drug discovery. Its main idea is to enrich a molecule representation by knowledge about known context or reference molecules. Our novel concept for molecule representation enrichment is to associate molecules from both the support set and the query set with a large set of reference (context) molecules through a modern Hopfield network. Intuitively, this enrichment step is analogous to a human expert who would associate a given molecule with familiar molecules whose properties are known. The enrichment step reinforces and amplifies the covariance structure of the data, while simultaneously removing spurious correlations arising from the decoration of molecules. Our approach is compared with other few-shot methods for drug discovery on the FS-Mol benchmark dataset. On FS-Mol, our approach outperforms all compared methods and therefore sets a new state-of-the art for few-shot learning in drug discovery. An ablation study shows that the enrichment step of our method is the key to improve the predictive quality. In a domain shift experiment, we further demonstrate the robustness of our method. | We introduce a new architecture for few-shot learning in drug discovery that enriches molecule representations by retrieving from a large set of known molecules. |
Recently, the pretrain-tuning paradigm in large-scale sequence models has made significant progress in Natural Language Processing and Computer Vision. However, such a paradigm is still hindered by intractable challenges in Reinforcement Learning (RL), including the lack of self-supervised large-scale pretraining methods based on offline data and efficient fine-tuning/prompt-tuning over unseen downstream tasks. In this work, we explore how prompts can help sequence-modeling-based offline Reinforcement Learning (offline-RL) algorithms. Firstly, we propose prompt tuning for offline RL, where a context vector sequence is concatenated with the input to guide the conditional generation. As such, we can pretrain a model on the offline dataset with supervised loss and learn a prompt to guide the policy to play the desired actions. Secondly, we extend the framework to the Meta-RL setting and propose Contextual Meta Transformer (CMT), which leverages the context among different tasks as the prompt to improve the performance on unseen tasks. We conduct extensive experiments across three different offline-RL settings: offline single-agent RL on the D4RL dataset, offline Meta-RL on the MuJoCo benchmark, and offline MARL on the SMAC benchmark. The results validate the strong performance, and generality of our methods. | This paper explores how prompts help sequence-modeling based offline-RL algorithms |
Pretrained Language Models (LMs) memorize a vast amount of knowledge during initial pretraining, including information that may violate the privacy of personal lives and identities. Previous work addressing privacy issues for language models has mostly focused on data preprocessing and differential privacy methods, both requiring re-training the underlying LM. We propose knowledge unlearning as an alternative method to reduce privacy risks for LMs post hoc. We show that simply applying the unlikelihood training objective to target token sequences is effective at forgetting them with little to no degradation of general language modeling performances; it sometimes even substantially improves the underlying LM with just a few iterations. We also find that sequential unlearning is better than trying to unlearn all the data at once and that unlearning is highly dependent on which kind of data (domain) is forgotten. By showing comparisons with a previous data preprocessing method known to mitigate privacy risks for LMs, we show that unlearning can give a stronger empirical privacy guarantee in scenarios where the data vulnerable to extraction attacks are known a priori while being orders of magnitude more computationally efficient. We release the code and dataset needed to replicate our results at http://www.omitted.link/. | We propose knowledge unlearning for efficiently providing empirical privacy guarantees for large language models as an alternative solution to existing methods. |
Long-tailed distribution of semantic categories, which has been often ignored in conventional methods, causes unsatisfactory performance in semantic segmentation on tail categories. In this paper, we focus on the problem of long-tailed semantic segmentation. Although some long-tailed recognition methods (e.g., re-sampling/re-weighting) have been proposed in other problems, they are likely to compromise crucial contextual information in semantic segmentation. Therefore, these methods are hardly adaptable to the problem of long-tailed semantic segmentation. To address this problem, we propose a novel method, named MEDOE, by ensembling and grouping contextual information. Specifically, our MEDOE is a two-sage framework comprising a multi-expert decoder (MED) and a multi-expert output ensemble (MOE). The MED includes several ``experts", each of which takes as input the dataset masked according to the specific categories based on frequency distribution and generates contextual information self-adaptively for classification. The MOE then ensembles the experts' outputs with learnable decision weights. As a model-agnostic framework, MEDOE can be flexibly and efficiently coupled with various popular deep neural networks (e.g., Deeplabv3+, OCRNet, and PSPNet) to improve the performance in long-tailed semantic segmentation. Experimental results show that the proposed framework outperforms the current methods on both Cityscapes and ADE20K datasets by up to 2\% in mIoU and 6\% in mAcc. | We proposed MEDOE framework to address the long-tailed distribution in semantic segmentation |
It is often useful to compactly summarize important properties of a training dataset so that they can be used later without storing and/or iterating over the entire dataset. We consider a specific case of this: approximating the function space distance (FSD) over the training set, i.e. the average distance between the outputs of two neural networks. We propose an efficient approximation to FSD for ReLU neural networks based on approximating the architecture as a linear network with stochastic gating. Despite requiring only one parameter per unit of the network, our approach outcompetes other parametric approximations with larger memory requirements. Applied to continual learning, our parametric approximation is competitive with state-of-the-art nonparametric approximations which require storing many training examples. Furthermore, we show its efficacy in influence function estimation, allowing influence functions to be accurately estimated without iterating over the full dataset. | We propose an efficient parametric approximation of neural network function space distance that is memory-efficient and can be successfully applied to continual learning and influence function estimation tasks. |
Recent studies have shown that structural perturbations are significantly effective in degrading the accuracy of Graph Neural Networks (GNNs) in the semi-supervised node classification (SSNC) task. However, why the gradient-based methods are so destructive is rarely explored. In this work, we discover an interesting phenomenon: the adversarial edges are not uniformly distributed on the graph. Nearly all perturbations are generated around the training nodes in poisoning attack. Combined with this phenomenon, we provide an explanation for the effectiveness of the gradient-based attack method from a data distribution perspective and revisit both poisoning attack and evasion attack in SSNC. From this new perspective, we empirically and theoretically discuss some other attack tendencies. Based on the analysis, we provide nine practical tips on both attack and defense and meanwhile leverage them to improve existing attack and defense methods. Moreover, we design a fast attack method and a self-training defense method, which outperform the state-of-the-art methods and can effectively scale to large graphs like ogbn-arxiv. We conduct extensive experiments on four benchmark datasets to verify our claims. | We revisit graph adversarial attack and defense from a data distribution perspective. |
Generative models (e.g., generative adversarial network (GAN)) have advanced zero-shot learning (ZSL). Studies on the generative ZSL methods typically produce visual features of unseen classes to mitigate the issue of lacking unseen samples based on the predefined class semantic prototypes. As these empirically designed prototypes are not able to faithfully represent the actual semantic prototypes of visual features (i.e., visual prototypes), existing methods limit their ability to synthesize visual features that accurately represent real features and prototypes. We formulate this phenomenon as a visual-semantic domain shift problem. It prevents the generative models from further improving the ZSL performance. In this paper, we propose a dynamic semantic prototype learning (DSP) method to align the empirical and actual semantic prototypes for synthesizing accurate visual features. The alignment is conducted by jointly refining semantic prototypes and visual features so that the generator synthesizes visual features which are close to the real ones. We utilize a visual$\rightarrow$semantic mapping network (V2SM) to map both the synthesized and real features into the class semantic space. The V2SM benefits the generator to synthesize visual representations with rich semantics. The real/synthesized visual features supervise our visual-oriented semantic prototype evolving network (VOPE) where the predefined class semantic prototypes are iteratively evolved to become dynamic semantic prototypes. Such prototypes are then fed back to the generative network as conditional supervision. Finally, we enhance visual features by fusing the evolved semantic prototypes into their corresponding visual features. Our extensive experiments on three benchmark datasets show that our DSP improves existing generative ZSL methods, \textit{e.g.}, the average improvements of the harmonic mean over four baselines (e.g., CLSWGAN, f-VAEGAN, TF-VAEGAN and FREE) by 8.5\%, 8.0\% and 9.7\% on CUB, SUN and AWA2, respectively. | Dynamic Semantic Prototype should be Considered in Generative Zero-Shot Learning |
Kernel matrices, as well as weighted graphs represented by them, are ubiquitous objects in machine learning, statistics and other related fields. The main drawback of using kernel methods (learning and inference using kernel matrices) is efficiency -- given $n$ input points, most kernel-based algorithms need to materialize the full $n \times n$ kernel matrix before performing any subsequent computation, thus incurring $\Omega(n^2)$ runtime. Breaking this quadratic barrier for various problems has therefore, been a subject of extensive research efforts.
We break the quadratic barrier and obtain \emph{subquadratic} time algorithms for several fundamental linear-algebraic and graph processing primitives, including approximating the top eigenvalue and eigenvector, spectral sparsification, solving linear systems, local clustering, low-rank approximation, arboricity estimation and counting weighted triangles. We build on the recently developed Kernel Density Estimation framework, which (after preprocessing in time subquadratic in $n$) can return estimates of row/column sums of the kernel matrix. In particular, we develop efficient reductions from \emph{weighted vertex} and \emph{weighted edge sampling} on kernel graphs, \emph{simulating random walks} on kernel graphs, and \emph{importance sampling} on matrices to Kernel Density Estimation and show that we can generate samples from these distributions in \emph{sublinear} (in the support of the distribution) time. Our reductions are the central ingredient in each of our applications and we believe they may be of independent interest. We empirically demonstrate the efficacy of our algorithms on low-rank approximation (LRA) and spectral sparsification, where we observe a $\textbf{9x}$ decrease in the number of kernel evaluations over baselines for LRA and a $\textbf{41x}$ reduction in the graph size for spectral sparsification. | We give a framework for using recently developed tools for kernel density estimation to solve downstream kernel problems in sub-quadratic time. |
Machine learning applications frequently come with multiple diverse objectives and constraints that can change over time. Accordingly, trained models can be tuned with sets of hyper-parameters that affect their predictive behavior (e.g., their run-time efficiency versus error rate). As the number of constraints and hyper-parameter dimensions grow, naively selected settings may lead to sub-optimal and/or unreliable results. We develop an efficient method for calibrating models such that their predictions provably satisfy multiple explicit and simultaneous statistical guarantees (e.g., upper-bounded error rates), while also optimizing any number of additional, unconstrained objectives (e.g., total run-time cost). Building on recent results in distribution-free, finite-sample risk control for general losses, we propose Pareto Testing: a two-stage process which combines multi-objective optimization with multiple hypothesis testing. The optimization stage constructs a set of promising combinations on the Pareto frontier. We then apply statistical testing to this frontier only to identify configurations that have (a) high utility with respect to our objectives, and (b) guaranteed risk levels with respect to our constraints, with specifiably high probability. We demonstrate the effectiveness of our approach to reliably accelerate the execution of large-scale Transformer models in natural language processing (NLP) applications. In particular, we show how Pareto Testing can be used to dynamically configure multiple inter-dependent model attributes—including the number of layers computed before exiting, number of attention heads pruned, or number of text tokens considered—to simultaneously control and optimize various accuracy and cost metrics. | This paper presents a statistically efficient strategy for performing multiple hypothesis testing in order to find risk-controlling model configurations that are also useful with respect to arbitrary auxiliary objectives. |
Intrinsic and extrinsic factors jointly affect users' decisions in item selection (e.g., click, purchase). Intrinsic factors reveal users' real interests and are invariant in different contexts (e.g., time, weather), whereas extrinsic factors can change w.r.t. different contexts. Analyzing these two factors is an essential yet challenging task in recommender systems. However, in existing studies, factor analysis is either largely neglected, or designed for a specific context (e.g., the time context in sequential recommendation), which limits the applicability of such models. In this paper, we propose a generic model, IEDR, to learn intrinsic and extrinsic factors from various contexts for recommendation. IEDR contains two key components: a contrastive learning component, and a disentangling component. The two components collaboratively enable our model to learn context-invariant intrinsic factors and context-based extrinsic factors from all available contexts. Experimental results on real-world datasets demonstrate the effectiveness of our model in factor learning and impart a significant improvement in recommendation accuracy over the state-of-the-art methods. | We propose a recommender system that capture intrinsic and extrinsic factors from various contexts to enhance the recommendation quality. |
Animals thrive in a constantly changing environment and leverage the temporal structure to learn well-factorized causal representations. In contrast, traditional neural networks suffer from forgetting in changing environments and many methods have been proposed to limit forgetting with different trade-offs. Inspired by the brain thalamocortical circuit, we introduce a simple algorithm that uses optimization at inference time to generate internal representations of the current task dynamically. The algorithm alternates between updating the model weights and a latent task embedding, allowing the agent to parse the stream of temporal experience into discrete events and organize learning about them. On a continual learning benchmark, it achieves competitive end average accuracy by mitigating forgetting, but importantly, the interaction between the weights dynamics and the latent dynamics organizes knowledge into flexible structures with a cognitive interface to control them. Tasks later in the sequence can be solved through knowledge transfer as they become reachable within the well-factorized latent space. The algorithm meets many of the desiderata of an ideal continually learning agent in open-ended environments, and its simplicity suggests fundamental computations in circuits with abundant feedback control loops such as the thalamocortical circuits in the brain | A brain-inspired algorithm that alternates optimizing in weight space with optimizing the latent embedding space in the same neural network leading to open-ended discovery of tasks and disentangled learning. |
Probabilistic hierarchical time-series forecasting is an important variant of time-series forecasting, where the goal is to model and forecast multivariate time-series that have hierarchical relations. Previous works all assume rigid consistency over the given hierarchies and do not adapt to real-world data that show deviation from this assumption. Moreover, recent state-of-art neural probabilistic methods also impose hierarchical relations on point predictions and samples of distribution. This does not account for full forecast distributions being coherent with the hierarchy and leads to poorly calibrated forecasts. We close both these gaps and propose PROFHIT, a probabilistic hierarchical forecasting model that jointly models forecast distributions over the entire hierarchy. PROFHIT (1) uses a flexible probabilistic Bayesian approach and (2) introduces soft distributional coherency regularization that enables end-to-end learning of the entire forecast distribution leveraging information from the underlying hierarchy. This enables robust and calibrated forecasts as well as adaptation to real-life data with varied hierarchical consistency. PROFHIT provides 41-88% better performance in accuracy and 23-33% better calibration over a wide range of dataset consistency. Furthermore, PROFHIT can robustly provide reliable forecasts even if up to 10% of input time-series data is missing, whereas other methods’ performance severely degrade by over 70%. | A novel probabilistic neural model for calibrated and accurate probabilistic forecasting on datasets of varying hierarchical consistency. |
Decoupling representation learning and classifier learning has been shown to be effective in classification with long-tailed data. There are two main ingredients in constructing a decoupled learning scheme; 1) how to train the feature extractor for representation learning so that it provides generalizable representations and 2) how to re-train the classifier that constructs proper decision boundaries by handling class imbalances in long-tailed data. In this work, we first apply Stochastic Weight Averaging (SWA), an optimization technique for improving the generalization of deep neural networks, to obtain better generalizing feature extractors for long-tailed classification. We then propose a novel classifier re-training algorithm based on stochastic representation obtained from the SWA-Gaussian, a Gaussian perturbed SWA, and a self-distillation strategy that can harness the diverse stochastic representations based on uncertainty estimates to build more robust classifiers. Extensive experiments on CIFAR10/100-LT, ImageNet-LT, and iNaturalist-2018 benchmarks show that our proposed method improves upon previous methods both in terms of prediction accuracy and uncertainty estimation. | We propose a novel classifier re-training algorithm for long-tailed classification. |
Recent years have witnessed the success of applying seq2seq models to graph parsing tasks, where the outputs are compositionally structured (e.g., a graph or a tree). However, these seq2seq approaches pose a challenge in quantifying the model’s compositional uncertainty on graph structures due to the gap between seq2seq output probability and structural probability on the graph. This work is the first to quantify and evaluate compositional uncertainty for seq2seq graph parsing tasks. First, we proposed a generic, probabilistically interpretable framework that allows correspondences between seq2seq output probability to structural probability on the graph. This framework serves as a powerful medium for quantifying a seq2seq model's compositional uncertainty on graph elements (i.e., nodes or edges). Second, to evaluate uncertainty quality in terms of calibration, we propose a novel metric called Compositional Expected Calibration Error (CECE) which can measure a model’s calibration behavior in predicting graph structures. By a thorough evaluation for compositional uncertainty on three different tasks across ten domains, we demonstrate that CECE is a better reflection for distributional shift compared to vanilla sequence ECE. Finally, we validate the effectiveness of compositional uncertainty considering the task of collaborative semantic parsing, where the model is allowed to send limited subgraphs for human review. The results show that the collaborative performance based on uncertain subgraph selection consistently outperforms random subgraph selection (30% average error reduction rate) and performs comparably to oracle subgraph selection (only 0.33 difference in average prediction error), indicating that compositional uncertainty is an ideal signal for model errors and can benefit various downstream tasks. | In this paper, we aim to quantify and evaluate compositional uncertainty for seq2seq graph parsing by proposing a simple probabilistic framework and rigorous evaluation metrics. |
Modern, multi-branched neural network architectures often possess complex interconnections between layers, which we call coupled channels (CCs). Structured pruning of CCs in these multi-branch networks is an under-researched problem, as most existing works are typically designed for pruning single-branch models like VGG-nets. While these methods yield accurate subnetworks, the improvements in inference times when applied to multi-branch networks are comparatively modest, as these methods do not prune CCs, which we observe contribute significantly to inference time. For instance, layers with CCs as input or output take more than 66% of the inference time in ResNet-50. Moreover, pruning in the data-free regime, where data is not used for pruning, is gaining traction owing to privacy concerns and computational costs associated with fine-tuning. Motivated by this, we study the problem of pruning CCs in the data-free regime. To facilitate the development of algorithms to prune CCs, we define Data Flow Couplings (DFCs) to enumerate the layers that constitute coupled connections and the associated transformation. Additionally, saliencies for pruning CCs cannot be gauged in isolation, as there may be discrepancies among the layerwise importance of CCs using conventional scoring strategies. This necessitates finding grouped saliencies to gauge the importance of all corresponding coupled elements in a network. We thus propose the Backwards Graph-based Saliency Computation (BGSC) algorithm, a data-free method that computes saliencies by estimating an upper bound to the reconstruction error of intermediate layers; we call this pruning strategy Data Flow driven Pruning of Coupled channels (DFPC). Finally, we show the efficacy of DFPC for models trained on standard datasets. Since we pruned coupled channels, we achieve up to 1.66x improvements in inference time for ResNet-101 trained on CIFAR-10 with a 5% accuracy drop without fine-tuning. With access to the ImageNet training set, we achieve significant improvements over the data-free method and see an improvement of at least 47.1% in speedup for a 2.3% accuracy drop for ResNet-50 against our baselines. | We propose a novel data-free algorithm to accelerate neural networks via pruning coupled channels. |
We propose a new approach, based on discrete filter pruning, to adapt off-the-shelf models into an embedded
environment. Importantly, we circumvent the usually prohibitive costs of model compression. Our method, Structured
Coarse Block Pruning (SCBP), prunes whole CNN kernels using width modifiers applied to a novel transformation of
convlayers into superblocks. SCBP uses set representations to construct a rudimentary search to provide candidate
networks. To test our approach, the original ResNet architectures serve as the baseline and also provide the 'seeds'
for our candidate search. The search produces a configurable number of compressed (derived) models. These derived models
are often ~20\% faster and ~50\% smaller than their unmodified counterparts. At the expense of accuracy, the size can
become even smaller and the inference latency lowered even further. The unique SCBP transformations yield many new model
variants, each with their own trade-offs, and does not require GPU clusters or expert humans for training and design. | convnet compression that is fast, not resource hungry and uses width modifiers applied with a new twist. |
The text retrieval task is mainly performed in two ways: the bi-encoder approach and the generative approach. The bi-encoder approach maps the document and query embeddings to common vector space and performs a nearest neighbor search. It stably shows high performance and efficiency across different domains but has an embedding space bottleneck as it interacts in L2 or inner product space. The generative retrieval model retrieves by generating a target sequence and overcomes the embedding space bottleneck by interacting in the parametric space. However, it fails to retrieve the information it has not seen during the training process as it depends solely on the information encoded in its own model parameters. To leverage the advantages of both approaches, we propose Contextualized Generative Retrieval model, which uses contextualized embeddings (output embeddings of a language model encoder) as vocab embeddings at the decoding step of generative retrieval. The model uses information encoded in both the non-parametric space of contextualized token embeddings and the parametric space of the generative retrieval model. Our approach of generative retrieval with contextualized vocab embeddings shows higher performance than generative retrieval with only vanilla vocab embeddings in the document retrieval task, an average of 6% higher performance in KILT (NQ, TQA) and 2X higher in NQ-320k, suggesting the benefits of using contextualized embedding in generative retrieval models. | By utilizing contextualized token embeddings in generative retrieval, it can utilize both the parametric space of the model and the non-parametric space of contextualized embeddings. |
The appearance of adversarial examples raises attention from both academia and industry. Along with the attack-defense arms race, adversarial training is the most effective against adversarial examples.
However, we find inequality phenomena occur during the $l_{\infty}$-adversarial training, that few features dominate the prediction made by the adversarially trained model. We systematically evaluate such inequality phenomena by extensive experiments and find such phenomena become more obvious when performing adversarial training with increasing adversarial strength (evaluated by $\epsilon$). We hypothesize such inequality phenomena make $l_{\infty}$-adversarially trained model less reliable than the standard trained model when few ``important features" are influenced. To validate our hypothesis, we proposed two simple attacks that either perturb or replace important features with noise or occlusion. Experiments show that $l_{\infty}$-adversarially trained model can be easily attacked when the few important features are influenced.
Our work shed light on the limitation of the practicality of $l_{\infty}$-adversarial training. | We find an intriguing phenomena of $l_{\infty}$ adversarial training, and this phenomena brings unrealized threats to adversarially trained model. |
Neural-symbolic AI (NeSy) methods allow neural networks to exploit symbolic background knowledge. NeSy has been shown to aid learning in the limited data regime and to facilitate inference on out-of-distribution data. Neural probabilistic logic programming (NPLP) is a popular NeSy approach that integrates probabilistic models with neural networks and logic programming. A major limitation of current NPLP systems, such as DeepProbLog, is their restriction to discrete and finite probability distributions, e.g., binary random variables. To overcome this limitation, we introduce DeepSeaProbLog, an NPLP language that supports discrete and continuous random variables on (possibly) infinite and even uncountable domains. Our main contributions are 1) the introduction of DeepSeaProbLog and its semantics, 2) an implementation of DeepSeaProbLog that supports inference and gradient-based learning, and 3) an experimental evaluation of our approach. | DeepSeaProbLog: a neural probabilistic logic programming language with discrete and continuous random variables. |
Recursive Neural Networks (RvNNs) generalize Recurrent Neural Networks (RNNs) by allowing sequential composition in a more flexible order, typically, based on some tree structure. While initially user-annotated tree structures were used, in due time, several approaches were proposed to automatically induce tree-structures from raw text to guide the recursive compositions in RvNNs. In this paper, we present an approach called Beam Tree Recursive Cell (or BT-Cell) based on a simple yet overlooked backpropagation-friendly framework. BT-Cell applies beam search on easy-first parsing for simulating RvNNs with automatic structure-induction. Our results show that BT-Cell achieves near-perfect performance on several aspects of challenging structure-sensitive synthetic tasks like ListOps and also comparable performance in realistic data to other RvNN-based models. We further introduce and analyze several extensions of BT-Cell based on relaxations of the hard top-k operators in beam search. We evaluate the models in different out of distribution splits in both synthetic and realistic data. Additionally, we identify a previously unknown failure case for neural models in generalization to unseen number of arguments in ListOps. We will release our code. | We apply beam search on easy-parsing strategy to simulate RvNN without ground truth tree supervision and experiment its different extensions. |
Despite the widespread use of unsupervised models, very few methods are designed to explain them. Most explanation methods explain a scalar model output. However, unsupervised models output representation vectors, the elements of which are not good candidates to explain because they lack semantic meaning. To bridge this gap, recent works defined a scalar explanation output: a dot product-based similarity in the representation space to the sample being explained (i.e., an explicand). Although this enabled explanations of unsupervised models, the interpretation of this approach can still be opaque because similarity to the explicand's representation may not be meaningful to humans. To address this, we propose contrastive corpus similarity, a novel and semantically meaningful scalar explanation output based on a reference corpus and a contrasting foil set of samples. We demonstrate that contrastive corpus similarity is compatible with many post-hoc feature attribution methods to generate COntrastive COrpus Attributions (COCOA) and quantitatively verify that features important to the corpus are identified. We showcase the utility of COCOA in two ways: (i) we draw insights by explaining augmentations of the same image in a contrastive learning setting (SimCLR); and (ii) we perform zero-shot object localization by explaining the similarity of image representations to jointly learned text representations (CLIP). | A novel method to explain representations (from unsupervised and supervised models) in terms of input features. |
Constraint satisfaction problems (CSPs) are about finding values of variables that satisfy the given constraints. We show that Transformer extended with recurrence is a viable approach to learning to solve CSPs in an end-to-end manner, having clear advantages over state-of-the-art methods such as Graph Neural Networks, SATNet, and some neuro-symbolic models. With the ability of Transformer to handle visual input, the proposed Recurrent Transformer can straightforwardly be applied to visual constraint reasoning problems while successfully addressing the symbol grounding problem. We also show how to leverage deductive knowledge of discrete constraints in the Transformer's inductive learning to achieve sample-efficient learning and semi-supervised learning for CSPs. | We show how Recurrent Transformer can be trained to perform logical reasoning on constraint satisfaction problems in an end-to-end manner and how to inject discrete constraints into its training |
Offline reinforcement learning (RL) aims at learning an effective policy from offline datasets without active interactions with the environment. The major challenge of offline RL is the distribution shift that appears when out-of-distribution actions are queried, which makes the policy improvement direction biased by extrapolation errors. Most existing methods address this problem by penalizing the policy for deviating from the behavior policy during policy improvement or making conservative updates for value functions during policy evaluation. In this work, we propose a novel MISA framework to approach offline RL from the perspective of Mutual Information between States and Actions in the dataset by directly constraining the policy improvement direction. Intuitively, mutual information measures the mutual dependence of actions and states, which reflects how a behavior agent reacts to certain environment states during data collection. To effectively utilize this information to facilitate policy learning, MISA constructs lower bounds of mutual information parameterized by the policy and Q-values. We show that optimizing this lower bound is equivalent to maximizing the likelihood of a one-step improved policy on the offline dataset. In this way, we constrain the policy improvement direction to lie in the data manifold. The resulting algorithm simultaneously augments the policy evaluation and improvement by adding a mutual information regularization. MISA is a general offline RL framework that unifies conservative Q-learning (CQL) and behavior regularization methods (e.g., TD3+BC) as special cases. Our experiments show that MISA performs significantly better than existing methods and achieves new state-of-the-art on various tasks of the D4RL benchmark. | We propose MISA, a general framework for offline reinforcemen learning with mutual information regularization |
Generating new molecules with specified chemical and biological properties via generative models has emerged as a promising direction for drug discovery. However, existing methods require extensive training/fine-tuning with a large dataset, often unavailable in real-world generation tasks. In this work, we propose a new retrieval-based framework for controllable molecule generation. We use a small set of exemplar molecules, i.e., those that (partially) satisfy the design criteria, to steer the pre-trained generative model towards synthesizing molecules that satisfy the given design criteria. We design a retrieval mechanism that retrieves and fuses the exemplar molecules with the input molecule, which is trained by a new self-supervised objective that predicts the nearest neighbor of the input molecule. We also propose an iterative refinement process to dynamically update the generated molecules and retrieval database for better generalization. Our approach is agnostic to the choice of generative models and requires no task-specific fine-tuning. On various tasks ranging from simple design criteria to a challenging real-world scenario for designing lead compounds that bind to the SARS-CoV-2 main protease, we demonstrate our approach extrapolates well beyond the retrieval database, and achieves better performance and wider applicability than previous methods. | We propose a first-of-its-kind retrieval-based framework for controllable molecule generation which can effectively extrapolate beyond the retrieval database and achieves state-of-the-art performance on various benchmarks. |
Deep neural networks have been shown to be vulnerable to small perturbations of their inputs known as adversarial attacks. In this paper, we consider the particular task of Neural Machine Translation (NMT), where security is often critical. We investigate the vulnerability of NMT models to adversarial attacks and propose a new attack algorithm called TransFool. It builds on a multi-term optimization problem and a gradient projection step to compute adversarial examples that fool NMT models. By integrating the embedding representation of a language model in the proposed attack, we generate fluent adversarial examples in the source language that maintain a high level of semantic similarity with the clean samples and render the attack largely undetectable. Experimental results demonstrate that, for multiple translation tasks and different NMT architectures, our white-box attack can severely degrade the translation quality for more than 60% of the sentences while the semantic similarity between the original sentence and the adversarial example stays very high. Moreover, we show that the proposed attack is transferable to unknown target models and can fool those quite easily. Finally, our method leads to improvement in terms of success rate, semantic similarity, and fluency compared to the existing attack strategies both in white-box and black-box settings. Hence, TransFool permits to better characterize the vulnerability of NMT systems and outlines the necessity to design strong defense mechanisms and more robust NMT systems for real-life applications. | We propose TransFool to build adversarial attacks against neural machine translation systems, which are fluent sentences and semantically similar to the original sentence, but highly degrade the translation quality. |
Commonly used optimization algorithms often show a trade-off between good generalization and fast training times. For instance, stochastic gradient descent (SGD) tends to have good generalization; however, adaptive gradient methods have superior training times. Momentum can help accelerate training with SGD, but so far there has been no principled way to select the momentum hyperparameter. Here we study implicit bias arising from the interplay between SGD with label noise and momentum in the training of overparameterized neural networks. We find that scaling the momentum hyperparameter $1-\beta$ with the learning rate to the power of $2/3$ maximally accelerates training, without sacrificing generalization. To analytically derive this result we develop an architecture-independent framework, where the main assumption is the existence of a degenerate manifold of global minimizers, as is natural in overparameterized models. Training dynamics display the emergence of two characteristic timescales that are well-separated for generic values of the hyperparameters. The maximum acceleration of training is reached when these two timescales meet, which in turn determines the scaling limit we propose. We perform experiments, including matrix sensing and ResNet on CIFAR10, which provide evidence for the robustness of these results. | We find the implicit bias induced by noise in SGD with momentum; this leads us to identify a scaling limit of the momentum hyperparameter in the learning rate that maximally accelerates training, without depleting generalization. |
Variational Autoencoders (VAEs) were originally motivated as probabilistic generative models in which one performs approximate Bayesian inference. The proposal of $\beta$-VAEs breaks this interpretation and generalizes VAEs to application domains beyond generative modeling (e.g., representation learning, clustering, or lossy data compression) by introducing an objective function that allows practitioners to trade off between the information content ("bit rate") of the latent representation and the distortion of reconstructed data. In this paper, we reconsider this rate/distortion trade-off in the context of hierarchical VAEs, i.e., VAEs with more than one layer of latent variables. We propose a method to control each layer's contribution to the rate independently. We identify the most general class of inference models to which our proposed method is applicable, and we derive theoretical bounds on the performance of downstream tasks as functions of the individual layers' rates. Our experiments demonstrate that the proposed method allows us to better tune hierarchical VAEs for a diverse set of practical use cases. | We generalize the rate/distortion theory of VAEs and analyze both theoeretically and analytically how manipulating each individual layer's rate affects performance. |
We present a novel, conditional generative probabilistic model of set-valued data with a tractable log density. This model is a continuous normalizing flow governed by permutation equivariant dynamics. These dynamics are driven by a learnable per-set-element term and pairwise interactions, both parametrized by deep neural networks. We illustrate the utility of this model via applications including (1) complex traffic scene generation conditioned on visually specified map information, and (2) object bounding box generation conditioned directly on images. We train our model by maximizing the expected likelihood of labeled conditional data under our flow, with the aid of a penalty that ensures the dynamics are smooth and hence efficiently solvable. Our method significantly outperforms non-permutation invariant baselines in terms of log likelihood and domain-specific metrics (offroad, collision, and combined infractions), yielding realistic samples that are difficult to distinguish from real data. | We present a novel, conditional generative probabilistic model of set-valued data with a tractable log density. |
Continuous neural representations have recently emerged as a powerful and flexible alternative to classical discretized representations of signals. However, training them to capture fine details in multi-scale signals is difficult and computationally expensive. Here we propose random weight factorization as a simple drop-in replacement for parameterizing and initializing conventional linear layers in coordinate-based multi-layer perceptrons (MLPs) that significantly accelerates and improves their training. We show how this factorization alters the underlying loss landscape and effectively enables each neuron in the network to learn using its own self-adaptive learning rate. This not only helps with mitigating spectral bias, but also allows networks to quickly recover from poor initializations and reach better local minima. We demonstrate how random weight factorization can be leveraged to improve the training of neural representations on a variety of tasks, including image regression, shape representation, computed tomography, inverse rendering, solving partial differential equations, and learning operators between function spaces. | A simple drop-in replacement of conventional dense layers for accelerating and improving the training of coordinate-based multi-layer perceptrons (MLPs). |
Models using structured state space sequence (S4) layers have achieved state-of-the-art performance on long-range sequence modeling tasks. An S4 layer combines linear state space models (SSMs), the HiPPO framework, and deep learning to achieve high performance. We build on the design of the S4 layer and introduce a new state space layer, the S5 layer. Whereas an S4 layer uses many independent single-input, single-output SSMs, the S5 layer uses one multi-input, multi-output SSM. We establish a connection between S5 and S4, and use this to develop the initialization and parameterization used by the S5 model. The result is a state space layer that can leverage efficient and widely implemented parallel scans, allowing S5 to match the computational efficiency of S4, while also achieving state-of-the-art performance on several long-range sequence modeling tasks. S5 averages $87.4\%$ on the long range arena benchmark, and $98.5\%$ on the most difficult Path-X task. | We introduce a new state space sequence modeling layer, building on the recent S4 layer, that increases the state of the art on many long-range benchmark tasks. |
Pretrained language models have shown superior performance on many natural language processing tasks, yet they still struggle at multi-step formal reasoning tasks like grade school math problems. One key challenge of finetuning them to solve such math reasoning problems is that many existing datasets only contain one reference solution for each problem, despite the fact that there are often alternative solutions resembling different reasoning paths to the final answer. This way, the finetuned models are biased towards the limited reference solutions, which limits their generalization to unseen examples. To mitigate this issue, we propose to let the model perform sampling during training and learn from both self-sampled fully-correct solutions, which yield the correct answer upon execution, and partially-correct solutions, whose intermediate state matches an intermediate state of a known correct solution. We show that our use of self-sampled correct and partially-correct solutions can benefit learning and help guide the sampling process, leading to more efficient exploration of the solution space. Additionally, we explore various training objectives to support learning from multiple solutions per example and find they greatly affect the performance. Experiments on two math reasoning datasets show the effectiveness of our method compared to learning from a single reference solution with MLE, where we improve PASS@100 from 35.5% to 44.5% for GSM8K, and 27.6% to 36.2% PASS@80 for MathQA. Such improvements are also consistent across different model sizes. | We propose to let pretrained language models sample additional solutions for each problem and learn from the self-sampled solutions that are correct or partially-correct. |
We propose a new paradigm to help Large Language Models (LLMs) generate more accurate factual knowledge without retrieving from an external corpus, called RECITation-augmented gEneration (RECITE). Different from retrieval-augmented language models that retrieve relevant documents before generating the outputs, given an input, RECITE first recites one or several relevant passages from LLMs’ own memory via sampling, and then produces the final answers. We show that RECITE is a powerful paradigm for knowledge-intensive NLP tasks. Specifically, we show that by utilizing recitation as the intermediate step, a recite-and-answer scheme can achieve new state-of-the-art performance in various closed-book question answering (CBQA) tasks. In experiments, we verify the effectiveness of RECITE on three pre-trained models (In-house LM, UL2, and OPT) and three CBQA tasks (Natural Questions, TriviaQA, and HotpotQA). Our code is available at "https://github.com/Edward-Sun/RECITE". | We propose a novel recitation-augmented generation framework to improve language models’ performance in the closed-book question-answering setting. |
Contemporary applications of machine learning raise important and overlooked theoretical questions regarding optimization in two-team games. Formally, two-team zero-sum games are defined as multi-player games where players are split into two competing sets of agents, each experiencing a utility identical to that of their teammates and opposite to that of the opposing team. We focus on the solution concept of Nash equilibria and prove $\textrm{CLS}$-hardness of computing them in this class of games. To further examine the capabilities of online learning algorithms in games with full-information feedback, we propose a benchmark of a simple ---yet nontrivial--- family of such games. These games do not enjoy the properties used to prove convergence for relevant algorithms. In particular, we use a dynamical systems perspective to demonstrate that gradient descent-ascent, its optimistic variant, optimistic multiplicative weights update, and extra gradient fail to converge (even locally) to a Nash equilibrium. On a brighter note, we propose a first-order method that leverages control theory techniques and under some conditions enjoys last-iterate local convergence to a Nash equilibrium. We also believe our proposed method is of independent interest for general min-max optimization. | Common no-regret algorithms fail to converge to a Nash equilibrium in two-team zero-sum games but a novel approach does converge locally. |
Recent advanced works achieve fair representations and predictions through regularization, adversarial debiasing, and contrastive learning in graph neural networks (GNNs). These methods \textit{implicitly} encode the sensitive attribute information in the well-trained model weight via \textit{backward propagation}. In practice, we not only pursue a fair machine learning model but also lend such fairness perception to the public. For current fairness methods,
how the sensitive attribute information usage makes the model achieve fair prediction still remains a black box. In this work, we first propose the concept \textit{transparency} to describe \textit{whether} the model embraces the ability of lending fairness perception to the public \textit{or not}. Motivated by the fact that current fairness models lack of transparency, we aim to pursue a fair machine learning model with transparency via \textit{explicitly} rendering sensitive attribute usage for fair prediction in \textit{forward propagation} . Specifically, we develop an effective and transparent \textsf{F}air \textsf{M}essage \textsf{P}assing (FMP) scheme adopting sensitive attribute information in forward propagation. In this way, FMP explicitly uncovers how sensitive attributes influence final prediction. Additionally, FMP scheme can aggregate useful information from neighbors and mitigate bias in a unified framework to simultaneously achieve graph smoothness and fairness objectives. An acceleration approach is also adopted to improve the efficiency of FMP. Experiments on node classification tasks demonstrate that the proposed FMP outperforms the state-of-the-art baselines in terms of fairness and accuracy on three real-world datasets. The code is available in {\color{blue}\url{https://anonymous.4open.science/r/FMP-AD84}}. | We aim to achieve fair message passsing with transparency to explictly use sensitive attributes in forward progagation instead of backward propagation.. |
One main challenge in federated learning is the large communication cost of exchanging weight updates from clients to the server at each round. While prior work has made great progress in compressing the weight updates through gradient compression methods, we propose a radically different approach that does not update the weights at all. Instead, our method freezes the weights at their initial \emph{random} values and learns how to sparsify the random network for the best performance. To this end, the clients collaborate in training a \emph{stochastic} binary mask to find the optimal sparse random network within the original one. At the end of the training, the final model is a sparse network with random weights -- or a subnetwork inside the dense random network. We show improvements in accuracy, communication (less than $1$ bit per parameter (bpp)), convergence speed, and final model size (less than $1$ bpp) over relevant baselines on MNIST, EMNIST, CIFAR-10, and CIFAR-100 datasets, in the low bitrate regime. | We propose an FL framework, where clients find a sparse random network using a stochastic strategy; and provide (1) lower communication cost, (2) higher accuracy, (3) faster convergence, and (4) at the end of the training, a compressed final model. |
Many fundamental properties of a quantum system are captured by its Hamiltonian and ground state. Despite the significance, ground states preparation (GSP) is classically intractable for large-scale Hamiltonians. Quantum neural networks (QNNs), which exert the power of modern quantum machines, have emerged as a leading protocol to conquer this issue. As such, the performance enhancement of QNNs becomes the core in GSP. Empirical evidence showed that QNNs with handcraft symmetric ans\"atze generally experience better trainability than those with asymmetric ans\"atze, while theoretical explanations remain vague. To fill this knowledge gap, here we propose the effective quantum neural tangent kernel (EQNTK) and connect this concept with over-parameterization theory to quantify the convergence of QNNs towards the global optima. We uncover that the advance of symmetric ans\"atze attributes to their large EQNTK value with low effective dimension, which requests few parameters and quantum circuit depth to reach the over-parameterization regime permitting a benign loss landscape and fast convergence. Guided by EQNTK, we further devise a symmetric pruning (SP) scheme to automatically tailor a symmetric ansatz from an over-parameterized and asymmetric one to greatly improve the performance of QNNs when the explicit symmetry information of Hamiltonian is unavailable. Extensive numerical simulations are conducted to validate the analytical results of EQNTK and the effectiveness of SP. | We prove how the symmetry enhances the training performance of QNNs and then devise an efficient symmetric pruning scheme to distill a symmetric ansatz from an over-parameterized and asymmetric ansatz. |
Detecting anomalies and the corresponding root causes in multivariate time series plays an important role in monitoring the behaviors of various real-world systems, e.g., IT system operations or manufacturing industry. Previous anomaly detection approaches model the joint distribution without considering the underlying mechanism of multivariate time series, making them computationally hungry and hard to identify root causes. In this paper, we formulate the anomaly detection problem from a causal perspective and view anomalies as instances that do not follow the regular causal mechanism to generate the multivariate data. We then propose a causality-based framework for detecting anomalies and root causes. It first learns the causal structure from data and then infers whether an instance is an anomaly relative to the local causal mechanism whose conditional distribution can be directly estimated from data. In light of the modularity property of causal systems (the causal processes to generate different variables are irrelevant modules), the original problem is divided into a series of separate, simpler, and low-dimensional anomaly detection problems so that where an anomaly happens (root causes) can be directly identified. We evaluate our approach with both simulated and public datasets as well as a case study on real-world AIOps applications, showing its efficacy, robustness, and practical feasibility. | A causality-based approach for anomaly detection and root cause analysis |
Very large state spaces with a sparse reward signal are difficult to explore. The lack of a sophisticated guidance results in a poor performance for numerous reinforcement learning algorithms. In these cases, the commonly used random exploration is often not helpful. The literature shows that this kind of environments require enormous efforts to systematically explore large chunks of the state space. Learned state representations can help here to improve the search by providing semantic context and build a structure on top of the raw observations. In this work we introduce a novel time-myopic state representation that clusters temporal close states together while providing a time prediction capability between them. By adapting this model to the Go-Explore paradigm (Ecoffet et al., 2021b), we demonstrate the first learned state representation that reliably estimates novelty instead of using the hand-crafted representation heuristic. Our method shows an improved solution for the detachment problem which still remains an issue at the Go-Explore Exploration Phase. We provide evidence that our proposed method covers the entire state space with respect to all possible time trajectories — without causing disadvantageous conflict-overlaps in the cell archive. Analogous to native Go-Explore, our approach is evaluated on the hard exploration environments MontezumaRevenge, Gravitar and Frostbite (Atari) in order to validate its capabilities on difficult tasks. Our experiments show that time-myopic Go-Explore is an effective alternative for the domain-engineered heuristic while also being more general. The source code of the method is available on GitHub. | To add flexibility to and possibly improve Go-Explore we study how its representation heuristic can be replaced by a time-predicting neural network. |
Clone detection is a well known task, which could be formulated on any programming language. Although to the best of our knowledge there is no cross-lingual clone detection task formulated. In this work we formulate such a task alongside with a specific training procedure CCT for a deep leaning language model. This procedure allows CCT-trained model to outperform the existing approaches on POJ-104 benchmark with result of 95.67\% MAP and on newly created cross-lingual clone detection benchmark XCD. Moreover, CCT model shows new state of the art results in code search task AdvTest 47.15\% MRR. | We present a novel approach for pre-training the language models for better code and text representation improving results in code search and clone detection. |
Vision-based ego-centric 3D human pose estimation (ego-HPE) is essential to support critical applications of xR-technologies. However, severe self-occlusions and strong distortion introduced by the fish-eye view from the head mounted camera, make ego-HPE extremely challenging. While current state-of-the-art (SOTA) methods try to address the distortion, they still suffer from large errors in the most critical joints (such as hands) due to self-occlusions. To this end, we propose a spatio-temporal transformer model that can attend to semantically rich feature maps obtained from popular convolutional backbones. Leveraging the complex spatio-temporal information encoded in ego-centric videos, we design a spatial concept called feature map tokens (FMT) which can attend to all the other spatial units in our spatio-temporal feature maps. Powered by this FMT-based transformer, we build Egocentric Spatio-Temporal Self-Attention Network (Ego-STAN), which uses heatmap-based representations and spatio-temporal attention specialized to address distortions and self-occlusions in ego-HPE.
Our quantitative evaluation on the contemporary sequential xR-EgoPose dataset, achieves a 38.2% improvement on the highest error joints against the SOTA ego-HPE model, while accomplishing a 22% decrease in the number of parameters. Finally, we also demonstrate the generalization capabilities of our model to real-world HPE tasks beyond ego-views. | spatio-temporal egocentric pose estimation using transformers. |
Semi-supervised learning is a critical tool in reducing machine learning’s dependence on labeled data. It has been successfully applied to structure data, such as image and language data, by exploiting the inherent spatial and semantic structure therein with pretrained models or data augmentation. Some of these methods are no longer applicable for the data where domain structures are not available because the pretrained models or data augmentation can not be used.
Due to simplicity, existing pseudo-labeling (PL) methods can be widely used without any domain assumption, but are vulnerable to noise samples and to greedy assignments given a predefined threshold which is typically unknown. This paper addresses this problem by proposing a Confident Sinkhorn Allocation (CSA), which assigns labels to only samples with high confidence scores and learns the best label allocation via optimal transport. CSA outperforms the current state-of-the-art in this practically important area of semi-supervised learning. | a new pseudo-labeling method for semi-supervised learning without domain knowledge |
The method of choice for integrating the time-dependent Fokker-Planck equation in high-dimension is to generate samples from the solution via integration of the associated stochastic differential equation. Here, we introduce an alternative scheme based on integrating an ordinary differential equation that describes the flow of probability. Acting as a transport map, this equation deterministically pushes samples from the initial density onto samples from the solution at any later time. Unlike integration of the stochastic dynamics, the method has the advantage of giving direct access to quantities that are challenging to estimate from trajectories alone, such as the probability current, the density itself, and its entropy. The probability flow equation depends on the gradient of the logarithm of the solution (its "score"), and so is a-priori unknown. To resolve this dependence, we model the score with a deep neural network that is learned on-the-fly by propagating a set of samples according to the instantaneous probability current. We consider several high-dimensional examples from the physics of interacting particle systems to highlight the efficiency and precision of the approach; we find that the method accurately matches analytical solutions computed by hand and moments computed via Monte-Carlo. | We develop an efficient method to solve the Fokker-Planck equation in high-dimension by learning the score of the solution. |
Devising a fair classifier that does not discriminate against different groups is an important problem in machine learning. Although researchers have proposed various ways of defining group fairness, most of them only focused on the immediate fairness, ignoring the long-term impact of a fair classifier under the dynamic scenario where each individual can improve its feature over time. Such dynamic scenarios happen in real world, e.g., college admission and credit loaning, where each rejected sample makes effort to change its features to get accepted afterwards. In this dynamic setting, the long-term fairness should equalize the samples’ feature distribution across different groups after the rejected samples make some effort to improve. In order to promote long-term fairness, we propose a new fairness notion called Equal Improvability (EI), which equalizes the potential acceptance rate of the rejected samples across different groups assuming a bounded level of effort will be spent by each rejected sample. We analyze the properties of EI and its connections with existing fairness notions. To find a classifier that satisfies the EI requirement, we propose and study three different approaches that solve EI regularized optimization problems. Through experiments on both synthetic and real datasets, we demonstrate that the proposed EI-regularized algorithms encourage us to find a fair classifier in terms of EI. Finally, we provide experimental results on dynamic scenarios which highlight the advantages of our EI metric in achieving the long-term fairness. Codes are available in anonymous GitHub repository. | We propose a new group fairness notion called Equal Improvability that equalizes the improvement of the rejected individuals across different groups. |
Anomaly detection from medical images is an important task for clinical screening and diagnosis. In general, a large dataset of normal images are available while only few abnormal images can be collected in clinical practice. By mimicking the diagnosis process of radiologists, we attempt to tackle this problem by learning a tractable distribution of normal images and identify anomalies by differentiating the original image and the reconstructed normal image. More specifically, we propose a normalizing flow-based autoencoder for an efficient and tractable representation of normal medical images. The anomaly score consists of the likelihood originated from the normalizing flow and the reconstruction error of the autoencoder, which allows to identify the abnormality and provide an interpretability at both image and pixel levels. Experimental evaluation on two medical images datasets showed that the proposed model outperformed the other approaches by a large margin, which validated the effectiveness and robustness of the proposed method. | We propose a normalizing flow based autoencoder for medical anomaly detection and it outperformed the other approaches by a large margin. |
Deep Neural network (DNN) pruning is an effective method to reduce the size of a model, improve the inference latency, and minimize the power consumption on DNN accelerators, at the risk of decreasing model accuracy. In this paper, we propose a novel differentiable pruning scheme, Iterative Differentiable Pruning or IDP which offers state-of-the-art qualities in model size, accuracy, and training cost. IDP creates attention-based pruning masks for a given sparsity target to achieve the state-of-the-art trade-offs between model accuracy and inference compute with negligible training overhead. We evaluated IDP on various computer vision and natural language processing tasks, and found that IDP delivers the state-of-the-art results. For MobileNet-v1 (which is a challenging DNN for pruning), IDP can achieve 68.2% top-1 ImageNet1k accuracy with 86.6% sparsity which is 2.3% higher accuracy than the latest state-of-the-art pruning algorithms. For ResNet18, IDP offers 69.5% top-1 ImageNet1k accuracy with 85.5% sparsity at the same training budget and 0.8% better top-1 accuracy than the state-of-the-art method. Also, IDP demonstrates over 83.1% accuracy on Multi-Genre Natural Language Inference with 90% sparsity for BERT, while the next best from the existing techniques shows 81.5% accuracy. | We proposed a differentiable pruning method, IDP which yields the state-of-the-art pruning quality on popular computer vision and natural language models, based on attention-based soft-mask. |
Transferring adversarial examples from surrogate (ML) models to evade target models is a common method for evaluating adversarial robustness in black-box settings. Researchers have invested substantial efforts to enhance transferability. Chiefly, attacks leveraging data augmentation have been found to help adversarial examples generalize better from surrogate to target models. Still, prior work has explored a limited set of augmentation techniques and their composition. To fill the gap, we conducted a systematic, comprehensive study of how data augmentation affects transferability. Particularly, we explored ten augmentation techniques of six categories originally proposed to help ML models generalize to unseen benign samples, and assessed how they influence transferability, both when applied individually and when composed. Our extensive experiments with the ImageNet dataset showed that simple color-space augmentations (e.g., color to greyscale) outperform the state of the art when combined with standard augmentations, such as translation and scaling. Additionally, except for two methods that may harm transferability, we found that composing augmentation methods impacts transferability monotonically (i.e., more methods composed $\rightarrow$ $\ge$transferability)---the best composition we found significantly outperformed the state of the art (e.g., 95.6% vs. 90.9% average transferability from normally trained surrogates to other normally trained models). We provide intuitive, empirically supported explanations for why certain augmentations fail to improve transferability. | We comprehensively studied data-augmentation methods for enhancing the transferability of adversarial examples, finding compositions that work best, and advancing the state of the art. |
Recent studies have shown that paddings in convolutional neural networks encode absolute position information which can negatively affect the model performance for certain tasks. However, existing metrics for quantifying the strength of positional information remain unreliable and frequently lead to erroneous results. To address this issue, we propose novel metrics for measuring and visualizing the encoded positional information. We formally define the encoded information as Position-information Pattern from Padding (PPP) and conduct a series of experiments to study its properties as well as its formation. The proposed metrics measure the presence of positional information more reliably than the existing metrics based on PosENet and tests in F-Conv. We also demonstrate that for any extant (and proposed) padding schemes, PPP is primarily a learning artifact and is less dependent on the characteristics of the underlying padding schemes. | We demonstrate a more accurate paradigm in quantifying the strength of positional information in CNN models |
We study the highly practical but comparatively under-studied problem of latent-domain adaptation, where a source model should be adapted to a target dataset that contains a mixture of unlabelled domain-relevant and domain-irrelevant examples. Furthermore, motivated by the requirements for data privacy and the need for embedded and resource-constrained devices of all kinds to adapt to local data distributions, we focus on the setting of feed-forward source-free domain adaptation, where adaptation should not require access to the source dataset, and also be back propagation-free. Our solution is to meta-learn a network capable of embedding the mixed-relevance target dataset and dynamically adapting inference for target examples using cross-attention. The resulting framework leads to consistent improvement on strong ERM baselines. We also show that our framework sometimes even improves on the upper bound of domain-supervised adaptation, where only domain-relevant instances are provided for adaptation. This suggests that human annotated domain labels may not always be optimal, and raises the possibility of doing better through automated instance selection. | Cross-attention based meta-learning approach for fast latent domain adaptation |
We present MemoNav, a novel memory model for image-goal navigation, which utilizes a working memory-inspired pipeline to improve navigation performance. Specifically, the node features on the topological map are stored in the short-term memory (STM), as these features are dynamically updated. The MemoNav retains the informative fraction of the STM via a forgetting module to improve navigation efficiency. To learn a global representation of 3D scenes, we introduce long-term memory (LTM) that continuously aggregates the STM. Afterward, a graph attention module encodes the retained STM and the LTM to generate working memory (WM). After encoding, the WM contains the informative features in the retained STM and the scene-level feature in the LTM and is finally used to generate actions. Consequently, the synergy of these three types of memory increases navigation performance by selectively retaining goal-relevant information and learning a high-level scene feature. When evaluated on multi-goal tasks, the MemoNav outperforms the SoTA methods at all difficulty levels in both Gibson and Matterport3D scenes. The MemoNav also achieves consistent improvements on traditional 1-goal tasks. Moreover, the qualitative results show that our model is less likely to be trapped in a deadlock. | The MemoNav learns three types of scene representations, which contain goal-relevant information and scene-level features and are utilized to improve navigation performance in both 1-goal and multi-goal ImageNav tasks. |
We study the application of variance reduction (VR) techniques to general non-convex stochastic optimization problems. In this setting, the recent work STORM (Cutkosky & Orabona, 2019) overcomes the drawback of having to compute gradients of “mega-batches” that earlier VR methods rely on. There, STORM utilizes recursive momentum to achieve the VR effect and is then later made fully adaptive in STORM+ (Levy et al., 2021), where full-adaptivity removes the requirement for obtaining certain problem-specific parameters such as the smoothness of the objective and bounds on the variance and norm of the stochastic gradients in order to set the step size. However, STORM+ crucially relies on the assumption that the function values are bounded, excluding a large class of useful functions. In this work, we propose META-STORM, a generalized framework of STORM+ that removes this bounded function values assumption while still attaining the optimal convergence rate for non-convex optimization. META-STORM not only maintains full-adaptivity, removing the need to obtain problem specific parameters, but also improves the convergence rate’s dependency on the problem parameters. Furthermore, META-STORM can utilize a large range of parameter settings that subsumes previous methods allowing for more flexibility in a wider range of settings. Finally, we demonstrate the effectiveness of META-STORM through experiments across common deep learning tasks. Our algorithm improves upon the previous work STORM+ and is competitive with widely used algorithms after the addition of per-coordinate update and exponential moving average heuristics. | We propose new fully adaptive variance reduced algorithms removing bounded function values and bounded gradients assumptions and improving upon previous work both in the theoretical convergence rate and empirical performance. |
Federated Learning (FL) is a nascent privacy-preserving learning framework under which the local data of participating clients is kept locally throughout model training. Scarce communication resources and data heterogeneity are two defining characteristics of FL. Besides, a FL system is often implemented in a harsh environment -- leaving the clients vulnerable to Byzantine attacks. To the best of our knowledge, no gradient compressors simultaneously achieve quantitative Byzantine resilience and privacy preservation. In this paper, we fill this gap via revisiting the stochastic sign SGD \cite{Jin2020}. We propose $\beta$-stochastic sign SGD, which contains a gradient compressor that encodes a client's gradient information in sign bits subject to the privacy budget $\beta>0$. We show that as long as $\beta>0$, $\beta$-stochastic sign SGD converges in the presence of partial client participation and mobile Byzantine faults, showing that it achieves quantifiable Byzantine-resilience and differential privacy simultaneously. In sharp contrast, when $\beta=0$, the compressor is not differentially private. Notably, for the special case when each of the stochastic gradients involved is bounded with known bounds, our gradient compressor with $\beta=0$ coincides with the compressor proposed in \cite{Jin2020}. As a byproduct, we show that when the clients report sign messages, the popular information aggregation rules simple mean, trimmed mean, median and majority vote are identical in terms of the output signs. Our theories are corroborated by experiments on MNIST and CIFAR-10 datasets. | Clients send stochastic sign-bits gradient estimates to a server, which aggregates updates based on majority vote. We show that this algorithm is provable differentially private, convergent, communication efficient, and Byzantine fault tolerant. |
Few-shot learning aims to transfer the knowledge acquired from training on a diverse set of tasks to unseen tasks from the same task distribution, with a limited amount of labeled data. The underlying requirement for effective few-shot generalization is to learn a good representation of the task manifold. This becomes more difficult when only a limited number of tasks are available for training. In such a few-task few-shot setting, it is beneficial to explicitly preserve the local neighborhoods from the task manifold and exploit this to generate artificial tasks for training. To this end, we introduce the notion of interval bounds from the provably robust training literature to few-shot learning. The interval bounds are used to characterize neighborhoods around the training tasks. These neighborhoods can then be preserved by minimizing the distance between a task and its respective bounds. We then use a novel strategy to artificially form new tasks for training by interpolating between the available tasks and their respective interval bounds. We apply our framework to both model-agnostic meta-learning as well as prototype-based metric-learning paradigms. The efficacy of our proposed approach is evident from the improved performance on several datasets from diverse domains in comparison to recent methods. | A method to densify the task distribution for few-task few-shot learning using task interpolation within interval-arithmetic-based bounds |
To avoid prohibitive computation cost of sending entire data, we propose four sparsification schemes Random-knots, Random-knots-Spatial, B-spline, Bspline-Spatial, and present corresponding nonparametric estimation of the covariance function. The covariance estimators are asymptotically equivalent to the sample covariance computed directly from the original data. And the estimated functional principal components effectively approximate the infeasible principal components under regularity conditions. The convergence rate reflects that leveraging spatial correlation and B-spline interpolation helps to reduce information loss. Data-driven selection method is further applied to determine the number of eigenfunctions in the model. Extensive numerical experiments are conducted to illustrate the theoretical results. | Novel sparsification schemes for functional data are proposed and the covariance estimation is shown to be asymptotically equivalent to sample covariance computed without sparsification. |
Based on the recent advancements in representation learning, we propose a novel framework for command-following robots with raw sensor inputs. Previous RL-based methods are either difficult to continuously improve after the deployment or require a large number of new labels during the fine-tuning. Motivated by (self-)supervised contrastive learning literature, we propose a novel representation, named VAR++, that generates an intrinsic reward function for command-following robot tasks by associating images with sound commands. After the robot is deployed in a new domain, the representation can be updated intuitively and data-efficiently by non-expert, and the robots are able to fulfill sound commands without any hand-crafted reward functions. We demonstrate our approach to various sound types and robotic tasks, including navigation and manipulation with raw sensor inputs. In the simulated experiments, we show that our system can continually self-improve in previously unseen scenarios given fewer new labeled data, yet achieves better performance, compared with previous methods.
| We learn a representation to generate a reward function to train command-following robots with reinforcement learning |
As with any machine learning problem with limited data, effective offline RL algorithms require careful regularization to avoid overfitting. One-step methods perform regularization by doing just a single step of policy improvement, while critic regularization methods do many steps of policy improvement with a regularized objective. These methods appear distinct. One-step methods, such as advantage-weighted regression and conditional behavioral cloning, truncate policy iteration after just one step. This ``early stopping'' makes one-step RL simple and stable, but can limit its asymptotic performance. Critic regularization typically requires more compute but has appealing lower-bound guarantees. In this paper, we draw a close connection between these methods: applying a multi-step critic regularization method with a regularization coefficient of 1 yields the same policy as one-step RL. While practical implementations violate our assumptions and critic regularization is typically applied with smaller regularization coefficients, our experiments nevertheless show that our analysis makes accurate, testable predictions about practical offline RL methods (CQL and one-step RL) with commonly-used hyperparameters. Our results that every problem can be solved with a single step of policy improvement, but rather that one-step RL might be competitive with critic regularization on RL problems that demand strong regularization. | We discuss some connections between one-step RL and critic regularization methods |
Curriculum learning (CL), whose core idea is to train from easy to hard, is a popular technique to accelerate reinforcement learning (RL) training. It has also been a trend to automate the curriculum generation process. Automatic CL works primarily focus on goal-conditioned RL problems, where an explicit indicator of training progress, e.g., reward or success rate, can be used to prioritize the training tasks. However, such a requirement is no longer valid in zero-sum games: there are no goals for the agents, and the accumulative reward of the learning policy can constantly fluctuate throughout training. In this work, we present the first theoretical framework of automatic curriculum learning in the setting of zero-sum games and derive a surprisingly simple indicator of training progress, i.e., the Q-value variance, which can be directly approximated by computing the variance of value network ensembles. With such a progression metric, we further adopt a particle-based task sampler to generate initial environment configurations for training, which is particularly lightweight, computation-efficient, and naturally multi-modal. Combining these techniques with multi-agent PPO training, we obtain our final algorithm, Zero-sum Automatic Curriculum Learning (ZACL). We first evaluate ZACL in a 2D particle-world environment, where ZACL produces much stronger policies than popular RL methods for zero-sum games using the same amount of samples. Then we show in the challenging hide-and-seek environment that ZACL can lead to all four emergent phases using a single desktop computer, which is reported for the first time in the literature. The project website is at https://sites.google.com/view/zacl. | In this work, we present the first theoretical framework of automatic curriculum learning in the setting of zero-sum game and derive a surprisingly simple indicator of training progress, i.e., the policy variance |
A treatment is usually appropriate for some group (the ``sick" group) on whom it has an effect, but it can also have a side-effect when given to subjects from another group (the ``healthy" group). In a non-targeted trial both sick and healthy subjects may be treated, producing
heterogeneous effects within the treated group. Inferring the correct treatment effect on the sick population is then
difficult, because the effect and side-effect are tangled. We propose an efficient nonparametric approach to untangling the effect and side-effect, called PCM (pre-cluster and merge). We prove its asymptotic consistency in a general setting and
show, on synthetic data,
more than a 10x improvement in accuracy over existing state-of-the-art.
| We propose an algorithm that provably recovers hidden effect groups in causal studies |
Few datasets contain self-identified demographic information, inferring demographic information risks introducing additional biases, and collecting and storing data on sensitive attributes can carry legal risks. Besides, categorical demographic labels do not necessarily capture all the relevant dimensions of human diversity. We propose to implicitly learn a set of continuous face-varying dimensions, without ever asking an annotator to explicitly categorize a person. We uncover the dimensions by learning on A View From Somewhere (AVFS) dataset of 638,180 human judgments of face similarity. We demonstrate the utility of our learned embedding space for predicting face similarity judgments, collecting continuous face attribute values, attribute classification, and comparative dataset diversity auditing. Moreover, using a novel conditional framework, we show that an annotator's demographics influences the \emph{importance} they place on different attributes when judging similarity, underscoring the \emph{need} for diverse annotator groups to avoid biases. Data and code are available at \url{https://github.com/SonyAI/a_view_from_somewhere}. | Implicit discovery of face-varying dimensions and annotator bias by learning on a novel face similarity dataset |
Bayesian meta-learning (BML) enables fitting expressive generative models to small datasets by incorporating inductive priors learned from a set of related tasks. The Neural Process (NP) is a prominent deep neural network-based BML architecture, which has shown remarkable results in recent years. In its standard formulation, the NP encodes epistemic uncertainty in an amortized, factorized, Gaussian variational (VI) approximation to the BML task posterior (TP), using reparametrized gradients. Prior work studies a range of architectural modifications to boost performance, such as attentive computation paths or improved context aggregation schemes, while the influence of the VI scheme remains under-explored. We aim to bridge this gap by introducing GMM-NP, a novel BML model, which builds on recent work that enables highly accurate, full-covariance Gaussian mixture (GMM) TP approximations by combining VI with natural gradients and trust regions. We show that GMM-NP yields tighter evidence lower bounds, which increases the efficiency of marginal likelihood optimization, leading to improved epistemic uncertainty estimation and accuracy. GMM-NP does not require complex architectural modifications, resulting in a powerful, yet conceptually simple BML model, which outperforms the state of the art on a range of challenging experiments, highlighting its applicability to settings where data is scarce. | We show that accurate inference of the task posterior is all you need for accurate Bayesian meta-learning. |
Consider the problem of clustering $n$ objects. One can apply multiple algorithms to produce $N$ potentially different clustersings of the same objects, that is, partitions of the $n$ objects into $K$ groups. Even a single randomized algorithm can output different clusterings. This often happens when one samples from the posterior of a Bayesian model, or runs multiple MCMC chains from random initializations. A natural task is then to form a consensus among these different clusterings. The challenge in an unsupervised setting is that the optimal matching between clusters of different inputs is unknown. We model this problem as finding a barycenter (also known as Fr\'{e}chet mean) relative to the misclassification rate. We show that by lifting the problem to the space of association matrices, one can derive aggregation algorithms that circumvent the knowledge of the optimal matchings. We analyze the statistical performance of aggregation algorithms under a stochastic label perturbation model, and show that a $K$-means type algorithm followed by a local refinement step can achieve near optimal performance, with a rate that decays exponentially fast in $N$. Numerical experiments show the effectiveness of the proposed methods. | We propose spectral algorithms for aggregating labels from multiple clustering algorithms without knowing the optimal matching between clusters, and we provide theoretical performance bounds. |
In this paper, we show that recent advances in self-supervised representation learning enable unsupervised object discovery and semantic segmentation with a performance that matches the state of the field on supervised semantic segmentation 10 years ago. We propose a methodology based on unsupervised saliency masks and self-supervised feature clustering to kickstart object discovery followed by training a semantic segmentation network on pseudo-labels to bootstrap the system on images with multiple objects. We show that while being conceptually simple our proposed baseline is surprisingly strong. We present results on PASCAL VOC that go far beyond the current state of the art (50.0 mIoU), and we report for the first time results on MS COCO for the whole set of 81 classes: our method discovers 34 categories with more than 20% IoU, while obtaining an average IoU of 19.6 for all 81 categories. | Strong and simple baseline for unsupervised segmentation methods obtained by leveraging and combining object-centric priors. |
We present a novel deep learning approach to approximate the solution of large, sparse, symmetric, positive-definite linear systems of equations.These systems arise from many problems in applied science, e.g., in numerical methods for partial differential equations. Algorithms for approximating the solution to these systems are often the bottleneck in problems that require their solution, particularly for modern applications that require many millions of unknowns. Indeed, numerical linear algebra techniques have been investigated for many decades to alleviate this computational burden. Recently, data-driven techniques have also shown promise for these problems. Motivated by the conjugate gradients algorithm that iteratively selects search directions for minimizing the matrix norm of the approximation error, we design an approach that utilizes a deep neural network to accelerate convergence via data-driven improvement of the search directions.
Our method leverages a carefully chosen convolutional network to approximate the action of the inverse of the linear operator up to an arbitrary constant. We train the network using unsupervised learning with a loss function equal to the $L^2$ difference between an input and the system matrix times the network evaluation, where the unspecified constant in the approximate inverse is accounted for.
We demonstrate the efficacy of our approach on spatially discretized Poisson equations with millions of degrees of freedom arising in computational fluid dynamics applications.Unlike state-of-the-art learning approaches, our algorithm is capable of reducing the linear system residual to a given tolerance in a small number of iterations, independent of the problem size.Moreover, our method generalizes effectively to various systems beyond those encountered during training. | We present a CNN-based algorithm for solving linear systems with millions of degrees of freedom in a way that rapidly achieves convergence to a specified tolerance, a significant improvement over learning methods that converge slowly or not at all. |
Traditionally, data valuation is posed as a problem of equitably splitting the validation performance of a learning algorithm among the training data. As a result, the calculated data values depend on many design choices of the underlying learning algorithm. However, this dependence is undesirable for many use cases of data valuation, such as setting priorities over different data sources in a data acquisition process and informing pricing mechanisms in a data marketplace. In these scenarios, data needs to be valued before the actual analysis and the choice of the learning algorithm is still undetermined then. Another side-effect of the dependence is that to assess the value of individual points, one needs to re-run the learning algorithm with and without a point, which incurs a large computation burden.
This work leapfrogs over the current limits of data valuation methods by introducing a new framework that can value training data in a way that is oblivious to the downstream learning algorithm. Our main results are as follows. $\textbf{(1)}$ We develop a proxy for the validation performance associated with a training set based on a non-conventional $\textit{class-wise}$ $\textit{Wasserstein distance}$ between the training and the validation set. We show that the distance characterizes the upper bound of the validation performance for any given model under certain Lipschitz conditions. $\textbf{(2)}$ We develop a novel method to value individual data based on the sensitivity analysis of the $\textit{class-wise}$ Wasserstein distance. Importantly, these values can be directly obtained $\textit{for free}$ from the output of off-the-shelf optimization solvers once the Wasserstein distance is computed. $\textbf{(3) }$We evaluate our new data valuation framework over various use cases related to detecting low-quality data
and show that, surprisingly, the learning-agnostic feature of our framework enables a $\textit{significant improvement}$ over the state-of-the-art performance while being $\textit{orders of magnitude faster.}$ | We propose LAVA: a novel model-agnostic approach to data valuation using a non-conventional, class-wise Wasserstein discrepancy. |
Fine-tuning pre-trained language models has become the prevalent paradigm for building downstream NLP models. Oftentimes fine-tuned models are readily available but their training data is not, due to data privacy or intellectual property concerns. This creates a barrier to fusing knowledge across individual models to yield a better single model. In this paper, we study the problem of merging individual models built on different training data sets to obtain a single model that performs well both across all data set domains and can generalize on out-of-domain data. We propose a data-less knowledge fusion method that merges models in their parameter space, guided by weights that minimize prediction differences between the merged model and the individual models. Over a battery of evaluation settings, we show that the proposed method significantly outperforms baselines such as Fisher-weighted averaging or model ensembling. Further, we find that our method is a promising alternative to multi-task learning that can preserve or sometimes improve over the individual models without access to the training data. Finally, model merging is more efficient than training a multi-task model, thus making it applicable to a wider set of scenarios. | We study the problem of merging individual models built on different training data sets and propose a novel merging algorithm. |
Multi-view image compression plays a critical role in 3D-related applications. Existing methods adopt a predictive coding architecture, which requires joint encoding to compress the corresponding disparity as well as residual information. This demands collaboration among cameras and enforces the epipolar geometric constraint between different views, which makes it challenging to deploy these methods in distributed camera systems with randomly overlapping fields of view. Meanwhile, distributed source coding theory indicates that efficient data compression of correlated sources can be achieved by independent encoding and joint decoding, which motivates us to design a learning-based distributed multi-view image coding (LDMIC) framework. With independent encoders, LDMIC introduces a simple yet effective joint context transfer module based on the cross-attention mechanism at the decoder to effectively capture the global inter-view correlations, which is insensitive to the geometric relationships between images. Experimental results show that LDMIC significantly outperforms both traditional and learning-based MIC methods while enjoying fast encoding speed. Code is released at https://github.com/Xinjie-Q/LDMIC. | We design a multi-view image compression framework based on symmetric distributed source coding paradigm, which achieves higher compression performance than previous multi-view image compression methods. |
Neural ordinary differential equations (NODEs) have received a lot of attention in recent years due to their memory efficiency. Different from traditional deep learning, it defines a continuous deep learning architecture based on the theory of ordinary differential equations (ODEs), which also improves the interpretability of deep learning. However, it has several obvious limitations, such as a NODE is not a universal approximator, it requires a large number of function evaluations (NFEs), and it has a slow convergence rate. We address these drawbacks by modeling and adding an oscillator to the framework of the NODEs. The oscillator enables the trajectories of our model to cross each other. We prove that our model is a universal approximator, even in the original input space. Due to the presence of oscillators, the flows learned by the model will be simpler, thus our model needs fewer NFEs and has a faster convergence speed. We apply our model to various tasks including classification and time series extrapolation, then compare several metrics including accuracy, NFEs, and convergence speed. The experiments show that our model can achieve better results compared to the existing baselines. | Oscillation Neural Ordinary Differential Equations |
Neural networks are known to use spurious correlations such as background information for classification. While prior work has looked at spurious correlations that are widespread in the training data, in this work, we investigate how sensitive neural networks are to $rare$ spurious correlations, which may be harder to detect and correct, and may lead to privacy leaks. We introduce spurious patterns correlated with a fixed class to a few training examples and find that it takes only a handful of such examples for the network to learn the correlation. Furthermore, these rare spurious correlations also impact accuracy and privacy. We empirically and theoretically analyze different factors involved in rare spurious correlations and propose mitigation methods accordingly. Specifically, we observe that $\ell_2$ regularization and adding Gaussian noise to inputs can reduce the undesirable effects. | Neural networks can learn spurious correlations caused by a small number of training examples. We empirically and theoretically study this phenomena. |
Activation compressed training (ACT) has been shown to be a promising way to reduce the memory cost of training deep neural networks (DNNs). However, existing work of ACT relies on searching for optimal bit-width during DNN training to reduce the quantization noise, which makes the procedure complicated and less transparent. To this end, we propose a simple and effective method to compress DNN training. Our method is motivated by an instructive observation: DNN backward propagation mainly utilizes the low-frequency component (LFC) of the activation maps, while the majority of memory is for caching the high-frequency component (HFC) during the training. This indicates the HFC of activation maps is highly redundant and compressible during DNN training, which inspires our proposed Dual Activation Precision (DIVISION). During the training, DIVISION preserves the high-precision copy of LFC and compresses the HFC into a light-weight copy with low numerical precision. This can significantly reduce the memory cost without negatively affecting the precision of backward propagation such that DIVISION maintains competitive model accuracy. Experimental results show DIVISION achieves over 10× compression of activation maps, and significantly higher training throughput than state-of-the-art ACT methods, without loss of model accuracy. The code is available at https://anonymous.4open.science/r/division-5CC0/
| A simple and transparent framework to reduce the memory cost of DNN training. |
It is widely observed that vanilla autoencoders can have low manifold learning accuracy given a noisy or small training dataset.
Recent work has discovered that it is important to regularize the decoder that explicitly parameterizes the manifold,
where a neighborhood graph is employed for decoder regularization. However, one caveat of this method is that it is not always straightforward to construct a correct graph. Alternatively, one may consider naive graph-free regularization methods such as minimizing the norm of the decoder's Jacobian or Hessian, but these norms are not coordinate-invariant (i.e. reparametrization-invariant) and hence do not capture any meaningful geometric quantity of the manifold nor result in geometrically meaningful manifold regularization effects.
Another recent work called the isometric regularization implicitly forces the manifold to have zero intrinsic curvature, resulting in some geometrically meaningful regularization effects. But, since the intrinsic curvature does not capture how the manifold is embedded in the data space from an extrinsic perspective, the regularization effects are often limited. In this paper, we propose a {\it minimum extrinsic curvature principle} for manifold regularization and {\bf Minimum Curvature Autoencoder (MCAE)}, a graph-free coordinate-invariant extrinsic curvature minimization framework for autoencoder regularization. Experiments with various standard datasets demonstrate that MCAE improves manifold learning accuracy compared to existing methods, especially showing strong robustness to noise. | We propose a minimum extrinsic curvature principle for manifold regularization and Minimum Curvature Autoencoder (MCAE), a graph-free coordinate-invariant extrinsic curvature minimization framework for autoencoder regularization. |
The problem of end-to-end learning of a communication system using an autoencoder -- consisting of an encoder, channel, and decoder modeled using neural networks -- has recently been shown to be an effective approach. A challenge faced in the practical adoption of this learning approach is that under changing channel conditions (e.g. a wireless link), it requires frequent retraining of the autoencoder in order to maintain a low decoding error rate. Since retraining is both time consuming and requires a large number of samples, it becomes impractical when the channel distribution is changing quickly. We propose to address this problem using a fast and sample-efficient (few-shot) domain adaptation method that does not change the encoder and decoder networks. Different from conventional training-time unsupervised or semi-supervised domain adaptation, here we have a trained autoencoder from a source distribution that we want to adapt (at test time) to a target distribution using only a small labeled dataset, and no unlabeled data. We focus on a generative channel model based on the Gaussian mixture density network (MDN), and propose a regularized, parameter-efficient adaptation of the MDN using a set of affine transformations. The learned affine transformations are then used to design an optimal transformation at the decoder input to compensate for the distribution shift, and effectively present to the decoder inputs close to the source distribution. Experiments on many simulated distribution changes common to the wireless setting, and a real mmWave FPGA testbed demonstrate the effectiveness of our method at adaptation using very few target domain samples~\footnote{Code for our work: \url{https://github.com/jayaram-r/domain-adaptation-autoencoder}}. | We propose a sample-efficient domain adaptation method for the autoencoder based end-to-end communication problem |
The energy consumption for training deep learning models is increasing at an alarming rate due to the growth of training data and model scale, resulting in a negative impact on carbon neutrality. Energy consumption is an especially pressing issue for AutoML algorithms because it usually requires repeatedly training large numbers of computationally intensive deep models to search for optimal configurations. This paper takes one of the most essential steps in developing energy-aware (EA) NAS methods, by providing a benchmark that makes EA-NAS research more reproducible and accessible. Specifically, we present the first large-scale energy-aware benchmark that allows studying AutoML methods to achieve better trade-offs between performance and search energy consumption, named EA-HAS-Bench. EA-HAS-Bench provides a large-scale architecture/hyperparameter joint search space, covering diversified configurations related to energy consumption. Furthermore, we propose a novel surrogate model specially designed for large joint search space, which proposes a Bezier curve-based model to predict learning curves with unlimited shape and length. Based on the proposed dataset, we new energy-aware AutoML method that arms existing AutoML algorithms to consider the search energy consumption, and our experiments show that the modified energy-aware AutoML methods achieve a better trade-off between energy consumption and model performance. | We provide the first HAS dataset aware of the overall search energy cost |
We present a novel graph neural network we call AgentNet, which is designed specifically for graph-level tasks. AgentNet is inspired by sublinear algorithms, featuring a computational complexity that is independent of the graph size. The architecture of AgentNet differs fundamentally from the architectures of traditional graph neural networks. In AgentNet, some trained \textit{neural agents} intelligently walk the graph, and then collectively decide on the output. We provide an extensive theoretical analysis of AgentNet: We show that the agents can learn to systematically explore their neighborhood and that AgentNet can distinguish some structures that are even indistinguishable by 2-WL. Moreover, AgentNet is able to separate any two graphs which are sufficiently different in terms of subgraphs. We confirm these theoretical results with synthetic experiments on hard-to-distinguish graphs and real-world graph classification tasks. In both cases, we compare favorably not only to standard GNNs but also to computationally more expensive GNN extensions. | We present a new agent-based sublinear and expressive GNN architecture for graph-level tasks. |
We consider the explanation problem of Graph Neural Networks (GNNs). Most existing GNN explanation methods identify the most important edges or nodes but fail to consider substructures, which are more important for graph data. One method considering subgraphs tries to search all possible subgraphs and identifies the most significant ones. However, the subgraphs identified may not be recurrent or statistically important for interpretation. This work proposes a novel method, named MotifExplainer, to explain GNNs by identifying important motifs, which are recurrent and statistically significant patterns in graphs. Our proposed motif-based methods can provide better human-understandable explanations than methods based on nodes, edges, and regular subgraphs. Given an instance graph and a pre-trained GNN model, our method first extracts motifs in the graph using domain-specific motif extraction rules. Then, a motif embedding is encoded by feeding motifs into the pre-trained GNN. Finally, we employ an attention-based method to identify the most influential motifs as explanations for the prediction results. The empirical studies on both synthetic and real-world datasets demonstrate the effectiveness of our method. | We propose a motif-based explainer that can provide better human-understandable explanations than methods based on nodes, edges, and regular subgraphs. |
Neural networks embed the geometric structure of a data manifold lying in a high-dimensional space into latent representations. Ideally, the distribution of the data points in the latent space should depend only on the task, the data, the loss, and other architecture-specific constraints. However, factors such as the random weights initialization, training hyperparameters, or other sources of randomness in the training phase may induce incoherent latent spaces that hinder any form of reuse. Nevertheless, we empirically observe that, under the same data and modeling choices, the angles between the encodings within distinct latent spaces do not change. In this work, we propose the latent similarity between each sample and a fixed set of anchors as an alternative data representation, demonstrating that it can enforce the desired invariances without any additional training. We show how neural architectures can leverage these relative representations to guarantee, in practice, invariance to latent isometries and rescalings, effectively enabling latent space communication: from zero-shot model stitching to latent space comparison between diverse settings. We extensively validate the generalization capability of our approach on different datasets, spanning various modalities (images, text, graphs), tasks (e.g., classification, reconstruction) and architectures (e.g., CNNs, GCNs, transformers). | Relative representations can be leveraged to enable solving tasks regarding "latent communication": from zero-shot model stitching to latent space comparison between diverse settings. |
We consider a deep matrix factorization model of covariance matrices trained with the Bures-Wasserstein distance. While recent works have made important advances in the study of the optimization problem for overparametrized low-rank matrix approximation, much emphasis has been placed on discriminative settings and the square loss. In contrast, our model considers another interesting type of loss and connects with the generative setting. We characterize the critical points and minimizers of the Bures-Wasserstein distance over the space of rank-bounded matrices. For low-rank matrices the Hessian of this loss can blow up, which creates challenges to analyze convergence of optimizaton methods. We establish convergence results for gradient flow using a smooth perturbative version of the loss and convergence results for finite step size gradient descent under certain assumptions on the initial weights | We prove convergence of gradient decent optimization of generative deep linear networks trained with the Bures-Wasserstein loss. |
One-shot non-autoregressive neural networks, different from RL-based ones, have been actively adopted for solving combinatorial optimization (CO) problems, which can be trained by the objective score in a self-supervised manner. Such methods have shown their superiority in efficiency (e.g. by parallelization) and potential for tackling predictive CO problems for decision-making under uncertainty. While the discrete constraints often become a bottleneck for gradient-based neural solvers, as currently handled in three typical ways: 1) adding a soft penalty in the objective, where a bounded violation of the constraints cannot be guaranteed, being critical to many constraint-sensitive scenarios; 2) perturbing the input to generate an approximate gradient in a black-box manner, though the constraints are exactly obeyed while the approximate gradients can hurt the performance on the objective score; 3) a compromise by developing soft algorithms whereby the output of neural networks obeys a relaxed constraint, and there can still occur an arbitrary degree of constraint-violation. Towards the ultimate goal of establishing a general framework for neural CO solver with the ability to control an arbitrary-small degree of constraint violation, in this paper, we focus on a more achievable and common setting: the cardinality constraints, which in fact can be readily encoded by a differentiable optimal transport (OT) layer. Based on this observation, we propose OT-based cardinality constraint encoding for end-to-end CO problem learning with two variants: Sinkhorn and Gumbel-Sinkhorn, whereby their violation of the constraints can be exactly characterized and bounded by our theoretical results. On synthetic and real-world CO problem instances, our methods surpass the state-of-the-art CO network and are comparable to (if not superior to) the commercial solver Gurobi. In particular, we further showcase a case study of applying our approach to the predictive portfolio optimization task on real-world asset price data, improving the Sharpe ratio from 1.1 to 2.0 of a strong LSTM+Gurobi baseline under the classic predict-then-optimize paradigm. | We present a Gumbel-Sinkhorn network for cardinality-constrained combinatorial optimization with theoretical and empirical notes. We surpass Erdos Goes Neural on optimization problems, and present an application on predictive portfolio optimization. |
We improve upon previous oblivious sketching and turnstile streaming results for $\ell_1$ and logistic regression, giving a much smaller sketching dimension achieving $O(1)$-approximation and yielding an efficient optimization problem in the sketch space. Namely, we achieve for any constant $c>0$ a sketching dimension of $\tilde{O}(d^{1+c})$ for $\ell_1$ regression and $\tilde{O}(\mu d^{1+c})$ for logistic regression, where $\mu$ is a standard measure that captures the complexity of compressing the data. For $\ell_1$-regression our sketching dimension is near-linear and improves previous work which either required $\Omega(\log d)$-approximation with this sketching dimension, or required a larger $\operatorname{poly}(d)$ number of rows. Similarly, for logistic regression previous work had worse $\operatorname{poly}(\mu d)$ factors in its sketching dimension. We also give a tradeoff that yields a $1+\varepsilon$ approximation in input sparsity time by increasing the total size to $(d\log(n)/\varepsilon)^{O(1/\varepsilon)}$ for $\ell_1$ and to $(\mu d\log(n)/\varepsilon)^{O(1/\varepsilon)}$ for logistic regression. Finally, we show that our sketch can be extended to approximate a regularized version of logistic regression where the data-dependent regularizer corresponds to the variance of the individual logistic losses. | We give the first constant factor approximate sketches for l1 and logistic regression in a turnstile stream with almost linear sketching dimension that result in an efficient optimization problem in the sketch space. |
One of the major open problems in machine learning is to characterize generalization in the overparameterized regime, where most traditional generalization bounds become inconsistent (Nagarajan and Kolter, 2019). In many scenarios, their failure can be attributed to obscuring the crucial interplay between the training algorithm and the underlying data distribution. To address this issue, we propose a concept named compatibility, which quantitatively characterizes generalization in a both data-relevant and algorithm relevant manner. By considering the entire training trajectory and focusing on early-stopping iterates, compatibility exploits the data and the algorithm information and is therefore a more suitable notion for generalization. We validate this by theoretically studying compatibility under the setting of solving overparameterized linear regression with gradient descent. Specifically, we perform a data-dependent trajectory analysis and derive a sufficient condition for compatibility in such a setting. Our theoretical results demonstrate that in the sense of compatibility, generalization holds with significantly weaker restrictions on the problem instance than the previous last iterate analysis. | We propose data-algorithm-compatibility to characterize generalization, and study it in the overparameterized linear regression regime. |
The low transferability of learned policies is one of the most critical problems limiting the applicability of learning-based solutions to decision-making tasks. In this paper, we present a way to align latent representations of states and actions between different domains by optimizing an adversarial objective. We train two models, a policy and a domain discriminator, with unpaired trajectories of proxy tasks through behavioral cloning as well as adversarial training. After the latent representations are aligned between domains, a domain-agnostic part of the policy trained with any method in the source domain can be immediately transferred to the target domain in a zero-shot manner. We empirically show that our simple approach achieves comparable performance to the latest methods in zero-shot cross-domain transfer. We also observe that our method performs better than other approaches in transfer between domains with different complexities, whereas other methods fail catastrophically. | We obtain a domain-invariant feature space by behavioral cloning and adversarial training using unpaired trajectories of proxy tasks, and use it for zero-shot cross-domain transfer. |
Although input attribution methods are mainstream in understanding predictions of DNNs for straightforward interpretations, the non-linearity of DNNs often makes the attributed scores unreliable in explaining a given prediction, deteriorating the faithfulness of the explanation.
However, the challenge could be mitigated by attributing scores to groups of explanatory components instead of the individuals, termed group attribution. While a group attribution would explain the group-wise contribution more reliably, it does not explain the component-wise contributions so that estimating component-wise scores yields less reliable explanation, indicating the trade-off of group attributions.
In this work, we introduce the generalized definition of reliability loss and group attribution, and formulate the optimization problem of the trade-off with these terms. We apply our formalization to Shapley value attribution to propose the optimization method G-SHAP. We show the effectiveness and explanatory benefits of our method through empirical results on image classification tasks. | We have proposed the group-wise attribution methods to yield more reliable explanation in understanding a model's prediction |
Records in a table are represented by a collection of heterogeneous scalar features. Previous work often made predictions for records in a paradigm that processed each feature as an operating unit, which requires to well cope with the heterogeneity. In this paper, we propose to encapsulate all feature values of a record into vectorial features and process them collectively rather than have to deal with individual ones, which directly captures the representations at the data level and benefits robust performances. Specifically, we adopt the concept of "capsules" to organize features into vectorial features, and devise a novel capsule neural network called "TabCaps" to process the vectorial features for classification. In TabCaps, a record is encoded into several vectorial features by some optimizable multivariate Gaussian kernels in the primary capsule layer, where each vectorial feature represents a specific "profile" of the input record and is transformed into senior capsule layer under the guidance of a new straightforward routing algorithm. The design of routing algorithm is motivated by the Bag-of-Words (BoW) model, which performs capsule feature grouping straightforwardly and efficiently, in lieu of the computationally complex clustering of previous routing algorithms. Comprehensive experiments show that TabCaps achieves competitive and robust performances in tabular data classification tasks. | We proposed a capsule neural network for tabular data classification. |
Distributed Mean Estimation (DME) is a fundamental building block in communication efficient federated learning. In DME, clients communicate their lossily compressed gradients to the parameter server, which estimates the average and updates the model.
State of the art DME techniques apply either unbiased quantization methods, resulting in large estimation errors, or biased quantization methods, where unbiasing the result requires that the server decodes each gradient individually, which markedly slows the aggregation time.
In this paper, we propose QUIC-FL, a DME algorithm that achieves the best of all worlds. QUIC-FL is unbiased, offers fast aggregation time, and is competitive with the most accurate (slow aggregation) DME techniques. To achieve this, we formalize the problem in a novel way that allows us to use standard solvers to design near-optimal unbiased quantization schemes. | A distributed mean estimation compression scheme with accuracy on-par with the state of the art while asymptotically improving the decoding time. |
In this paper, we propose AutoSKDBERT, a new knowledge distillation paradigm for BERT compression, that stochastically samples a teacher from a predefined teacher team following a categorical distribution in each step, to transfer knowledge into student. AutoSKDBERT aims to discover the optimal categorical distribution which plays an important role to achieve high performance. The optimization procedure of AutoSKDBERT can be divided into two phases: 1) phase-1 optimization distinguishes effective teachers from ineffective teachers, and 2) phase-2 optimization further optimizes the sampling weights of the effective teachers to obtain satisfactory categorical distribution. Moreover, after phase-1 optimization completion, AutoSKDBERT adopts teacher selection strategy to discard the ineffective teachers whose sampling weights are assigned to the effective teachers. Particularly, to alleviate the gap between categorical distribution optimization and evaluation, we also propose a stochastic single-weight optimization strategy which only updates the weight of the sampled teacher in each step. Extensive experiments on GLUE benchmark show that the proposed AutoSKDBERT achieves state-of-the-art score compared to previous compression approaches on several downstream tasks, including pushing MRPC F1 and accuracy to 93.2 (0.6 point absolute improvement) and 90.7 (1.2 point absolute improvement), RTE accuracy to 76.9 (2.9 point absolute improvement). | AutoSKDBERT stochastically selects a teacher model from a predefined teacher team to distill student model in each iteration with a learnable categorical distribution. |
In complex reinforcement learning (RL) problems, policies with similar rewards may have substantially different behaviors. Yet, to not only optimize rewards but also discover as many diverse strategies as possible remains a challenging problem. A natural approach to this task is constrained population-based training (PBT), which simultaneously learns a collection of policies subject to diversity constraints. However, due to the unaffordable computation cost of PBT, we adopt an alternative approach, iterative learning (IL), which repeatedly learns a single novel policy that is sufficiently different from previous ones. We first analyze these two frameworks and prove that, for any policy pool derived by PBT, we can always use IL to obtain another policy pool of the same rewards and competitive diversity scores. In addition, we also present a novel state-based diversity measure with two tractable realizations. Such a metric can impose a stronger and much smoother diversity constraint than existing action-based metrics. Combining IL and the state-based diversity measure, we develop a powerful diversity-driven RL algorithm, State-based Intrinsic-reward Policy Optimization (SIPO), with provable convergence properties. We empirically examine our algorithm in complex multi-agent environments including StarCraft Multi-Agent Challenge and Google Research Football. SIPO is able to consistently derive strategically diverse and human-interpretable policies that cannot be discovered by existing baselines. | We develop an iterative RL algorithm for discovering diverse high-reward strategies with provable convergence properties. |
Incorporating second-order gradient information (curvature) into optimization can dramatically reduce the number of iterations required to train machine learning models. In natural gradient descent, such information comes from the Fisher information matrix which yields a number of desirable properties. As exact natural gradient updates are intractable for large models, successful methods such as KFAC and sequels approximate the Fisher in a structured form that can easily be inverted. However, this requires model/layer-specific tensor algebra and certain approximations that are often difficult to justify. Here, we use ideas from Legendre-Fenchel duality to learn a direct and efficiently evaluated model for the product of the inverse Fisher with any vector, in an online manner, leading to natural gradient steps that get progressively more accurate over time despite noisy gradients. We prove that the resulting “Fisher-Legendre” (FishLeg) optimizer converges to a (global) minimum of non-convex functions satisfying the PL condition, which applies in particular to deep linear networks. On standard auto-encoder benchmarks, we show empirically that FishLeg outperforms standard first-order optimization methods, and performs on par with or better than other second-order methods, especially when using small batches. Thanks to its generality, we expect our approach to facilitate the handling of a variety neural network layers in future work. | We introduce a new approach to estimate the natural gradient via Legendre-Fenchel duality, provide a convergence proof, and show competitive performance on a number of benchmarks. |
To protect user privacy and meet legal regulations, federated learning (FL) is attracting significant attention. Training neural machine translation (NMT) models with traditional FL algorithm (e.g., FedAvg) typically relies on multi-round model-based interactions. However, it is impractical and inefficient for machine translation tasks due to the vast communication overheads and heavy synchronization. In this paper, we propose a novel federated nearest neighbor (FedNN) machine translation framework that, instead of multi-round model-based interactions, leverages one-round memorization-based interaction to share knowledge across different clients to build low-overhead privacy-preserving systems. The whole approach equips the public NMT model trained on large-scale accessible data with a $k$-nearest-neighbor ($k$NN) classifier and integrates the external datastore constructed by private text data in all clients to form the final FL model. A two-phase datastore encryption strategy is introduced to achieve privacy-preserving during this process. Extensive experiments show that FedNN significantly reduces computational and communication costs compared with FedAvg, while maintaining promising performance in different FL settings. | We propose a novel federated nearest neighbor machine translation framework to build low-overhead privacy-preserving MT systems in FL settings. |
User embeddings (vectorized representations of a user) are essential in recommendation systems. Numerous approaches have been proposed to construct a representation for the user in order to find similar items for retrieval tasks, and they have been proven effective in industrial recommendation systems. Recently people have discovered the power of using multiple embeddings to represent a user, with the hope that each embedding represents the user's interest in a certain topic. With multi-interest representation, it's important to model the user's preference over the different topics and how the preference change with time. However, existing approaches either fail to estimate the user's affinity to each interest or unreasonably assume every interest of every user fades with an equal rate with time, thus hurting the performance of candidate retrieval. In this paper, we propose the Multi-Interest Preference (MIP) model, an approach that not only produces multi-interest for users by using the user's sequential engagement more effectively but also automatically learns a set of weights to represent the preference over each embedding so that the candidates can be retrieved from each interest proportionally. Extensive experiments have been done on various industrial-scale datasets to demonstrate the effectiveness of our approach. | A joint-modeling of unbiased multiple user interests and the interest weights. |
Marginal-based methods achieve promising performance in the synthetic data competition hosted by the National Institute of Standards and Technology (NIST).
To deal with high-dimensional data, the distribution of synthetic data is represented by a probabilistic graphical model (e.g., a Bayesian network), while the raw data distribution is approximated by a collection of low-dimensional marginals.
Differential privacy (DP) is guaranteed by introducing random noise to each low-dimensional marginal distribution.
Despite its promising performance in practice, the statistical properties of marginal-based methods are rarely studied in the literature.
In this paper, we study DP data synthesis algorithms based on Bayesian networks (BN) from a statistical perspective. We establish a rigorous accuracy guarantee for BN-based algorithms, where the errors are measured by the total variation (TV) distance or the $L^2$ distance.
Related to downstream machine learning tasks, an upper bound for the utility error of the DP synthetic data is also derived. To complete the picture, we establish a lower bound for TV accuracy that holds for every $\epsilon$-DP synthetic data generator. | We analyze the differentially private marginal-based data synthesis algorithms in a statistical framework and establish a theoretical guarantee for the accuracy and utility. |
Natural environments have temporal structure at multiple timescales. This property is reflected in biological learning and memory but typically not in machine learning systems. We advance a multiscale learning method in which each weight in a neural network is decomposed as a sum of subweights with different learning and decay rates. Thus knowledge becomes distributed across different timescales, enabling rapid adaptation to task changes while avoiding catastrophic interference. First, we prove previous models that learn at multiple timescales, but with complex coupling between timescales, are equivalent to multiscale learning via a reparameterization that eliminates this coupling. The same analysis yields a new characterization of momentum learning, as a fast weight with a negative learning rate. Second, we derive a model of Bayesian inference over $1/f$ noise, a common temporal pattern in many online learning domains that involves long-range (power law) autocorrelations. The generative side of the model expresses $1/f$ noise as a sum of diffusion processes at different timescales, and the inferential side tracks these latent processes using a Kalman filter. We then derive a variational approximation to the Bayesian model and show how it is an extension of the multiscale learner. The result is an optimizer that can be used as a drop-in replacement in an arbitrary neural network architecture. Third, we evaluate the ability of these methods to handle nonstationarity by testing them in online prediction tasks characterized by $1/f$ noise in the latent parameters. We find that the Bayesian model significantly outperforms online stochastic gradient descent and two batch heuristics that rely preferentially or exclusively on more recent data. Moreover, the variational approximation performs nearly as well as the full Bayesian model, and with memory requirements that are linear in the size of the network.
| Models that learn at multiple timescales perform well in tasks with complex temporal structure |
We study the problem of deployment efficient reinforcement learning (RL) with linear function approximation under the \emph{reward-free} exploration setting. This is a well-motivated problem because deploying new policies is costly in real-life RL applications. Under the linear MDP setting with feature dimension $d$ and planning horizon $H$, we propose a new algorithm that collects at most $\widetilde{O}(\frac{d^2H^5}{\epsilon^2})$ trajectories within $H$ deployments to identify $\epsilon$-optimal policy for any (possibly data-dependent) choice of reward functions. To the best of our knowledge, our approach is the first to achieve optimal deployment complexity and optimal $d$ dependence in sample complexity at the same time, even if the reward is known ahead of time. Our novel techniques include an exploration-preserving policy discretization and a generalized G-optimal experiment design, which could be of independent interest. Lastly, we analyze the related problem of regret minimization in low-adaptive RL and provide information-theoretic lower bounds for switching cost and batch complexity. | We design algorithms for reward free RL under linear MDP with near-optimal deployment complexity and sample complexity. |
We present Newton losses, a method for incorporating second-order information of losses by approximating them with quadratic functions. The presented method is applied only to the loss function and allows training the neural network with gradient descent. As loss functions are usually substantially cheaper to compute than the neural network, Newton losses can be used at a relatively small additional cost. We find that they yield superior performance, especially when applied to non-convex and hard-to-optimize loss functions such as algorithmic losses, which have been popularized in recent research. | Applying Newton to the loss and gradient descent to the neural network. |
Many deep neural networks are vulnerable to backdoor poisoning attacks, in which an adversary strategically injects a backdoor trigger into a small fraction of the training data. The trigger can later be applied during inference to manipulate prediction labels. While the data label could be changed to arbitrary values by an adversary, the extent of corruption injected into the feature values is strictly limited to keep the backdoor attack in disguise, which leads to a resemblance between the backdoor attack and a milder attack that involves only noisy labels.
This paper investigates an intriguing question: \textit{Can we leverage algorithms that defend against noisy label corruptions to defend against general backdoor attacks?} We first discuss the limitations of directly using current noisy-label defense algorithms to defend against backdoor attacks. We then propose a meta-algorithm for both supervised and semi-supervised settings that transforms an existing noisy label defense algorithm into one that protects against backdoor attacks. Extensive experiments on different settings show that, by introducing a lightweight alteration for minimax optimization to the existing noisy-label defense algorithms, the robustness against backdoor attacks can be substantially improved, while the initial form of those algorithms would fail in the presence of a backdoor attack. | We propose a principled approach to defend the backdoor attack by using existed robust algorithm against label noise. |
Animals are able to rapidly infer from limited experience when sets of state action pairs have equivalent reward and transition dynamics. On the other hand, modern reinforcement learning systems must painstakingly learn through trial and error that sets of state action pairs are value equivalent---requiring an often prohibitively large amount of samples from their environment. MDP homomorphisms have been proposed that reduce the observed MDP of an environment to an abstract MDP, which can enable more sample efficient policy learning. Consequently, impressive improvements in sample efficiency have been achieved when a suitable MDP homomorphism can be constructed a priori---usually by exploiting a practioner's knowledge of environment symmetries. We propose a novel approach to constructing a homomorphism in discrete action spaces, which uses a partial model of environment dynamics to infer which state action pairs lead to the same state---reducing the size of the state-action space by a factor equal to the cardinality of the action space. We call this method equivalent effect abstraction. We demonstrate empirically that equivalent effect abstraction can improve sample efficiency in a model-free setting and planning efficiency for model based approaches. | MDP Homomorphism with a forwards-backwards model |
Pseudo-labeling has proven to be a promising semi-supervised learning (SSL) paradigm. Existing pseudo-labeling methods commonly assume that the class distributions of training data are balanced. However, such an assumption is far from realistic scenarios and thus severely limits the performance of current pseudo-labeling methods under the context of class-imbalance. To alleviate this problem, we design a bias adaptive classifier that targets the imbalanced SSL setups. The core idea is to automatically assimilate the training bias caused by class imbalance via the bias adaptive classifier, which is composed of a novel bias attractor and the original linear classifier. The bias attractor is designed as a light-weight residual network and learned through a bi-level learning framework, which enables the bias adaptive classifier to fit imbalanced training data, while the linear classifier can provide unbiased label prediction for each class. We conduct extensive experiments under various imbalanced semi-supervised setups, and the results demonstrate that our method can be applied to different pseudo-labeling models and is superior to current state-of-the-art methods. | This work proposes a bi-level learning framework to learn a tailored classifier for imbalanced semi-supervised learning. |
Recently, sequence learning methods have been applied to the problem of off-policy
Reinforcement Learning, including the seminal work on Decision Transformers,
which employs transformers for this task. Since transformers are parameter-heavy,
cannot benefit from history longer than a fixed window size, and are not computed
using recurrence, we set out to investigate the suitability of the S4 family of
models, which are based on state-space layers and have been shown to outperform
transformers, especially in modeling long-range dependencies. In this work, we
present two main algorithms: (i) an off-policy training procedure that works with
trajectories, while still maintaining the training efficiency of the S4 model. (ii) An
on-policy training procedure that is trained in a recurrent manner, benefits from
long-range dependencies, and is based on a novel stable actor-critic mechanism.
Our results indicate that our method outperforms multiple variants of decision
transformers, as well as the other baseline methods on most tasks, while reducing
the latency, number of parameters, and training time by several orders of magnitude,
making our approach more suitable for real-world RL | Replacing transformers with state-space layers for RL modeling. Also extended to on-policy training. |
Sequential labeling is a fundamental NLP task, forming the backbone of many applications.
Supervised learning of Seq2Seq models (like T5) has shown great success on these problems.
However there remains a significant disconnect between the training objectives of these models vs the metrics and desiderata we care about in practical applications.
For example, a practical sequence tagging application may want to optimize for a certain precision-recall trade-off (of the top-k predictions) which is quite different from the standard objective of maximizing the likelihood of the gold labeled sequence.
Thus to bridge this gap, we propose GROOT -- a simple yet effective framework for Generative Reward Optimization Of Text sequences.
GROOT works by training a generative sequential labeling model to match the decoder output distribution with that of the (black-box) reward function.
Using an iterative training regime, we first generate prediction candidates, then correct errors in them, and finally contrast those candidates (based on their reward values).
As demonstrated via extensive experiments on four public benchmarks, GROOT significantly improves all reward metrics.
Furthermore, GROOT also leads to improvements of the overall decoder distribution as evidenced by the quality gains of the top-k candidates. | This paper proposes a novel framework for iteratively training Seq2Seq models to directly optimize a given blackbox reward metric, showing its effectiveness on sequential labeling tasks. |
Recent methods for imitation learning directly learn a $Q$-function using an implicit reward formulation rather than an explicit reward function. However, these methods generally require implicit reward regularization to improve stability and often mistreat absorbing states. Previous works show that a squared norm regularization on the implicit reward function is effective, but do not provide a theoretical analysis of the resulting properties of the algorithms. In this work, we show that using this regularizer under a mixture distribution of the policy and the expert provides a particularly illuminating perspective: the original objective can be understood as squared Bellman error minimization, and the corresponding optimization problem minimizes a bounded $\chi^2$-Divergence between the expert and the mixture distribution. This perspective allows us to address instabilities and properly treat absorbing states. We show that our method, Least Squares Inverse Q-Learning (LS-IQ), outperforms state-of-the-art algorithms, particularly in environments with absorbing states. Finally, we propose to use an inverse dynamics model to learn from observations only. Using this approach, we retain performance in settings where no expert actions are available. | We propose a novel perspective on implicit L2 reward regularization for inverse reinforcement learning. |
Video is a promising source of knowledge for embodied agents to learn models of the world's dynamics. Large deep networks have become increasingly effective at modeling complex video data in a self-supervised manner, as evaluated by metrics based on human perceptual similarity or pixel-wise comparison. However, it remains unclear whether current metrics are accurate indicators of performance on downstream tasks. We find empirically that for planning robotic manipulation, existing metrics can be unreliable at predicting execution success. To address this, we propose a benchmark for action-conditioned video prediction in the form of a control benchmark that evaluates a given model for simulated robotic manipulation through sampling-based planning. Our benchmark, Video Prediction for Visual Planning ($\text{VP}^2$), includes simulated environments with $11$ task categories and $310$ task instance definitions, a full planning implementation, and training datasets containing scripted interaction trajectories for each task category. A central design goal of our benchmark is to expose a simple interface -- a single forward prediction call -- so it is straightforward to evaluate almost any action-conditioned video prediction model. We then leverage our benchmark to study the effects of scaling model size, quantity of training data, and model ensembling by analyzing five highly-performant video prediction models, finding that while scale can improve perceptual quality when modelling visually diverse settings, other attributes such as uncertainty awareness can also aid planning performance. | We find that existing video evaluation metrics are not always indicative of a model's performance during control, and propose a benchmark that directly evaluates video prediction models on simulated manipulation tasks by using them for planning. |
Models from the Vision Transformer (ViT) family have recently provided breakthrough results across image classification tasks such as ImageNet. Yet, they still face barriers to deployment, notably the fact that their accuracy can be severely impacted by compression techniques such as pruning. In this paper, we take a step towards addressing this issue by introducing \textit{Optimal ViT Surgeon (oViT)}, a new state-of-the-art method for the weight sparsification of Vision Transformers (ViT) models. At the technical level, oViT introduces a new weight pruning algorithm which leverages second-order information, specifically adapted to be both highly-accurate and efficient in the context of ViTs. We complement this accurate one-shot pruner with an in-depth investigation of gradual pruning, augmentation, and recovery schedules for ViTs, which we show to be critical for successful ViT compression. We validate our method via extensive experiments on classical ViT and DeiT models, as well as on newer variants, such as XCiT, EfficientFormer and Swin. Moreover, our results are even relevant to recently-proposed highly-accurate ResNets. Our results show for the first time that ViT-family models can in fact be pruned to high sparsity levels (e.g. $\geq 75\%$) with low impact on accuracy ($\leq 1\%$ relative drop), and that our approach outperforms prior methods by significant margins at high sparsities. In addition, we show that our method is compatible with structured pruning methods and quantization, and that it can lead to significant speedups on a sparsity-aware inference engine.
| We have proposed a new framework for efficient compression of Vision Transformers with the novel pruning method leveraging second order information and optimization of the training procedure. |
Federated learning (FL) is an efficient learning framework that assists distributed machine learning when data cannot be shared with a centralized server. Recent advancements in FL use predefined architecture-based learning for all the clients. However, given that clients' data are invisible to the server and data distributions are non-identical across clients, a predefined architecture discovered in a centralized setting may not be an optimal solution for all the clientsin FL. Motivated by this challenge, we introduce SPIDER, an algorithmic framework that aims to Search Personalized neural architecture for feDERated learning. SPIDER is designed based on two unique features: (1) alternately optimizing one architecture-homogeneous global model (Supernet) in a generic FL manner and one architecture-heterogeneous local model that is connected to the global model by weight-sharing-based regularization (2) achieving architecture-heterogeneous local model by an operation-level perturbation based neural architecture search method. Experimental results demonstrate that SPIDER outperforms other state-of-the-art personalization methods with much fewer times of hyperparameter tuning. | SPIDER searches and trains heterogeneous architectures in a federated learning setting to achieve the objective of personalization. |