conference
stringclasses 6
values | title
stringlengths 2
176
| abstract
stringlengths 2
5k
| decision
stringclasses 11
values |
|---|---|---|---|
ICLR.cc/2019/Conference | Modeling the Long Term Future in Model-Based Reinforcement Learning | In model-based reinforcement learning, the agent interleaves between model learning and planning. These two components are inextricably intertwined. If the model is not able to provide sensible long-term prediction, the executed planer would exploit model flaws, which can yield catastrophic failures. This paper focuses on building a model that reasons about the long-term future and demonstrates how to use this for efficient planning and exploration. To this end, we build a latent-variable autoregressive model by leveraging recent ideas in variational inference. We argue that forcing latent variables to carry future information through an auxiliary task substantially improves long-term predictions. Moreover, by planning in the latent space, the planner's solution is ensured to be within regions where the model is valid. An exploration strategy can be devised by searching for unlikely trajectories under the model. Our methods achieves higher reward faster compared to baselines on a variety of tasks and environments in both the imitation learning and model-based reinforcement learning settings. | Accept (Poster) |
ICLR.cc/2022/Conference | The Evolution of Out-of-Distribution Robustness Throughout Fine-Tuning | Although machine learning models typically experience a drop in performance on out-of-distribution data, accuracies on in- versus out-of-distribution data are widely observed to follow a single linear trend when evaluated across a testbed of models. Models that are more accurate on the out-of-distribution data relative to this baseline exhibit “effective robustness” and are exceedingly rare. Identifying such models, and understanding their properties, is key to improving out-of-distribution performance. We conduct a thorough empirical investigation of effective robustness during fine-tuning and surprisingly find that models pre-trained on larger datasets exhibit effective robustness during training that vanishes at convergence. We study how properties of the data influence effective robustness, and we show that it increases with the larger size, more diversity, and higher example difficulty of the dataset. We also find that models that display effective robustness are able to correctly classify 10% of the examples that no other current testbed model gets correct. Finally, we discuss several strategies for scaling effective robustness to the high-accuracy regime to improve the out-of-distribution accuracy of state-of-the-art models. | Reject |
ICLR.cc/2023/Conference | Making Substitute Models More Bayesian Can Enhance Transferability of Adversarial Examples | The transferability of adversarial examples across deep neural networks (DNNs) is the crux of many black-box attacks. Many prior efforts have been devoted to improving the transferability via increasing the diversity in inputs of some substitute models. In this paper, by contrast, we opt for the diversity in substitute models and advocate to attack a Bayesian model for achieving desirable transferability. Deriving from the Bayesian formulation, we develop a principled strategy for possible finetuning, which can be combined with many off-the-shelf Gaussian posterior approximations over DNN parameters. Extensive experiments have been conducted to verify the effectiveness of our method, on common benchmark datasets, and the results demonstrate that our method outperforms recent state-of-the-arts by large margins (roughly 19% absolute increase in average attack success rate on ImageNet), and, by combining with these recent methods, further performance gain can be obtained. Our code: https://github.com/qizhangli/MoreBayesian-attack. | Accept: poster |
ICLR.cc/2023/Conference | Deep autoregressive density nets vs neural ensembles for model-based offline reinforcement learning | We consider the problem of offline reinforcement learning where only a set of system transitions is made available for policy optimization. Following recent advances in the field, we consider a model-based reinforcement learning algorithm that infers the system dynamics from the available data and performs policy optimization on imaginary model rollouts. This approach is vulnerable to exploiting model errors which can lead to catastrophic failures on the real system. The standard solution is to rely on ensembles for uncertainty heuristics and to avoid exploiting the model where it is too uncertain. We challenge the popular belief that we must resort to ensembles by showing that better performance can be obtained with a single well-calibrated autoregressive model on the D4RL benchmark. We also analyze static metrics of model-learning and conclude on the important model properties for the final performance of the agent. | Reject |
ICLR.cc/2023/Conference | SAGE: Semantic-Aware Global Explanations for Named Entity Recognition | In the last decades, deep learning approaches achieved impressive results in many research fields,
such as Computer Vision and Natural Language Processing (NLP).
NLP in particular has greatly benefit from unsupervised methods that allow to learn distributed representation of language.
On the race for better performances Language Models have reached hundred of billions parameters nowadays.
Despite the remarkable results, deep models are still far from being fully exploited in real world applications.
Indeed, these approaches are black-boxes, i.e. they are not interpretable by design nor explainable, which is often crucial to make decisions in business.
Several task-agnostic methods have been proposed in literature to explain models' decisions.
Most techniques rely on the "local" assumption, i.e. explanations are made example-wise.
In this paper instead, we present a post-hoc method to produce highly interpretable global rules to explain NLP classifiers.
Rules are extracted with a data mining approach on a semantically enriched input representation, instead of using words/wordpieces solely.
Semantic information yields more abstract and general rules that are both more explanatory and less complex, while being also better at reflecting the model behaviour.
In the experiments we focus on Named Entity Recognition, an NLP task where explainability is under-investigated.
We explain the predictions of BERT NER classifiers trained on two popular benchmarks, CoNLL03 and Ontonotes, and compare our model against LIME. | Reject |
ICLR.cc/2022/Conference | Variability of Neural Networks and Han-Layer: A Variability-Inspired Model | What makes an artificial neural network easier to train or to generalize better than its peers? We introduce a notion of variability to view such issues under the setting of a fixed number of parameters which is, in general, a dominant cost-factor. Experiments verify that variability correlates positively to the number of activations and negatively to a phenomenon called Collapse to Constants, which is related but not identical to vanishing gradient. Further experiments on stylized problems show that variability is indeed a key performance indicator for fully-connected neural networks. Guided by variability considerations, we propose a new architecture called Householder-absolute neural layers, or Han-layers for short, to build high variability networks with a guaranteed immunity to gradient vanishing or exploding.
On small stylized models, Han-layer networks exhibit a far superior generalization ability over fully-connected networks. Extensive empirical results demonstrate that, by judiciously replacing fully-connected layers in large-scale networks such as MLP-Mixers, Han-layers can greatly reduce the number of model parameters while maintaining or improving generalization performance. We will also briefly discuss current limitations of the proposed Han-layer architecture.
| Reject |
ICLR.cc/2023/Conference | Limitations of Piecewise Linearity for Efficient Robustness Certification | Certified defenses against small-norm adversarial examples have received growing attention in recent years; though certified accuracies of state-of-the-art methods remain far below their non-robust counterparts, despite the fact that benchmark datasets have been shown to be well-separated at far larger radii than the literature generally attempts to certify. In this work, we offer insights that identify potential factors in this performance gap. Specifically, our analysis reveals that piecewise linearity imposes fundamental limitations on the tightness of leading certification techniques. These limitations are felt in practical terms as a greater need for capacity in models hoped to be certified efficiently. Moreover, this is _in addition_ to the capacity necessary to learn a robust boundary, studied in prior work. However, we argue that addressing the limitations of piecewise linearity through scaling up model capacity may give rise to potential difficulties---particularly regarding robust generalization---therefore, we conclude by suggesting that developing _smooth_ activation functions may be the way forward for advancing the performance of certified neural networks. | Reject |
ICLR.cc/2018/Conference | SEARNN: Training RNNs with global-local losses | We propose SEARNN, a novel training algorithm for recurrent neural networks (RNNs) inspired by the "learning to search" (L2S) approach to structured prediction. RNNs have been widely successful in structured prediction applications such as machine translation or parsing, and are commonly trained using maximum likelihood estimation (MLE). Unfortunately, this training loss is not always an appropriate surrogate for the test error: by only maximizing the ground truth probability, it fails to exploit the wealth of information offered by structured losses. Further, it introduces discrepancies between training and predicting (such as exposure bias) that may hurt test performance. Instead, SEARNN leverages test-alike search space exploration to introduce global-local losses that are closer to the test error. We first demonstrate improved performance over MLE on two different tasks: OCR and spelling correction. Then, we propose a subsampling strategy to enable SEARNN to scale to large vocabulary sizes. This allows us to validate the benefits of our approach on a machine translation task. | Accept (Poster) |
ICLR.cc/2018/Conference | Improving generalization by regularizing in $L^2$ function space | Learning rules for neural networks necessarily include some form of regularization. Most regularization techniques are conceptualized and implemented in the space of parameters. However, it is also possible to regularize in the space of functions. Here, we propose to measure networks in an $L^2$ Hilbert space, and test a learning rule that regularizes the distance a network can travel through $L^2$-space each update. This approach is inspired by the slow movement of gradient descent through parameter space as well as by the natural gradient, which can be derived from a regularization term upon functional change. The resulting learning rule, which we call Hilbert-constrained gradient descent (HCGD), is thus closely related to the natural gradient but regularizes a different and more calculable metric over the space of functions. Experiments show that the HCGD is efficient and leads to considerably better generalization. | Reject |
ICLR.cc/2023/Conference | AutoSKDBERT: Learn to Stochastically Distill BERT | 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). | Reject |
ICLR.cc/2021/Conference | Simplifying Models with Unlabeled Output Data | We focus on prediction problems with high-dimensional outputs that are subject to output validity constraints, e.g. a pseudocode-to-code translation task where the code must compile. For these problems, labeled input-output pairs are expensive to obtain, but "unlabeled" outputs, i.e. outputs without corresponding inputs, are freely available and provide information about output validity (e.g. code on GitHub). In this paper, we present predict-and-denoise, a framework that can leverage unlabeled outputs. Specifically, we first train a denoiser to map possibly invalid outputs to valid outputs using synthetic perturbations of the unlabeled outputs. Second, we train a predictor composed with this fixed denoiser. We show theoretically that for a family of functions with a high-dimensional discrete valid output space, composing with a denoiser reduces the complexity of a 2-layer ReLU network needed to represent the function and that this complexity gap can be arbitrarily large. We evaluate the framework empirically on several datasets, including image generation from attributes and pseudocode-to-code translation. On the SPoC pseudocode-to-code dataset, our framework improves the proportion of code outputs that pass all test cases by 3-5% over a baseline Transformer. | Reject |
ICLR.cc/2022/Conference | EfficientPhys: Enabling Simple, Fast, and Accurate Camera-Based Vitals Measurement | Camera-based physiological measurement is a growing field with neural models providing state-the-art-performance. Prior research have explored various "end-to-end'' models; however these methods still require several preprocessing steps. These additional operations are often non-trivial to implement making replication and deployment difficult and can even have a higher computational budget than the "core'' network itself. In this paper, we propose two novel and efficient neural models for camera-based physiological measurement called EfficientPhys that remove the need for face detection, segmentation, normalization, color space transformation or any other preprocessing steps. Using an input of raw video frames, our models achieve state-of-the-art accuracy on three public datasets. We show that this is the case whether using a transformer or convolutional backbone. We further evaluate the latency of the proposed networks and show that our most light weight network also achieves a 33\% improvement in efficiency.
| Reject |
ICLR.cc/2023/Conference | Multiplane NeRF-Supervised Disentanglement of Depth and Camera Pose from Videos | We propose to perform self-supervised disentanglement of depth and camera pose from large-scale videos. We introduce an Autoencoder-based method to reconstruct the input video frames for training, without using any ground-truth annotations of depth and camera. The encoders for our model will estimate the monocular depth and camera pose as the disentangled representations. The decoder will then construct a Multiplane NeRF representation based on the depth encoder feature, and perform rendering to reconstruct the input frames with the estimated camera. The disentanglement is learned with the reconstruction error, based on the assumption that the scene structure does not change in short periods of time in videos. Once the model is learned, it can be applied to multiple applications including depth estimation, camera pose estimation, and single image novel view synthesis. We show substantial improvements over previous self-supervised approaches on all tasks and even better results than counterparts trained with camera ground-truths in some applications. Our code will be made publicly available. | Withdrawn |
ICLR.cc/2022/Conference | Adapt to Adaptation: Learning to Personalize for Cross-Silo Federated Learning | The goal of conventional federated learning (FL) is to train a global model for a federation of clients with decentralized data, reducing the systemic privacy risk of centralized training. The distribution shift across non-IID datasets, also known as the data heterogeneity, often poses a challenge for this one-global-model-fits-all solution. In this work, we propose APPLE, a personalized cross-silo FL framework that adaptively learns how much each client can benefit from other clients’ models. We also introduce a method to flexibly control the focus of training APPLE between global and local objectives. We empirically evaluate our method's convergence and generalization behavior and performed extensive experiments on two benchmark datasets and two medical imaging datasets under two non-IID settings. The results show that the proposed personalized FL framework, APPLE, achieves state-of-the-art performance compared to several other personalized FL approaches in the literature. | Withdrawn |
ICLR.cc/2023/Conference | Offline RL for Natural Language Generation with Implicit Language Q Learning | Large language models distill broad knowledge from text corpora. However, they can be inconsistent when it comes to completing user specified tasks. This issue can be addressed by finetuning such models via supervised learning on curated datasets, or via reinforcement learning. In this work, we propose a novel offline RL method, implicit language Q-learning (ILQL), designed for use on language models, that combines both the flexible utility maximization framework of RL algorithms with the ability of supervised learning to leverage previously collected data, as well as its simplicity and stability. Our method employs a combination of value conservatism alongside an implicit dataset support constraint in learning value functions, which are then used to guide language model generations towards maximizing user-specified utility functions. In addition to empirically validating ILQL, we present a detailed empirical analysis of situations where offline RL can be useful in natural language generation settings, demonstrating how it can be a more effective utility optimizer than prior approaches for end-to-end dialogue, and how it can effectively optimize high variance reward functions based on subjective judgement, such as whether to label a comment as toxic or not. | Accept: poster |
ICLR.cc/2020/Conference | DDSP: Differentiable Digital Signal Processing | Most generative models of audio directly generate samples in one of two domains: time or frequency. While sufficient to express any signal, these representations are inefficient, as they do not utilize existing knowledge of how sound is generated and perceived. A third approach (vocoders/synthesizers) successfully incorporates strong domain knowledge of signal processing and perception, but has been less actively researched due to limited expressivity and difficulty integrating with modern auto-differentiation-based machine learning methods. In this paper, we introduce the Differentiable Digital Signal Processing (DDSP) library, which enables direct integration of classic signal processing elements with deep learning methods. Focusing on audio synthesis, we achieve high-fidelity generation without the need for large autoregressive models or adversarial losses, demonstrating that DDSP enables utilizing strong inductive biases without losing the expressive power of neural networks. Further, we show that combining interpretable modules permits manipulation of each separate model component, with applications such as independent control of pitch and loudness, realistic extrapolation to pitches not seen during training, blind dereverberation of room acoustics, transfer of extracted room acoustics to new environments, and transformation of timbre between disparate sources. In short, DDSP enables an interpretable and modular approach to generative modeling, without sacrificing the benefits of deep learning. The library will is available at https://github.com/magenta/ddsp and we encourage further contributions from the community and domain experts.
| Accept (Spotlight) |
ICLR.cc/2022/Conference | Reward Learning as Doubly Nonparametric Bandits: Optimal Design and Scaling Laws | Specifying reward functions for complex tasks like object manipulation or driving is challenging to do by hand. Reward learning seeks to address this by learning a reward model using human feedback on selected query policies. This shifts the burden of reward specification to the optimal design of the queries. We propose a theoretical framework for studying reward learning and the associated optimal experiment design problem. Our framework models rewards and policies as nonparametric functions belonging to subsets of Reproducing Kernel Hilbert Spaces (RKHSs). The learner receives (noisy) oracle access to a true reward and must output a policy that performs well under the true reward. For this setting, we first derive non-asymptotic excess risk bounds for a simple plug-in estimator based on ridge regression. We then solve the query design problem by optimizing these risk bounds with respect to the choice of query set and obtain a finite sample statistical rate, which depends primarily on the eigenvalue spectrum of a certain linear operator on the RKHSs. Despite the generality of these results, our bounds are stronger than previous bounds developed for more specialized problems. We specifically show that the well-studied problem of Gaussian process (GP) bandit optimization is a special case of our framework, and that our bounds either improve or are competitive with known regret guarantees for the Mat\'ern kernel. | Reject |
ICLR.cc/2020/Conference | Imagine That! Leveraging Emergent Affordances for Tool Synthesis in Reaching Tasks | In this paper we investigate an artificial agent's ability to perform task-focused tool synthesis via imagination. Our motivation is to explore the richness of information captured by the latent space of an object-centric generative model - and how to exploit it. In particular, our approach employs activation maximisation of a task-based performance predictor to optimise the latent variable of a structured latent-space model in order to generate tool geometries appropriate for the task at hand. We evaluate our model using a novel dataset of synthetic reaching tasks inspired by the cognitive sciences and behavioural ecology. In doing so we examine the model's ability to imagine tools for increasingly complex scenario types, beyond those seen during training. Our experiments demonstrate that the synthesis process modifies emergent, task-relevant object affordances in a targeted and deliberate way: the agents often specifically modify aspects of the tools which relate to meaningful (yet implicitly learned) concepts such as a tool's length, width and configuration. Our results therefore suggest, that task relevant object affordances are implicitly encoded as directions in a structured latent space shaped by experience. | Reject |
ICLR.cc/2022/Conference | BLUnet: Arithmetic-free Inference with Bit-serialised Table Lookup Operation for Efficient Deep Neural Networks | Deep neural networks (DNNs) are both computation and memory intensive. Large amounts of costly arithmetic multiply-accumulate (MAC) operations and data movement hinder its application to edge AI where DNN models are required to run on energy-constrained platforms. Table lookup operations have potential advantages over traditional arithmetic multiplication and addition operations in terms of both energy consumption and latency in hardware implementations for DNN design. Moreover, the integration of weights into the table lookup operation eliminates costly weight movements. However, the challenge of using table lookups is in scaling. In particular, the size and lookup times of tables grow exponentially with the fan-in of the tables. In this paper, we propose BLUnet, a table lookup-based DNN model with bit-serialized input to overcome this challenge. Using binarized time series inputs, we successfully solve the fan-in issue of lookup tables. BLUnet not only achieves high efficiency but also the same accuracies as MAC-based neural networks. We experimented with popular models in computer vision applications to confirm this. Our experimental results show that compared to MAC-based baseline designs as well as the state-of-the-art solutions, BLUnet achieves orders of magnitude improvement in energy efficiencies. | Withdrawn |
ICLR.cc/2023/Conference | GuardHFL: Privacy Guardian for Heterogeneous Federated Learning | Heterogeneous federated learning (HFL) enables clients with different computation and communication capabilities to collaboratively train their own customized models via a query-response paradigm on auxiliary datasets. However, such paradigm raises serious privacy issues due to the leakage of highly sensitive query samples and response predictions. Although existing secure querying solutions may be extended to enhance the privacy of HFL with non-trivial adaptation, they suffer from two key limitations: (1) lacking customized protocol designs and (2) relying on heavy cryptographic primitives, which could lead to poor performance. In this work, we put forth GuardHFL, the first-of-its-kind efficient and privacy-preserving HFL framework. GuardHFL is equipped with a novel HFL-friendly secure querying scheme that is built on lightweight secret sharing and symmetric-key techniques. Its core is a set of customized multiplication and comparison protocols, which substantially boost the execution efficiency. Extensive evaluations demonstrate that GuardHFL outperforms the state-of-the-art works by up to two orders of magnitude in efficiency. | Reject |
ICLR.cc/2021/Conference | DarKnight: A Data Privacy Scheme for Training and Inference of Deep Neural Networks | Protecting the privacy of input data is of growing importance as machine learning methods reach new application domains.
In this paper, we provide a unified training and inference framework for large DNNs while protecting input privacy and computation integrity. Our approach called DarKnight uses a novel data blinding strategy using matrix masking to create input obfuscation within a trusted execution environment (TEE). Our rigorous mathematical proof demonstrates that our blinding process provides an information-theoretic privacy guarantee by bounding information leakage. The obfuscated data can then be offloaded to any GPU for accelerating linear operations on blinded data. The results from linear operations on blinded data are decoded before performing non-linear operations within the TEE. This cooperative execution allows DarKnight to exploit the computational power of GPUs to perform linear operations while exploiting TEEs to protect input privacy. We implement DarKnight on an Intel SGX TEE augmented with a GPU to evaluate its performance. | Reject |
ICLR.cc/2023/Conference | Global-Scale Species Mapping From Crowdsourced Data | Estimating the geographical range of a species from in situ observational data is a challenging and important geospatial prediction problem. Given a set of locations indicating where a species has been observed, the goal is to learn a model that can predict how likely it is for the species to be present at any other location. While this is a well-studied problem, traditional approaches are unable to take advantage of more recently available large-scale datasets that cover many locations and species. We propose a new approach that jointly estimates the geographical ranges of tens of thousands of different species simultaneously. We develop a series of benchmark evaluation tasks that measure different aspects of the species range and spatial representation learning problems. We show that our approach scales both in terms of amount of training data and species, where adding more data enables the models to learn better spatial representations that generalize to other species. Despite being only trained on weakly supervised crowdsourced data, our models can approach the predictions of current expert-developed gold standard models. | Reject |
ICLR.cc/2022/Conference | Unified Recurrence Modeling for Video Action Anticipation | Forecasting future events based on evidence of current conditions is an innate skill of human beings, and key for predicting the outcome of any decision making. In artificial vision for example, we would like to predict the next human action before it is actually performed, without observing the future video frames associated to it. Computer vision models for action anticipation are expected to collect the subtle evidence in the preamble of the target actions. In prior studies recurrence modeling often leads to better performance, and the strong temporal inference is assumed to be a key element for reasonable prediction. To this end, we propose a unified recurrence modeling for video action anticipation by generalizing the recurrence mechanism from sequence into graph representation via message passing. The information flow in space-time can be described by the interaction between vertices and edges, and the changes of vertices for each incoming frame reflects the underlying dynamics. Our model leverages self-attention for all building blocks in the graph modeling, and we introduce different edge learning strategies can be end-to-end optimized while updating the vertices. Our experimental results demonstrate that our modeling method is light-weight, efficient, and outperforms all previous works on the large-scale EPIC-Kitchen dataset. | Reject |
ICLR.cc/2021/Conference | Sparse Linear Networks with a Fixed Butterfly Structure: Theory and Practice | A butterfly network consists of logarithmically many layers, each with a linear number of non-zero weights (pre-specified). The fast Johnson-Lindenstrauss transform (FJLT) can be represented as a butterfly network followed by a random projection to a subset of the coordinates. Moreover, a random matrix based on FJLT with high probability approximates the action of any matrix on a vector. Motivated by these facts, we propose to replace a dense linear layer in any neural network by an architecture based on the butterfly network. The proposed architecture significantly improves upon the quadratic number of weights required in a standard dense layer to nearly linear with little compromise in expressibility of the resulting operator. In a collection of wide variety of experiments, including supervised prediction on both the NLP and vision data, we show that this not only produces results that match and often outperform existing well-known architectures, but it also offers faster training and prediction in deployment. To understand the optimization problems posed by neural networks with a butterfly network, we study the optimization landscape of the encoder-decoder network, where the encoder is replaced by a butterfly network followed by a dense linear layer in smaller dimension. Theoretical result presented in the paper explain why the training speed and outcome are not compromised by our proposed approach. Empirically we demonstrate that the network performs as well as the encoder-decoder network. | Reject |
ICLR.cc/2022/Conference | A Zest of LIME: Towards Architecture-Independent Model Distances | Definitions of the distance between two machine learning models either characterize the similarity of the models' predictions or of their weights. While similarity of weights is attractive because it implies similarity of predictions in the limit, it suffers from being inapplicable to comparing models with different architectures. On the other hand, the similarity of predictions is broadly applicable but depends heavily on the choice of model inputs during comparison. In this paper, we instead propose to compute distance between black-box models by comparing their Local Interpretable Model-Agnostic Explanations (LIME). To compare two models, we take a reference dataset, and locally approximate the models on each reference point with linear models trained by LIME. We then compute the cosine distance between the concatenated weights of the linear models. This yields an approach that is both architecture-independent and possesses the benefits of comparing models in weight space. We empirically show that our method, which we call Zest, can be applied to two problems that require measurements of model similarity: detecting model stealing and machine unlearning. | Accept (Poster) |
ICLR.cc/2021/Conference | Deep Partition Aggregation: Provable Defenses against General Poisoning Attacks | Adversarial poisoning attacks distort training data in order to corrupt the test-time behavior of a classifier. A provable defense provides a certificate for each test sample, which is a lower bound on the magnitude of any adversarial distortion of the training set that can corrupt the test sample's classification.
We propose two novel provable defenses against poisoning attacks: (i) Deep Partition Aggregation (DPA), a certified defense against a general poisoning threat model, defined as the insertion or deletion of a bounded number of samples to the training set --- by implication, this threat model also includes arbitrary distortions to a bounded number of images and/or labels; and (ii) Semi-Supervised DPA (SS-DPA), a certified defense against label-flipping poisoning attacks. DPA is an ensemble method where base models are trained on partitions of the training set determined by a hash function. DPA is related to both subset aggregation, a well-studied ensemble method in classical machine learning, as well as to randomized smoothing, a popular provable defense against evasion (inference) attacks. Our defense against label-flipping poison attacks, SS-DPA, uses a semi-supervised learning algorithm as its base classifier model: each base classifier is trained using the entire unlabeled training set in addition to the labels for a partition. SS-DPA significantly outperforms the existing certified defense for label-flipping attacks (Rosenfeld et al., 2020) on both MNIST and CIFAR-10: provably tolerating, for at least half of test images, over 600 label flips (vs. < 200 label flips) on MNIST and over 300 label flips (vs. 175 label flips) on CIFAR-10. Against general poisoning attacks where no prior certified defenses exists, DPA can certify $\geq$ 50% of test images against over 500 poison image insertions on MNIST, and nine insertions on CIFAR-10. These results establish new state-of-the-art provable defenses against general and label-flipping poison attacks. Code is available at https://github.com/alevine0/DPA | Accept (Poster) |
ICLR.cc/2018/Conference | Certifying Some Distributional Robustness with Principled Adversarial Training | Neural networks are vulnerable to adversarial examples and researchers have proposed many heuristic attack and defense mechanisms. We address this problem through the principled lens of distributionally robust optimization, which guarantees performance under adversarial input perturbations. By considering a Lagrangian penalty formulation of perturbing the underlying data distribution in a Wasserstein ball, we provide a training procedure that augments model parameter updates with worst-case perturbations of training data. For smooth losses, our procedure provably achieves moderate levels of robustness with little computational or statistical cost relative to empirical risk minimization. Furthermore, our statistical guarantees allow us to efficiently certify robustness for the population loss. For imperceptible perturbations, our method matches or outperforms heuristic approaches.
| Accept (Oral) |
ICLR.cc/2022/Conference | PRNet: A Progressive Regression Network for No-Reference User-Generated-Content Video Quality Assessment | Non-professional video, commonly known as User Generated Content (UGC) has become very popular in today’s video sharing applications. However, objectively perceptual quality assessment of UGC-videos is still a challenge problem, which is arose from many reasons. First, the pristine sources of UGC-videos are not available, which makes the appropriate technique is the no-reference NR video quality assessment VQA (NR-VQA). Another factor leads the NR-UGC-VQA to a challenge is that subjective mean option scores (MOS) of all the UGC-datasets are not uniformly distributed. The largest UGC video dataset---YouTube-UGC still faces a problem that the database has right-skewed MOS distribution. In addition, authentic degradations occurred in the videos are not unique, therefore, not predicable. For example, an over- or under-exposure image/video, brightness and contrast static information is important for evaluation. Only employing verified priori statistic knowledge or generalized learning knowledge may not cover all possible distortions. To solve these problems, we introduce a novel NR-VQA framework---Progressive Regress Network (PRNet) in this paper. For the skewed MOS problem, a progressive regression model is proposed, which utilizes the coarse-to-fine strategy during the training process. This strategy can turn sparse subjective human rating scores into integers with denser samples, which can solve the in-balanced sample problem and make the training progress smoother. For the unpredictable distortions problem, a wide and deep model based on our PRNet is developed, which employs both low-level features generated from natural scene statistics (NSS) and high-level semantic features extracted by deep neural networks, to fuse memorizing priori knowledge and generalizing learning features. Our experimental results demonstrate that our proposed method PRNet achieves state-of-the-art performance in currently three main popular UGC-VQA datasets (KoNVid-1K, LIVE-VQC, and YouTube-UGC). | Withdrawn |
ICLR.cc/2021/Conference | Constraint-Driven Explanations of Black-Box ML Models | Modern machine learning techniques have enjoyed widespread success, but are plagued by lack of transparency in their decision making, which has led to the emergence of the field of explainable AI. One popular approach called LIME, seeks to explain an opaque model's behavior, by training a surrogate interpretable model to be locally faithful on perturbed instances.
Despite being model-agnostic and easy-to-use, it is known that LIME's explanations can be unstable and are susceptible to adversarial attacks as a result of Out-Of-Distribution (OOD) sampling. Quality of explanations is also calculated heuristically, and lacks a strong theoretical foundation. In spite of numerous attempts to remedy some of these issues, making the LIME framework more trustworthy and reliable remains an open problem.
In this work, we demonstrate that the OOD sampling problem stems from rigidity of the perturbation procedure. To resolve this issue, we propose a theoretically sound framework based on uniform sampling of user-defined subspaces. Through logical constraints, we afford the end-user the flexibility to delineate the precise subspace of the input domain to be explained. This not only helps mitigate the problem of OOD sampling, but also allow experts to drill down and uncover bugs deep inside the model. For testing the quality of generated explanations, we develop an efficient estimation algorithm that is able to certifiably measure the true value of metrics such as fidelity up to any desired degree of accuracy, which can help in building trust in the generated explanations. Our framework called CLIME can be applied to any ML model, and extensive experiments demonstrate its versatility on real-world problems.
| Reject |
ICLR.cc/2022/Conference | Analyzing the Implicit Position Encoding Ability of Transformer Decoder | A common limitation of Transformer Encoder's self-attention mechanism is that it cannot automatically capture the information of word order, so one needs to feed the explicit position encodings into the target model. On the other hand, Transformer Decoder with the auto-regressive attention masks is naturally sensitive to the word order information. In this work, based on the analysis of implicit position encoding power of Transformer Decoder, we obtain the conditions that at least two or more layers are required for the Decoder to encode word positions. To examine the correlations between the implicit and explicit position encodings respectively from the Transformer Encoder and Decoder, extensive experiments conducted on two large Wikipedia datasets demonstrate that all kinds of explicit position encoding mechanisms improve the performance of Decoder, but the gap of learnable position embeddings is smaller than the others. To make use of the power of implicit position encoding, we propose a new model, called \textit{DecBERT}, and fine-tune it on GLUE benchmarks. Experimental results show that (1) the implicit position encoding ability is strong enough to enhance language modeling and perform well on downstream tasks; and (2) our model accelerates the pre-training process and achieves superior performances than the baseline systems when pre-training with the same amount of computational resource. | Withdrawn |
ICLR.cc/2020/Conference | Defense against Adversarial Examples by Encoder-Assisted Search in the Latent Coding Space | Deep neural networks were shown to be vulnerable to crafted adversarial perturbations, and thus bring serious safety problems. To solve this problem, we proposed $\text{AE-GAN}_\text{+sr}$, a framework for purifying input images by searching a closest natural reconstruction with little computation. We first build a reconstruction network AE-GAN, which adapted auto-encoder by introducing adversarial loss to the objective function. In this way, we can enhance the generative ability of decoder and preserve the abstraction ability of encoder to form a self-organized latent space. In the inference time, when given an input, we will start a search process in the latent space which aims to find the closest reconstruction to the given image on the distribution of normal data. The encoder can provide a good start point for the searching process, which saves much computation cost. Experiments show that our method is robust against various attacks and can reach comparable even better performance to similar methods with much fewer computations. | Reject |
ICLR.cc/2023/Conference | Capturing the Motion of Every Joint: 3D Human Pose and Shape Estimation with Independent Tokens | In this paper we present a novel method to estimate 3D human pose and shape from monocular videos. This task requires directly recovering pixel-alignment 3D human pose and body shape from monocular images or videos, which is challenging due to its inherent ambiguity. To improve precision, existing methods highly rely on the initialized mean pose and shape as prior estimates and parameter regression with an iterative error feedback manner. In addition, video-based approaches model the overall change over the image-level features to temporally enhance the single-frame feature, but fail to capture the rotational motion at the joint level, and cannot guarantee local temporal consistency. To address these issues, we propose a novel Transformer-based model with a design of independent tokens. First, we introduce three types of tokens independent of the image feature: \textit{joint rotation tokens, shape token, and camera token}.
By progressively interacting with image features through Transformer layers, these tokens learn to encode the prior knowledge of human 3D joint rotations, body shape, and position information from large-scale data, and are updated to estimate SMPL parameters conditioned on a given image. Second, benefiting from the proposed token-based representation, we further use a temporal model to focus on capturing the rotational temporal information of each joint, which is empirically conducive to preventing large jitters in local parts. Despite being conceptually simple, the proposed method attains superior performances on the 3DPW and Human3.6M datasets. Using ResNet-50 and Transformer architectures, it obtains 42.0 mm error on the PA-MPJPE metric of the challenging 3DPW, outperforming state-of-the-art counterparts by a large margin. Code will be publicly available\footnote{\url{https://github.com/yangsenius/INT_HMR_Model}}. | Accept: notable-top-25% |
ICLR.cc/2020/Conference | JAUNE: Justified And Unified Neural language Evaluation | We review the limitations of BLEU and ROUGE -- the most popular metrics used to assess reference summaries against hypothesis summaries, and introduce JAUNE: a set of criteria for what a good metric should behave like and propose concrete ways to use recent Transformers-based Language Models to assess reference summaries against hypothesis summaries.
| Reject |
ICLR.cc/2021/Conference | Quickest change detection for multi-task problems under unknown parameters | We consider the quickest change detection problem where both the parameters of pre- and post- change distributions are unknown, which prevent the use of classical simple hypothesis testing. Without additional assumptions, optimal solutions are not tractable as they rely on some minimax and robust variant of the objective. As a consequence, change points might be detected too late for practical applications (in economics, health care or maintenance for instance).
Other approaches solve a relaxed version of the problem through the use of particular probability distributions or the use of domain knowledge.
We tackle this problem in the more complex Markovian case and we provide a new scalable approximate algorithm with near optimal performance that runs in $\mathcal{O}(1)$. | Reject |
ICLR.cc/2022/Conference | Zero-Round Active Learning | Active learning (AL) aims at reducing labeling efforts by identifying the most valuable unlabeled data points from a large pool. Traditional AL frameworks have two limitations: First, they perform data selection in a multi-round manner, which is time-consuming and impractical. Second, they usually assume that there are a small amount of labeled data points available \emph{in the same domain as} the data in the unlabeled pool.
Our paper investigates a new setting in active learning---how to conduct active learning without relying on pre-labeled data, which is an under-explored yet of great practical value.
Besides, we propose $D^2ULO$ as a solution that solves both issues, which leverages the idea of domain adaptation (DA) to train a data utility model that can effectively predict the utility for any given unlabeled data in the target domain once labeled. The trained data utility model can then be used to select high-utility data and at the same time, provide an estimate for the utility of the selected data. Our algorithm does not rely on any feedback from annotators in the target domain and hence, which is able to not only work standalone but also benefit existing multi-round active learning algorithms by providing a warm-start.
Our experiments show that $D^2ULO$ outperforms the existing state-of-the-art AL strategies equipped with domain adaptation over various domain shift settings (e.g., real-to-real data and synthetic-to-real data). Particularly, $D^2ULO$ is applicable to the scenario where source and target labels have mismatch, which is not supported by the existing works. | Withdrawn |
ICLR.cc/2021/Conference | A Simple Approach To Define Curricula For Training Neural Networks | In practice, sequence of mini-batches generated by uniform sampling of examples from the entire data is used for training neural networks. Curriculum learning is a training strategy that sorts the training examples by their difficulty and gradually exposes them to the learner. In this work, we propose two novel curriculum learning algorithms and empirically show their improvements in performance with convolutional and fully-connected neural networks on multiple real image datasets. Our dynamic curriculum learning algorithm tries to reduce the distance between the network weight and an optimal weight at any training step by greedily sampling examples with gradients that are directed towards the optimal weight. The curriculum ordering determined by our dynamic algorithm achieves a training speedup of $\sim 45\%$ in our experiments. We also introduce a new task-specific curriculum learning strategy that uses statistical measures such as standard deviation and entropy values to score the difficulty of data points in natural image datasets. We show that this new approach yields a mean training speedup of $\sim 43\%$ in the experiments we perform. Further, we also use our algorithms to learn why curriculum learning works. Based on our study, we argue that curriculum learning removes noisy examples from the initial phases of training, and gradually exposes them to the learner acting like a regularizer that helps in improving the generalization ability of the learner. | Reject |
ICLR.cc/2022/Conference | Latent Image Animator: Learning to Animate Images via Latent Space Navigation | Due to the remarkable progress of deep generative models, animating images has become increasingly efficient, whereas associated results have become increasingly realistic. Current animation-approaches commonly exploit structure representation extracted from driving videos. Such structure representation is instrumental in transferring motion from driving videos to still images. However, such approaches fail in case the source image and driving video encompass large appearance variation. Moreover, the extraction of structure information requires additional modules that endow the animation-model with increased complexity. Deviating from such models, we here introduce the Latent Image Animator (LIA), a self-supervised autoencoder that evades need for structure representation. LIA is streamlined to animate images by linear navigation in the latent space. Specifically, motion in generated video is constructed by linear displacement of codes in the latent space. Towards this, we learn a set of orthogonal motion directions simultaneously, and use their linear combination, in order to represent any displacement in the latent space. Extensive quantitative and qualitative analysis suggests that our model systematically and significantly outperforms state-of-art methods on VoxCeleb, Taichi and TED-talk datasets w.r.t. generated quality. | Accept (Poster) |
ICLR.cc/2020/Conference | HaarPooling: Graph Pooling with Compressive Haar Basis | Deep Graph Neural Networks (GNNs) are instrumental in graph classification and graph-based regression tasks. In these tasks, graph pooling is a critical ingredient by which GNNs adapt to input graphs of varying size and structure. We propose a new graph pooling operation based on compressive Haar transforms, called HaarPooling. HaarPooling is computed following a chain of sequential clusterings of the input graph. The input of each pooling layer is transformed by the compressive Haar basis of the corresponding clustering. HaarPooling operates in the frequency domain by the synthesis of nodes in the same cluster and filters out fine detail information by compressive Haar transforms. Such transforms provide an effective characterization of the data and preserve the structure information of the input graph. By the sparsity of the Haar basis, the computation of HaarPooling is of linear complexity. The GNN with HaarPooling and existing graph convolution layers achieves state-of-the-art performance on diverse graph classification problems. | Reject |
ICLR.cc/2023/Conference | Multi-Prompt Alignment for Multi-source Unsupervised Domain Adaptation | Most existing methods for multi-source unsupervised domain adaptation (UDA) rely on a common feature encoder to extract domain-invariant features. However, learning such an encoder involves updating the parameters of the entire network, which makes the optimization computationally expensive, particularly when coupled with min-max objectives. Inspired by recent advances in prompt learning that adapts high-capacity deep models for downstream tasks in a computationally economic way, we introduce Multi-Prompt Alignment (MPA), a simple yet efficient two-stage framework for multi-source UDA. Given a source and target domain pair, MPA first trains an individual prompt to minimize the domain gap through a contrastive loss, while tuning only a small set of parameters. Then, MPA derives a low-dimensional latent space through an auto-encoding process that maximizes the agreement of multiple learned prompts. The resulting embedding further facilitates generalization to unseen domains. Extensive experiments show that our method achieves state-of-the-art results on popular benchmark datasets while requiring substantially fewer tunable parameters. To the best of our knowledge, we are the first to apply prompt learning to the multi-source UDA problem and our method achieves the highest reported average accuracy of 54.1% on DomainNet, the most challenging UDA dataset to date, with only 15.9M parameters trained. More importantly, we demonstrate that the learned embedding space can be easily adapted to novel unseen domains. | Reject |
ICLR.cc/2023/Conference | DetectBench: An Object Detection Benchmark for OOD Generalization Algorithms | The consensus about practical machine learning tasks, such as object detection, is still the test data are drawn from the same distribution as the training data, which is known as IID (Independent and Identically Distributed). However, it can not avoid being confronted with OOD (Out-of-Distribution) scenarios in real practice. It is risky to apply an object detection algorithm without figuring out its OOD generalization performance. On the other hand, a plethora of OOD generalization algorithms has been proposed to amortize the gap between the in-house and open-world performances of machine learning systems. However, their effectiveness was only demonstrated in the image classification tasks. It is still an opening question of how these algorithms perform on complex and practical tasks. In this paper, we first specify the setting of OOD-OD (OOD generalization object detection). Then, we propose DetectBench consisting of four OOD-OD benchmark datasets to evaluate various object detection and OOD generalization algorithms. From extensive experiments on DetectBench, we find that existing OOD generalization algorithms fail dramatically when applied to the more practical object detection tasks. This raises questions over the current progress on a large number of these algorithms and whether they can be effective in practice beyond simple toy examples. For future work, we sincerely hope that DetectBench can serve as a foothold for OOD-OD research. | Reject |
ICLR.cc/2022/Conference | Model-augmented Prioritized Experience Replay | Experience replay is an essential component in off-policy model-free reinforcement learning (MfRL). Due to its effectiveness, various methods for calculating priority scores on experiences have been proposed for sampling. Since critic networks are crucial to policy learning, TD-error, directly correlated to $Q$-values, is one of the most frequently used features to compute the scores. However, critic networks often under- or overestimate $Q$-values, so it is often ineffective to learn to predict $Q$-values by sampled experiences based heavily on TD-error. Accordingly, it is valuable to find auxiliary features, which positively support TD-error in calculating the scores for efficient sampling. Motivated by this, we propose a novel experience replay method, which we call model-augmented prioritized experience replay (MaPER), that employs new learnable features driven from components in model-based RL (MbRL) to calculate the scores on experiences. The proposed MaPER brings the effect of curriculum learning for predicting $Q$-values better by the critic network with negligible memory and computational overhead compared to the vanilla PER. Indeed, our experimental results on various tasks demonstrate that MaPER can significantly improve the performance of the state-of-the-art off-policy MfRL and MbRL which includes off-policy MfRL algorithms in its policy optimization procedure. | Accept (Poster) |
ICLR.cc/2019/Conference | $A^*$ sampling with probability matching | Probabilistic methods often need to draw samples from a nontrivial distribution. $A^*$ sampling is a nice algorithm by building upon a top-down construction of a Gumbel process, where a large state space is divided into subsets and at each round $A^*$ sampling selects a subset to process. However, the selection rule depends on a bound function, which can be intractable. Moreover, we show that such a selection criterion can be inefficient. This paper aims to improve $A^*$ sampling by addressing these issues. To design a suitable selection rule, we apply \emph{Probability Matching}, a widely used method for decision making, to $A^*$ sampling. We provide insights into the relationship between $A^*$ sampling and probability matching by analyzing a nontrivial special case in which the state space is partitioned into two subsets. We show that in this case probability matching is optimal within a constant gap. Furthermore, as directly applying probability matching to $A^*$ sampling is time consuming, we design an approximate version based on Monte-Carlo estimators. We also present an efficient implementation by leveraging special properties of Gumbel distributions and well-designed balanced trees. Empirical results show that our method saves a significantly amount of computational resources on suboptimal regions compared with $A^*$ sampling. | Reject |
ICLR.cc/2020/Conference | ODE Analysis of Stochastic Gradient Methods with Optimism and Anchoring for Minimax Problems and GANs | Despite remarkable empirical success, the training dynamics of generative adversarial networks (GAN), which involves solving a minimax game using stochastic gradients, is still poorly understood. In this work, we analyze last-iterate convergence of simultaneous gradient descent (simGD) and its variants under the assumption of convex-concavity, guided by a continuous-time analysis with differential equations. First, we show that simGD, as is, converges with stochastic sub-gradients under strict convexity in the primal variable. Second, we generalize optimistic simGD to accommodate an optimism rate separate from the learning rate and show its convergence with full gradients. Finally, we present anchored simGD, a new method, and show convergence with stochastic subgradients. | Reject |
ICLR.cc/2023/Conference | Approximated Anomalous Diffusion: Gaussian Mixture Score-based Generative Models | Score-based generative models (SGMs) can generate high-quality samples via Langevin dynamics with a drift term and a diffusion term (Gaussian noise) iteratively calculated and added to a sample until convergence. In biological systems, it is observed that the neural population can conduct heavy-tailed L\'{e}vy dynamics for sampling-based probabilistic representation through neural fluctuations. Critically, unlike the existing sampling process of SGMs, L\'{e}vy dynamics can produce both large jumps and small roaming to explore the sampling space, resulting in better sampling results than Langevin dynamics with a lacking of large jumps. Motivated by this contrast, we explore a new class of SGMs with the sampling based on the L\'{e}vy dynamics. However, exact numerical simulation of the L\'{e}vy dynamics is significantly more challenging and intractable. We hence propose an approximation solution by leveraging Gaussian mixture noises during training to achieve the desired large jumps and small roaming properties. Theoretically, GM-SGMs conduct a probabilistic graphical model used by empirical Bayes for sampling, expanding the maximum a posteriori (MAP) estimation applied by conventional SGMs. Expensive experiments on the challenging image generation tasks show that our GM-SGMs exhibit superior sampling quality over prior art SGMs across various sampling iterations. | Reject |
ICLR.cc/2023/Conference | Linear convergence for natural policy gradient with log-linear policy parametrization | We analyze the convergence rate of the \emph{unregularized} natural policy gradient algorithm with log-linear policy parametrizations in infinite-horizon discounted Markov decision processes. In the deterministic case, when the Q-value is known and can be approximated by a linear combination of a known feature function up to a bias error, we show that a geometrically-increasing step size yields a linear convergence rate towards an optimal policy. We then consider the sample-based case, when the best representation of the Q-value function among linear combinations of a known feature function is known up to an estimation error. In this setting, we show that the algorithm enjoys the same linear guarantees as in the deterministic case up to an error term that depends on the estimation error, the bias error, and the condition number of the feature covariance matrix. Our results build upon the general framework of policy mirror descent and extend previous findings for the softmax tabular parametrization to the log-linear policy class. | Reject |
ICLR.cc/2022/Conference | An evaluation of quality and robustness of smoothed explanations | Explanation methods play a crucial role in helping to understand the decisions of deep neural networks (DNNs) to develop trust that is critical for the adoption of predictive models. However, explanation methods are easily manipulated through visually imperceptible perturbations that generate misleading explanations. The geometry of the decision surface of the DNNs has been identified as the main cause of this phenomenon and several \emph{smoothing} approaches have been proposed to build more robust explanations.
In this work, we provide a thorough evaluation of the quality and robustness of the explanations derived by smoothing approaches. Their different properties are evaluated with extensive experiments, which reveal the settings where the smoothed explanations are better, and also worse than the explanations derived by the common Gradient method. By making the connection with the literature on adversarial attacks, we further show that such smoothed explanations are robust primarily against additive $\ell_p$-norm attacks. However, a combination of additive and non-additive attacks can still manipulate these explanations, which reveals shortcomings in their robustness properties. | Reject |
ICLR.cc/2018/Conference | Towards Unsupervised Classification with Deep Generative Models | Deep generative models have advanced the state-of-the-art in semi-supervised classification, however their capacity for deriving useful discriminative features in a completely unsupervised fashion for classification in difficult real-world data sets, where adequate manifold separation is required has not been adequately explored. Most methods rely on defining a pipeline of deriving features via generative modeling and then applying clustering algorithms, separating the modeling and discriminative processes. We propose a deep hierarchical generative model which uses a mixture of discrete and continuous distributions to learn to effectively separate the different data manifolds and is trainable end-to-end. We show that by specifying the form of the discrete variable distribution we are imposing a specific structure on the model's latent representations. We test our model's discriminative performance on the task of CLL diagnosis against baselines from the field of computational FC, as well as the Variational Autoencoder literature. | Reject |
ICLR.cc/2021/Conference | Graph-Based Neural Network Models with Multiple Self-Supervised Auxiliary Tasks | Self-supervised learning is currently gaining a lot of attention, as it allows neural networks to learn robust representations from large quantities of unlabeled data. Additionally, multi-task learning can further improve representation learning by training networks simultaneously on related tasks, leading to significant performance improvements. In this paper, we propose a general framework to improve graph-based neural network models by combining self-supervised auxiliary learning tasks in a multi-task fashion. Since Graph Convolutional Networks are among the most promising approaches for capturing relationships among structured data points, we use them as a building block to achieve competitive results on standard semi-supervised graph classification tasks. | Withdrawn |
ICLR.cc/2023/Conference | Adversarial Examples Guided Pseudo-label Refinement for Decentralized Domain Adaptation | Unsupervised domain adaptation (UDA) methods usually assume data from multiple domains can be put together for centralized adaptation. Unfortunately, this assumption impairs data privacy, which leads to the failure of traditional methods in practical scenarios. To cope with the above issue, we present a new approach named Adversarial Examples Guided Pseudo-label Refinement for Decentralized Domain Adaptation (AGREE), which conducts target adaptation in an iterative training process during which only models can be delivered across domains. More specifically, to train a promising target model, we leverage Adversarial Examples (AEs) to filter out error prone predictions of source models towards each target sample based on both robustness and confidence, and then treat the most frequent prediction as the pseudo-label. Besides, to improve central model aggregation, we introduce Knowledge Contribution (KC) to compute reasonable aggregation weights. Extensive experiments conducted on several standard datasets verify the superiority of the proposed AGREE. Especially, our AGREE achieves the new state-of-the-art performance on the DomainNet and Office-Caltech10 datasets. The implementation code will be publicly available. | Withdrawn |
ICLR.cc/2018/Conference | Discrete-Valued Neural Networks Using Variational Inference | The increasing demand for neural networks (NNs) being employed on embedded devices has led to plenty of research investigating methods for training low precision NNs. While most methods involve a quantization step, we propose a principled Bayesian approach where we first infer a distribution over a discrete weight space from which we subsequently derive hardware-friendly low precision NNs. To this end, we introduce a probabilistic forward pass to approximate the intractable variational objective that allows us to optimize over discrete-valued weight distributions for NNs with sign activation functions. In our experiments, we show that our model achieves state of the art performance on several real world data sets. In addition, the resulting models exhibit a substantial amount of sparsity that can be utilized to further reduce the computational costs for inference. | Reject |
ICLR.cc/2020/Conference | Neural Epitome Search for Architecture-Agnostic Network Compression | Traditional compression methods including network pruning, quantization, low rank factorization and knowledge distillation all assume that network architectures and parameters should be hardwired. In this work, we propose a new perspective on network compression, i.e., network parameters can be disentangled from the architectures. From this viewpoint, we present the Neural Epitome Search (NES), a new neural network compression approach that learns to find compact yet expressive epitomes for weight parameters of a specified network architecture end-to-end. The complete network to compress can be generated from the learned epitome via a novel transformation method that adaptively transforms the epitomes to match shapes of the given architecture. Compared with existing compression methods, NES allows the weight tensors to be independent of the architecture design and hence can achieve a good trade-off between model compression rate and performance given a specific model size constraint. Experiments demonstrate that, on ImageNet, when taking MobileNetV2 as backbone, our approach improves the full-model baseline by 1.47% in top-1 accuracy with 25% MAdd reduction and AutoML for Model Compression (AMC) by 2.5% with nearly the same compression ratio. Moreover, taking EfficientNet-B0 as baseline, our NES yields an improvement of 1.2% but are with 10% less MAdd. In particular, our method achieves a new state-of-the-art results of 77.5% under mobile settings (<350M MAdd). Code will be made publicly available. | Accept (Poster) |
ICLR.cc/2023/Conference | Robust Quantity-Aware Aggregation for Federated Learning | Federated learning (FL) enables multiple clients to collaboratively train models without sharing their local data, and becomes an important privacy-preserving machine learning framework. However, classical FL faces serious security and robustness problem, e.g., malicious clients can poison model updates and at the same time claim large quantities to amplify the impact of their model updates in the model aggregation. Existing defense methods for FL, while all handling malicious model updates, either treat all quantities benign or simply ignore/truncate the quantities of all clients. The former is vulnerable to quantity-enhanced attack, while the latter leads to sub-optimal performance since the local data on different clients is usually in significantly different sizes. In this paper, we propose a robust quantity-aware aggregation algorithm for federated learning, called FedRA, to perform the aggregation with awareness of local data quantities while being able to defend against quantity-enhanced attacks. More specifically, we propose a method to filter malicious clients by jointly considering the uploaded model updates and data quantities from different clients, and performing quantity-aware weighted averaging on model updates from remaining clients. Moreover, as the number of malicious clients participating in the federated learning may dynamically change in different rounds, we also propose a malicious client number estimator to predict how many suspicious clients should be filtered in each round. Experiments on four public datasets demonstrate the effectiveness of our FedRA method in defending FL against quantity-enhanced attacks. Our code is available at \url{https://anonymous.4open.science/r/FedRA-4C1E}.
| Withdrawn |
ICLR.cc/2023/Conference | Adversarial Policies Beat Professional-Level Go AIs | We attack the state-of-the-art Go-playing AI system, KataGo, by training an adversarial policy that plays against a frozen KataGo victim. Our attack achieves a >99% win-rate against KataGo without search, and a >80% win-rate when KataGo uses enough search to be near-superhuman. To the best of our knowledge, this is the first successful end-to-end attack against a Go AI playing at the level of a top human professional. Notably, the adversary does not win by learning to play Go better than KataGo---in fact, the adversary is easily beaten by human amateurs. Instead, the adversary wins by tricking KataGo into ending the game prematurely at a point that is favorable to the adversary. Our results demonstrate that even professional-level AI systems may harbor surprising failure modes. | Reject |
ICLR.cc/2023/Conference | Understanding Curriculum Learning in Policy Optimization for Online Combinatorial Optimization | Over the recent years, reinforcement learning (RL) starts to show promising results in tackling combinatorial optimization (CO) problems, in particular when coupled with curriculum learning to facilitate training. Despite emerging empirical evidence, theoretical study on why RL helps is still at its early stage. This paper presents the first systematic study on policy optimization methods for online CO problems. We show that online CO problems can be naturally formulated as latent Markov Decision Processes (LMDPs), and prove convergence bounds on natural policy gradient (NPG) for solving LMDPs. Furthermore, our theory explains the benefit of curriculum learning: it can find a strong sampling policy and reduce the distribution shift, a critical quantity that governs the convergence rate in our theorem. For a canonical online CO problem, Secretary Problem, we formally prove that distribution shift is reduced exponentially with curriculum learning even if the curriculum is randomly generated. Our theory also shows we can simplify the curriculum learning scheme used in prior work from multi-step to single-step. Lastly, we provide extensive experiments on Secretary Problem and Online Knapsack to verify our findings. | Reject |
ICLR.cc/2021/Conference | Compressing gradients in distributed SGD by exploiting their temporal correlation | We propose SignXOR, a novel compression scheme that exploits temporal correlation of gradients for the purpose of gradient compression. Sign-based schemes such as Scaled-sign and SignSGD (Bernstein et al., 2018; Karimireddy et al., 2019) compress gradients by storing only the sign of gradient entries. These methods, however, ignore temporal correlations between gradients. The equality or non-equality of signs of gradients in two consecutive iterations can be represented by a binary vector, which can be further compressed depending on its entropy. By implementing a rate-distortion encoder we increase the temporal correlation of gradients, lowering entropy and improving compression. We achieve theoretical convergence of SignXOR by employing the two-way error-feedback approach introduced by Zheng et al. (2019). Zheng et al. (2019) show that two-way compression with error-feedback achieves the same asymptotic convergence rate as SGD, although convergence is slower by a constant factor. We strengthen their analysis to show that the rate of convergence of two-way compression with errorfeedback asymptotically is the same as that of SGD. As a corollary we prove that two-way SignXOR compression with error-feedback achieves the same asymptotic rate of convergence as SGD. We numerically evaluate our proposed method on the CIFAR-100 and ImageNet datasets and show that SignXOR requires less than 50% of communication traffic compared to sending sign of gradients. To the best of our knowledge we are the first to present a gradient compression scheme that exploits temporal correlation of gradients. | Reject |
ICLR.cc/2020/Conference | AdvCodec: Towards A Unified Framework for Adversarial Text Generation | Machine learning (ML) especially deep neural networks (DNNs) have been widely applied to real-world applications. However, recent studies show that DNNs are vulnerable to carefully crafted \emph{adversarial examples} which only deviate from the original data by a small magnitude of perturbation.
While there has been great interest on generating imperceptible adversarial examples in continuous data domain (e.g. image and audio) to explore the model vulnerabilities, generating \emph{adversarial text} in the discrete domain is still challenging.
The main contribution of this paper is to propose a general targeted attack framework \advcodec for adversarial text generation which addresses the challenge of discrete input space and be easily adapted to general natural language processing (NLP) tasks.
In particular, we propose a tree based autoencoder to encode discrete text data into continuous vector space, upon which we optimize the adversarial perturbation. With the tree based decoder, it is possible to ensure the grammar correctness of the generated text; and the tree based encoder enables flexibility of making manipulations on different levels of text, such as sentence (\advcodecsent) and word (\advcodecword) levels. We consider multiple attacking scenarios, including appending an adversarial sentence or adding unnoticeable words to a given paragraph, to achieve arbitrary \emph{targeted attack}. To demonstrate the effectiveness of the proposed method, we consider two most representative NLP tasks: sentiment analysis and question answering (QA). Extensive experimental results show that \advcodec has successfully attacked both tasks. In particular, our attack causes a BERT-based sentiment classifier accuracy to drop from $0.703$ to $0.006$, and a BERT-based QA model's F1 score to drop from $88.62$ to $33.21$ (with best targeted attack F1 score as $46.54$). Furthermore, we show that the white-box generated adversarial texts can transfer across other black-box models, shedding light on an effective way to examine the robustness of existing NLP models. | Reject |
ICLR.cc/2023/Conference | Boosting the Cycle Counting Power of Graph Neural Networks with I$^2$-GNNs | Message Passing Neural Networks (MPNNs) are a widely used class of Graph Neural Networks (GNNs). The limited representational power of MPNNs inspires the study of provably powerful GNN architectures. However, knowing one model is more powerful than another gives little insight about what functions they can or cannot express. It is still unclear whether these models are able to approximate specific functions such as counting certain graph substructures, which is essential for applications in biology, chemistry and social network analysis. Motivated by this, we propose to study the counting power of Subgraph MPNNs, a recent and popular class of powerful GNN models that extract rooted subgraphs for each node, assign the root node a unique identifier and encode the root node's representation within its rooted subgraph. Specifically, we prove that Subgraph MPNNs fail to count more-than-4-cycles at node level, implying that node representations cannot correctly encode the surrounding substructures like ring systems with more than four atoms. To overcome this limitation, we propose I$^2$-GNNs to extend Subgraph MPNNs by assigning different identifiers for the root node and its neighbors in each subgraph. I$^2$-GNNs' discriminative power is shown to be strictly stronger than Subgraph MPNNs and partially stronger than the 3-WL test. More importantly, I$^2$-GNNs are proven capable of counting all 3, 4, 5 and 6-cycles, covering common substructures like benzene rings in organic chemistry, while still keeping linear complexity. To the best of our knowledge, it is the first linear-time GNN model that can count 6-cycles with theoretical guarantees. We validate its counting power in cycle counting tasks and demonstrate its competitive performance in molecular prediction benchmarks. | Accept: poster |
ICLR.cc/2023/Conference | Attention-Guided Backdoor Attacks against Transformers | With the popularity of transformers in natural language processing (NLP) applications, there are growing concerns about their security. Most existing NLP attack methods focus on injecting stealthy trigger words/phrases. In this paper, we focus on the interior structure of neural networks and the Trojan mechanism. Focusing on the prominent NLP transformer models, we propose a novel Trojan Attention Loss (TAL), which enhances the Trojan behavior by directly manipulating the attention pattern. Our loss significantly improves the attack efficacy; it achieves better successful rates and with a much smaller poisoning rate (i.e., a smaller proportion of poisoned samples). It boosts attack efficacy for not only traditional dirty-label attacks, but also the more challenging clean-label attacks. TAL is also highly compatible with most existing attack methods and its flexibility enables this loss easily adapted to other backbone transformer models. | Reject |
ICLR.cc/2021/Conference | Interactive Visualization for Debugging RL | Visualization tools for supervised learning (SL) allow users to interpret, introspect, and gain an intuition for the successes and failures of their models. While reinforcement learning (RL) practitioners ask many of the same questions while debugging agent policies, existing tools aren't a great fit for the RL setting as these tools address challenges typically found in the SL regime. Whereas SL involves a static dataset, RL often entails collecting new data in challenging environments with partial observability, stochasticity, and non-stationary data distributions. This necessitates the creation of alternate visual interfaces to help us better understand agent policies trained using RL. In this work, we design and implement an interactive visualization tool for debugging and interpreting RL. Our system identifies and addresses important aspects missing from existing tools such as (1) visualizing alternate state representations (different from those seen by the agent) that researchers could use while debugging RL policies; (2) interactive interfaces tailored to metadata stored while training RL agents (3) a conducive workflow designed around RL policy debugging. We provide an example workflow of how this system could be used, along with ideas for future extensions. | Reject |
ICLR.cc/2023/Conference | SmilesFormer: Language Model for Molecular Design | The objective of drug discovery is to find novel compounds with desirable chemical properties. Generative models have been utilized to sample molecules at the intersection of multiple property constraints. In this paper we pose molecular design as a language modeling problem where the model implicitly learns the vocabulary and composition of valid molecules, hence it is able to generate new molecules of interest. We present SmilesFormer, a Transformer-based model which is able to encode molecules, molecule fragments, and fragment compositions as latent variables, which are in turn decoded to stochastically generate novel molecules. This is achieved by fragmenting the molecules into smaller combinatorial groups, then learning the mapping between the input fragments and valid SMILES sequences.
The model is able to optimize molecular properties through a stochastic latent space traversal technique. This technique systematically searches the encoded latent space to find latent vectors that are able to produce molecules to meet the multi-property objective. The model was validated through various de novo molecular design tasks, achieving state-of-the-art performances when compared to previous methods. Furthermore, we used the proposed method to demonstrate a drug rediscovery pipeline for Donepezil, a known Acetylcholinesterase Inhibitor. | Reject |
ICLR.cc/2020/Conference | MissDeepCausal: causal inference from incomplete data using deep latent variable models | Inferring causal effects of a treatment, intervention or policy from observational data is central to many applications. However, state-of-the-art methods for causal inference seldom consider the possibility that covariates have missing values, which is ubiquitous in many real-world analyses. Missing data greatly complicate causal inference procedures as they require an adapted unconfoundedness hypothesis which can be difficult to justify in practice. We circumvent this issue by considering latent confounders whose distribution is learned through variational autoencoders adapted to missing values. They can be used either as a pre-processing step prior to causal inference but we also suggest to embed them in a multiple imputation strategy to take into account the variability due to missing values. Numerical experiments demonstrate the effectiveness of the proposed methodology especially for non-linear models compared to competitors. | Reject |
ICLR.cc/2023/Conference | Gradient-based Algorithms for Pessimistic Bilevel Optimization | As a powerful framework for a variety of machine learning problems, bilevel optimization has attracted much attention. While many modern gradient-based algorithms have been devised for optimistic bilevel optimization, pessimistic bilevel optimization (PBO) is still less-explored and only studied under the linear settings. To fill this void, we investigate PBO with nonlinear inner- and outer-level objective functions in this work, by reformulating it into a single-level constrained optimization problem. In particular, two gradient-based algorithms are first proposed to solve the reformulated problem, i.e., the switching gradient method (SG-PBO) and the primal-dual method (PD-PBO). Through carefully handling the bias errors in gradient estimations resulted by the nature of bilevel optimization, we show that both SG-PBO and PD-PBO converge to the global minimum of the reformulated problem when it is strongly convex, which immediately implies the convergence to the original PBO. Moreover, we propose the proximal scheme (Prox-PBO) with the convergence guarantee for the nonconvex reformulated problem. To the best of our knowledge, this is the first work that investigates gradient-based algorithms and provides convergence analysis for PBO under non-linear settings. We further conduct experiments on an illustrative example and a robust hyperparameter learning problem, which clearly validate our algorithmic design and theoretical analysis. | Withdrawn |
ICLR.cc/2021/Conference | Robust Learning via Golden Symmetric Loss of (un)Trusted Labels | Learning robust deep models against noisy labels becomes ever critical when today's data is commonly collected from open platforms and subject to adversarial corruption. The information on the label corruption process, i.e., corruption matrix, can greatly enhance the robustness of deep models but still fall behind in combating hard classes. In this paper, we propose to construct a golden symmetric loss (GSL) based on the estimated confusion matrix as to avoid overfitting to noisy labels and learn effectively from hard classes. GSL is the weighted sum of the corrected regular cross entropy and reverse cross entropy. By leveraging a small fraction of trusted clean data, we estimate the corruption matrix and use it to correct the loss as well as to determine the weights of GSL. We theoretically prove the robustness of the proposed loss function in the presence of dirty labels. We provide a heuristics to adaptively tune the loss weights of GSL according to the noise rate and diversity measured from the dataset. We evaluate our proposed golden symmetric loss on both vision and natural language deep models subject to different types of label noise patterns. Empirical results show that GSL can significantly outperform the existing robust training methods on different noise patterns, showing accuracy improvement up to 18% on CIFAR-100 and 1% on real world noisy dataset of Clothing1M. | Reject |
ICLR.cc/2023/Conference | Meta-Learning via Classifier(-free) Guidance | State-of-the-art meta-learning techniques do not optimize for zero-shot adaptation to unseen tasks, a setting in which humans excel. On the contrary, meta-learning algorithms learn hyperparameters and weight initializations that explicitly optimize for few-shot learning performance. In this work, we take inspiration from recent advances in generative modeling and language-conditioned image synthesis to propose meta-learning techniques that use natural language guidance to achieve higher zero-shot performance compared to the state-of-the-art. We do so by recasting the meta-learning problem as a multi-modal generative modeling problem: given a task, we consider its adapted neural network weights and its natural language description as equivalent multi-modal task representations. We first train an unconditional generative hypernetwork model to produce neural network weights; then we train a second "guidance" model that, given a natural language task description, traverses the hypernetwork latent space to find high-performance task-adapted weights in a zero-shot manner. We explore two alternative approaches for latent space guidance: "HyperCLIP"-based classifier guidance and a conditional Hypernetwork Latent Diffusion Model ("HyperLDM"), which we show to benefit from the classifier-free guidance technique common in image generation. Finally, we demonstrate that our approaches outperform existing meta-learning methods with zero-shot learning experiments on our Meta-VQA dataset, which we specifically constructed to reflect the multi-modal meta-learning setting. | Reject |
ICLR.cc/2022/Conference | Non-reversible Parallel Tempering for Uncertainty Approximation in Deep Learning | Parallel tempering (PT), also known as replica exchange, is the go-to workhorse for simulations of multi-modal distributions. The key to the success of PT is to adopt efficient swap schemes. The popular deterministic even-odd (DEO) scheme exploits the non-reversibility property and has successfully reduced the communication cost from $O(P^2)$ to $O(P)$ given sufficient many $P$ chains. However, such an innovation largely disappears given limited chains in big data problems due to the extremely few bias-corrected swaps. To handle this issue, we generalize the DEO scheme to promote the non-reversibility and obtain an optimal communication cost $O(P\log P)$. In addition, we also analyze the bias when we adopt stochastic gradient descent (SGD) with large and constant learning rates as exploration kernels. Such a user-friendly nature enables us to conduct large-scale uncertainty approximation tasks without much tuning costs. | Reject |
ICLR.cc/2020/Conference | A Simple and Scalable Shape Representation for 3D Reconstruction | Deep learning applied to the reconstruction of 3D shapes has seen growing interest. A popular approach to 3D reconstruction and generation in recent years has been the CNN decoder-encoder model often applied in voxel space. However this often scales very poorly with the resolution limiting the effectiveness of these models. Several sophisticated alternatives for decoding to 3D shapes have been proposed typically relying on alternative deep learning architectures. We show however in this work that standard benchmarks in 3D reconstruction can be tackled with a surprisingly simple approach: a linear decoder obtained by principal component analysis on the signed distance transform of the surface. This approach allows easily scaling to larger resolutions. We show in multiple experiments it is competitive with state of the art methods and also allows the decoder to be fine-tuned on the target task using a loss designed for SDF transforms, obtaining further gains. | Reject |
ICLR.cc/2020/Conference | FRICATIVE PHONEME DETECTION WITH ZERO DELAY | People with high-frequency hearing loss rely on hearing aids that employ frequency lowering algorithms. These algorithms shift some of the sounds from the high frequency band to the lower frequency band where the sounds become more perceptible for the people with the condition. Fricative phonemes have an important part of their content concentrated in high frequency bands. It is important that the frequency lowering algorithm is activated exactly for the duration of a fricative phoneme, and kept off at all other times. Therefore, timely (with zero delay) and accurate fricative phoneme detection is a key problem for high quality hearing aids. In this paper we present a deep learning based fricative phoneme detection algorithm that has zero detection delay and achieves state-of-the-art fricative phoneme detection accuracy on the TIMIT Speech Corpus. All reported results are reproducible and come with easy to use code that could serve as a baseline for future research.
| Reject |
ICLR.cc/2023/Conference | Dual personalization for federated recommendation on devices | Federated recommendation is a new Internet service architecture that aims to provide privacy-preserving recommendation services in federated settings. Existing solutions are used to combine distributed recommendation algorithms and privacy-preserving mechanisms. Thus it inherently takes the form of heavyweight models at the server and hinders the deployment of on-device intelligent models to end-users. This paper proposes a novel Personalized Federated Recommendation (PFedRec) framework to learn many user-specific lightweight models to be deployed on smart devices rather than a heavyweight model on a server. Moreover, we propose a new dual personalization mechanism to effectively learn fine-grained personalization on both users and items. The overall learning process is formulated into a unified federated optimization framework. Specifically, unlike previous methods that share exactly the same item embeddings across users in a federated system, dual personalization allows mild finetuning of item embeddings for each user to generate user-specific views for item representations which can be integrated into existing federated recommendation methods to gain improvements immediately. Experiments on multiple benchmark datasets have demonstrated the effectiveness of PFedRec and the dual personalization mechanism. Moreover, we provide visualizations and in-depth analysis of the personalization techniques in item embedding, which shed novel insights on the design of RecSys in federated settings. | Reject |
ICLR.cc/2023/Conference | Generative Adversarial Training for Neural Combinatorial Optimization Models | Recent studies show that deep neural networks can be trained to learn good heuristics for various Combinatorial Optimization Problems (COPs). However, it remains a great challenge for the trained deep optimization models to generalize to distributions different from the training one. To address this issue, we propose a general framework, Generative Adversarial Neural Combinatorial Optimization (GANCO) which is equipped with another deep model to generate training instances for the optimization model, so as to improve its generalization ability. The two models are trained alternatively in an adversarial way, where the generation model is trained by reinforcement learning to find instance distributions hard for the optimization model. We apply the GANCO framework to two recent deep combinatorial optimization models, i.e., AM and POMO. Extensive experiments on various COPs such as Traveling Salesman Problem, Capacitated Vehicle Routing Problem, and 0-1 Knapsack Problem show that GANCO can significantly improve the generalization ability of optimization models on various instance distributions. | Withdrawn |
ICLR.cc/2021/Conference | Efficient Graph Neural Architecture Search | Recently, graph neural networks (GNN) have been demonstrated effective in various graph-based tasks.
To obtain state-of-the-art (SOTA) data-specific GNN architectures, researchers turn to the neural architecture search (NAS) methods.
However, it remains to be a challenging problem to conduct efficient architecture search for GNN.
In this work, we present a novel framework for Efficient GrAph Neural architecture search (EGAN).
By designing a novel and expressive search space, an efficient one-shot NAS method based on stochastic relaxation and natural gradient is proposed.
Further, to enable architecture search in large graphs, a transfer learning paradigm is designed.
Extensive experiments, including node-level and graph-level tasks, are conducted. The results show that the proposed EGAN can obtain SOTA data-specific architectures, and reduce the search cost by two orders of magnitude compared to existing NAS baselines. | Reject |
ICLR.cc/2022/Conference | UniNet: Unified Architecture Search with Convolution, Transformer, and MLP | Recently, transformer and multi-layer perceptron (MLP) architectures have achieved impressive results on various vision tasks. A few works investigated manually combining those operators to design visual network architectures, and can achieve satisfactory performances to some extent. In this paper, we propose to jointly search the optimal combination of convolution, transformer, and MLP for building a series of all-operator network architectures with high performances on visual tasks. We empirically identify that the widely-used strided convolution or pooling based down-sampling modules become the performance bottlenecks when the operators are combined to form a network. To better tackle the global context captured by the transformer and MLP operators, we propose two novel context-aware down-sampling modules, which can better adapt to the global information encoded by transformer and MLP operators. To this end, we jointly search all operators and down-sampling modules in a unified search space. Notably, Our searched network UniNet (Unified Network) outperforms state-of-the-art pure convolution-based architecture, EfficientNet, and pure transformer-based architecture, Swin-Transformer, on multiple public visual benchmarks, ImageNet classification, COCO object detection, and ADE20K semantic segmentation. | Withdrawn |
ICLR.cc/2021/Conference | Undistillable: Making A Nasty Teacher That CANNOT teach students | Knowledge Distillation (KD) is a widely used technique to transfer knowledge from pre-trained teacher models to (usually more lightweight) student models. However, in certain situations, this technique is more of a curse than a blessing. For instance, KD poses a potential risk of exposing intellectual properties (IPs): even if a trained machine learning model is released in ``black boxes'' (e.g., as executable software or APIs without open-sourcing code), it can still be replicated by KD through imitating input-output behaviors. To prevent this unwanted effect of KD, this paper introduces and investigates a concept called $\textit{Nasty Teacher}$: a specially trained teacher network that yields nearly the same performance as a normal one, but would significantly degrade the performance of student models learned by imitating it. We propose a simple yet effective algorithm to build the nasty teacher, called $\textit{self-undermining knowledge distillation}$. Specifically, we aim to maximize the difference between the output of the nasty teacher and a normal pre-trained network. Extensive experiments on several datasets demonstrate that our method is effective on both standard KD and data-free KD, providing the desirable KD-immunity to model owners for the first time. We hope our preliminary study can draw more awareness and interest in this new practical problem of both social and legal importance. Our codes and pre-trained models can be found at: $\url{https://github.com/VITA-Group/Nasty-Teacher}$. | Accept (Spotlight) |
ICLR.cc/2023/Conference | 3D UX-Net: A Large Kernel Volumetric ConvNet Modernizing Hierarchical Transformer for Medical Image Segmentation | The recent 3D medical ViTs (e.g., SwinUNETR) achieve the state-of-the-art performances on several 3D volumetric data benchmarks, including 3D medical image segmentation. Hierarchical transformers (e.g., Swin Transformers) reintroduced several ConvNet priors and further enhanced the practical viability of adapting volumetric segmentation in 3D medical datasets. The effectiveness of hybrid approaches is largely credited to the large receptive field for non-local self-attention and the large number of model parameters. We hypothesize that volumetric ConvNets can simulate the large receptive field behavior of these learning approaches with fewer model parameters using depth-wise convolution. In this work, we propose a lightweight volumetric ConvNet, termed 3D UX-Net, which adapts the hierarchical transformer using ConvNet modules for robust volumetric segmentation. Specifically, we revisit volumetric depth-wise convolutions with large kernel (LK) size (e.g. starting from $7\times7\times7$) to enable the larger global receptive fields, inspired by Swin Transformer. We further substitute the multi-layer perceptron (MLP) in Swin Transformer blocks with pointwise depth convolutions and enhance model performances with fewer normalization and activation layers, thus reducing the number of model parameters. 3D UX-Net competes favorably with current SOTA transformers (e.g. SwinUNETR) using three challenging public datasets on volumetric brain and abdominal imaging: 1) MICCAI Challenge 2021 FLARE, 2) MICCAI Challenge 2021 FeTA, and 3) MICCAI Challenge 2022 AMOS. 3D UX-Net consistently outperforms SwinUNETR with improvement from 0.929 to 0.938 Dice (FLARE2021) and 0.867 to 0.874 Dice (Feta2021). We further evaluate the transfer learning capability of 3D UX-Net with AMOS2022 and demonstrates another improvement of $2.27\%$ Dice (from 0.880 to 0.900). The source code with our proposed model are available at https://github.com/MASILab/3DUX-Net. | Accept: poster |
ICLR.cc/2019/Conference | ACCELERATING NONCONVEX LEARNING VIA REPLICA EXCHANGE LANGEVIN DIFFUSION | Langevin diffusion is a powerful method for nonconvex optimization, which enables the escape from local minima by injecting noise into the gradient. In particular, the temperature parameter controlling the noise level gives rise to a tradeoff between ``global exploration'' and ``local exploitation'', which correspond to high and low temperatures. To attain the advantages of both regimes, we propose to use replica exchange, which swaps between two Langevin diffusions with different temperatures. We theoretically analyze the acceleration effect of replica exchange from two perspectives: (i) the convergence in $\chi^2$-divergence, and (ii) the large deviation principle. Such an acceleration effect allows us to faster approach the global minima. Furthermore, by discretizing the replica exchange Langevin diffusion, we obtain a discrete-time algorithm. For such an algorithm, we quantify its discretization error in theory and demonstrate its acceleration effect in practice. | Accept (Poster) |
ICLR.cc/2020/Conference | Generalization through Memorization: Nearest Neighbor Language Models | We introduce $k$NN-LMs, which extend a pre-trained neural language model (LM) by linearly interpolating it with a $k$-nearest neighbors ($k$NN) model. The nearest neighbors are computed according to distance in the pre-trained LM embedding space, and can be drawn from any text collection, including the original LM training data. Applying this transformation to a strong Wikitext-103 LM, with neighbors drawn from the original training set, our $k$NN-LM achieves a new state-of-the-art perplexity of 15.79 -- a 2.9 point improvement with no additional training. We also show that this approach has implications for efficiently scaling up to larger training sets and allows for effective domain adaptation, by simply varying the nearest neighbor datastore, again without further training. Qualitatively, the model is particularly helpful in predicting rare patterns, such as factual knowledge. Together, these results strongly suggest that learning similarity between sequences of text is easier than predicting the next word, and that nearest neighbor search is an effective approach for language modeling in the long tail. | Accept (Poster) |
ICLR.cc/2021/Conference | On the role of planning in model-based deep reinforcement learning | Model-based planning is often thought to be necessary for deep, careful reasoning and generalization in artificial agents. While recent successes of model-based reinforcement learning (MBRL) with deep function approximation have strengthened this hypothesis, the resulting diversity of model-based methods has also made it difficult to track which components drive success and why. In this paper, we seek to disentangle the contributions of recent methods by focusing on three questions: (1) How does planning benefit MBRL agents? (2) Within planning, what choices drive performance? (3) To what extent does planning improve generalization? To answer these questions, we study the performance of MuZero (Schrittwieser et al., 2019), a state-of-the-art MBRL algorithm with strong connections and overlapping components with many other MBRL algorithms. We perform a number of interventions and ablations of MuZero across a wide range of environments, including control tasks, Atari, and 9x9 Go. Our results suggest the following: (1) Planning is most useful in the learning process, both for policy updates and for providing a more useful data distribution. (2) Using shallow trees with simple Monte-Carlo rollouts is as performant as more complex methods, except in the most difficult reasoning tasks. (3) Planning alone is insufficient to drive strong generalization. These results indicate where and how to utilize planning in reinforcement learning settings, and highlight a number of open questions for future MBRL research. | Accept (Poster) |
ICLR.cc/2023/Conference | Plateau in Monotonic Linear Interpolation --- A "Biased" View of Loss Landscape for Deep Networks | Monotonic linear interpolation (MLI) --- on the line connecting a random initialization with the minimizer it converges to, the loss and accuracy are monotonic --- is a phenomenon that is commonly observed in the training of neural networks. Such a phenomenon may seem to suggest that optimization of neural networks is easy. In this paper, we show that the MLI property is not necessarily related to the hardness of optimization problems, and empirical observations on MLI for deep neural networks depend heavily on the biases. In particular, we show that interpolating both weights and biases linearly leads to very different influences on the final output, and when different classes have different last-layer biases on a deep network, there will be a long plateau in both the loss and accuracy interpolation (which existing theory of MLI cannot explain). We also show how the last-layer biases for different classes can be different even on a perfectly balanced dataset using a simple model. Empirically we demonstrate that similar intuitions hold on practical networks and realistic datasets. | Accept: poster |
ICLR.cc/2022/Conference | Reinforcement Learning under a Multi-agent Predictive State Representation Model: Method and Theory | We study reinforcement learning for partially observable multi-agent systems where each agent only has access to its own observation and reward and aims to maximize its cumulative rewards. To handle partial observations, we propose graph-assisted predictive state representations (GAPSR), a scalable multi-agent representation learning framework that leverages the agent connectivity graphs to aggregate local representations computed by each agent. In addition, our representations are readily able to incorporate dynamic interaction graphs and kernel space embeddings of the predictive states, and thus have strong flexibility and representation power.
Based on GAPSR, we propose an end-to-end MARL algorithm that simultaneously infers the predictive representations and uses the representations as the input of a policy optimization algorithm. Empirically, we demonstrate the efficacy of the proposed algorithm provided on both a MAMuJoCo robotic learning experiment and a multi-agent particle learning environment. | Accept (Spotlight) |
ICLR.cc/2020/Conference | Rethinking Generalized Matrix Factorization for Recommendation: The Importance of Multi-hot Encoding | Learning good representations of users and items is crucially important to recommendation with implicit feedback. Matrix factorization is the basic idea to derive the representations of users and items by decomposing the given interaction matrix. However, existing matrix factorization based approaches share the limitation in that the interaction between user embedding and item embedding is only weakly enforced by fitting the given individual rating value, which may lose potentially useful information. In this paper, we propose a novel Augmented Generalized Matrix Factorization (AGMF) approach that is able to incorporate the historical interaction information of users and items for learning effective representations of users and items. Despite the simplicity of our proposed approach, extensive experiments on four public implicit feedback datasets demonstrate that our approach outperforms state-of-the-art counterparts. Furthermore, the ablation study demonstrates that by using multi-hot encoding to enrich user embedding and item embedding for Generalized Matrix Factorization, better performance, faster convergence, and lower training loss can be achieved. | Withdrawn |
ICLR.cc/2021/Conference | Semi-supervised regression with skewed data via adversarially forcing the distribution of predicted values | Advances in scientific fields including drug discovery or material design are accompanied by numerous trials and errors. However, generally only representative experimental results are reported. Because of this reporting bias, the distribution of labeled result data can deviate from their true distribution. A regression model can be erroneous if it is built on these skewed data. In this work, we propose a new approach to improve the accuracy of regression models that are trained using a skewed dataset. The method forces the regression outputs to follow the true distribution; the forcing algorithm regularizes the regression results while keeping the information of the training data. We assume the existence of enough unlabeled data that follow the true distribution, and that the true distribution can be roughly estimated from domain knowledge or a few samples. During training neural networks to generate a regression model, an adversarial network is used to force the distribution of predicted values to follow the estimated ‘true’ distribution. We evaluated the proposed approach on four real-world datasets (pLogP, Diamond, House, Elevators). In all four datasets, the proposed approach reduced the root mean squared error of the regression by around 55 percent to 75 percent compared to regression models without adjustment of the distribution. | Reject |
ICLR.cc/2023/Conference | Critical Learning Periods Augmented Model Poisoning Attacks to Byzantine-Robust Federated Learning | Existing attacks in federated learning (FL) control a set of malicious clients and share a fixed number of malicious gradients with the central server in each training round, to achieve a desired tradeoff between attack impact and resilience against defenses. In this paper, we show that such a tradeoff is not fundamental and an adaptive attack budget not only improves the impact of attack $\mathcal{A}$ but makes it more resilient to defenses. Inspired by recent findings on critical learning periods (CLP), where small gradient errors have irrecoverable impact on model accuracy, we advocate CLP augmented model poisoning attacks $\mathcal{A}$-CLP, which merely augment attack $\mathcal{A}$ with an adaptive attack budget scheme. $\mathcal{A}$-CLP inspects the changes in federated gradient norms to identify CLP and adaptively adjusts the number of malicious clients that share their malicious gradients with the central server in each round, leading to dramatically improved attack impact compared to $\mathcal{A}$ itself by up to 6.85$\times$, with a smaller attack budget and hence improved resilience of $\mathcal{A}$ by up to 2$\times$. Based on understandings on $\mathcal{A}$-CLP, we further relax the inner attack subroutine $\mathcal{A}$ in $\mathcal{A}$-CLP, and propose SimAttack-CLP, a lightweight CLP augmented similarity-based attack, which is more flexible and impactful. | Withdrawn |
ICLR.cc/2023/Conference | Adversarial Driving Policy Learning by Misunderstanding the Traffic Flow | Acquiring driving policies that can transfer to unseen environments is essential for driving in dense traffic flows. Adversarial training is a promising path to improve robustness under disturbances. Most prior works leverage few agents to induce driving policy's failures. However, we argue that directly implementing this training framework into dense traffic flow degrades transferability in unseen environments. In this paper, we propose a novel robust policy training framework that is capable of applying adversarial training based on a coordinated traffic flow. We start by building up a coordinated traffic flow where agents are allowed to communicate Social Value Orientations (SVOs). Adversary emerges when the traffic flow misunderstands the SVO of driving agent. We utilize this property to formulate a minimax optimization problem where the driving policy maximizes its own reward and a spurious adversarial policy minimizes it. Experiments demonstrate that our adversarial training framework significantly improves zero-shot transfer performance of the driving policy in dense traffic flows compared to existing algorithms. | Reject |
ICLR.cc/2021/Conference | Efficient Long-Range Convolutions for Point Clouds | The efficient treatment of long-range interactions for point clouds is a challenging problem in many scientific machine learning applications. To extract global information, one usually needs a large window size, a large number of layers, and/or a large number of channels. This can often significantly increase the computational cost. In this work, we present a novel neural network layer that directly incorporates long-range information for a point cloud. This layer, dubbed the long-range convolutional (LRC)-layer, leverages the convolutional theorem coupled with the non-uniform Fourier transform. In a nutshell, the LRC-layer mollifies the point cloud to an adequately sized regular grid, computes its Fourier transform, multiplies the result by a set of trainable Fourier multipliers, computes the inverse Fourier transform, and finally interpolates the result back to the point cloud. The resulting global all-to-all convolution operation can be performed in nearly-linear time asymptotically with respect to the number of input points. The LRC-layer is a particularly powerful tool when combined with local convolution as together they offer efficient and seamless treatment of both short and long range interactions. We showcase this framework by introducing a neural network architecture that combines LRC-layers with short-range convolutional layers to accurately learn the energy and force associated with a $N$-body potential. We also exploit the induced two-level decomposition and propose an efficient strategy to train the combined architecture with a reduced number of samples. | Reject |
ICLR.cc/2023/Conference | EUCLID: Towards Efficient Unsupervised Reinforcement Learning with Multi-choice Dynamics Model | Unsupervised reinforcement learning (URL) poses a promising paradigm to learn useful behaviors in a task-agnostic environment without the guidance of extrinsic rewards to facilitate the fast adaptation of various downstream tasks. Previous works focused on the pre-training in a model-free manner while lacking the study of transition dynamics modeling that leaves a large space for the improvement of sample efficiency in downstream tasks. To this end, we propose an Efficient Unsupervised Reinforcement Learning Framework with Multi-choice Dynamics model (EUCLID), which introduces a novel model-fused paradigm to jointly pre-train the dynamics model and unsupervised exploration policy in the pre-training phase, thus better leveraging the environmental samples and improving the downstream task sampling efficiency. However, constructing a generalizable model which captures the local dynamics under different behaviors remains a challenging problem. We introduce the multi-choice dynamics model that covers different local dynamics under different behaviors concurrently, which uses different heads to learn the state transition under different behaviors during unsupervised pre-training and selects the most appropriate head for prediction in the downstream task. Experimental results in the manipulation and locomotion domains demonstrate that EUCLID achieves state-of-the-art performance with high sample efficiency, basically solving the state-based URLB benchmark and reaching a mean normalized score of 104.0±1.2% in downstream tasks with 100k fine-tuning steps, which is equivalent to DDPG’s performance at 2M interactive steps with 20× more data. More visualization videos are released on our homepage. | Accept: poster |
ICLR.cc/2020/Conference | One-Shot Pruning of Recurrent Neural Networks by Jacobian Spectrum Evaluation | Recent advances in the sparse neural network literature have made it possible to prune many large feed forward and convolutional networks with only a small quantity of data. Yet, these same techniques often falter when applied to the problem of recovering sparse recurrent networks. These failures are quantitative: when pruned with recent techniques, RNNs typically obtain worse performance than they do under a simple random pruning scheme. The failures are also qualitative: the distribution of active weights in a pruned LSTM or GRU network tend to be concentrated in specific neurons and gates, and not well dispersed across the entire architecture. We seek to rectify both the quantitative and qualitative issues with recurrent network pruning by introducing a new recurrent pruning objective derived from the spectrum of the recurrent Jacobian. Our objective is data efficient (requiring only 64 data points to prune the network), easy to implement, and produces 95 % sparse GRUs that significantly improve on existing baselines. We evaluate on sequential MNIST, Billion Words, and Wikitext. | Accept (Poster) |
ICLR.cc/2021/Conference | How to compare adversarial robustness of classifiers from a global perspective | Adversarial robustness of machine learning models has attracted considerable attention over recent years. Adversarial attacks undermine the reliability of and trust in machine learning models, but the construction of more robust models hinges on a rigorous understanding of adversarial robustness as a property of a given model. Point-wise measures for specific threat models are currently the most popular tool for comparing the robustness of classifiers and are used in most recent publications on adversarial robustness. In this work, we use robustness curves to show that point-wise measures fail to capture important global properties that are essential to reliably compare the robustness of different classifiers. We introduce new ways in which robustness curves can be used to systematically uncover these properties and provide concrete recommendations for researchers and practitioners when assessing and comparing the robustness of trained models. Furthermore, we characterize scale as a way to distinguish small and large perturbations, and relate it to inherent properties of data sets, demonstrating that robustness thresholds must be chosen accordingly. We hope that our work contributes to a shift of focus away from point-wise measures of robustness and towards a discussion of the question what kind of robustness could and should reasonably be expected. We release code to reproduce all experiments presented in this paper, which includes a Python module to calculate robustness curves for arbitrary data sets and classifiers, supporting a number of frameworks, including TensorFlow, PyTorch and JAX. | Reject |
ICLR.cc/2022/Conference | Towards Federated Learning on Time-Evolving Heterogeneous Data | Federated Learning (FL) is an emerging learning paradigm that preserves privacy by ensuring client data locality on edge devices. The optimization of FL is challenging in practice due to the diversity and heterogeneity of the learning system. Despite recent research efforts on improving the optimization of heterogeneous data, the impact of time-evolving heterogeneous data in real-world scenarios, such as changing client data or intermittent clients joining or leaving during training, has not been well studied.
In this work, we propose Continual Federated Learning (CFL), a flexible framework, to capture the time-evolving heterogeneity of FL. CFL covers complex and realistic scenarios---which are challenging to evaluate in previous FL formulations---by extracting the information of past local datasets and approximating the local objective functions. Theoretically, we demonstrate that CFL methods achieve a faster convergence rate than FedAvg in time-evolving scenarios, with the benefit being dependent on approximation quality. In a series of experiments, we show that the numerical findings match the convergence analysis, and CFL methods significantly outperform the other SOTA FL baselines. | Reject |
ICLR.cc/2023/Conference | WaGI: Wavelet-based GAN Inversion for Preserving High-Frequency Image Details | Recent GAN inversion models focus on preserving image-specific details through various methods, e.g., generator tuning or feature mixing.
While those are helpful for preserving details compared to naive low-rate latent inversion, they still fail to maintain high-frequency features precisely. In this paper, we point out that existing GAN inversion models have inherent limitations in both structural and training aspects, which preclude the delicate reconstruction of high-frequency features. Especially, we prove that the widely-used loss term in GAN inversion, i.e., is biased to mainly reconstructing low-frequency features. To overcome this problem, we propose a novel GAN inversion model, coined WaGI, which enables handling high-frequency features explicitly, by using a novel wavelet-based loss term and a newly proposed wavelet fusion scheme. To the best of our knowledge, WaGI is the first approach to interpret GAN inversion in the frequency domain. We demonstrate that WaGI shows outstanding results on both inversion and editing, compared to existing state-of-the-art GAN inversion models. Especially, WaGI robustly preserves high-frequency features of images even in the editing scenario. We will release our code with the pre-trained model after the review. | Reject |
ICLR.cc/2020/Conference | Reasoning-Aware Graph Convolutional Network for Visual Question Answering | Relational reasoning methods based on graph networks are currently state-of-the-art models for Visual Question Answering (VQA) tasks involving real images. Although graph networks are used in these models to enrich visual representations by encoding question-adaptive inter-object relations, these simple graph networks is arguably insufficient to perform visual reasoning for VQA tasks. In this paper, we propose a Reasoning-Aware Graph Convolutional Networks (RA-GCN) that goes one step further towards visual reasoning for GCNs. Our first contribution is the introduction of visual reasoning ability into conventional GCNs. Secondly, we strengthen the expressive power of GCNs via introducing node-sensitive kernel parameters based on edge features to address the limitation of shared transformation matrix for each node in GCNs. Finally, we provide a novel iterative reasoning network architecture for solving VQA task via embedding the RA-GCN module into an iterative process. We evaluate our model on the VQA-CP v2, GQA and Clevr dataset. Our final RA-GCN network successfully achieves state-of-the-art accuracy which is 42.3% on the VQA-CP v2, and highly competitive 62.4% accuracy on the GQA, as well as 90.0% on val split of Clevr dataset. | Withdrawn |
ICLR.cc/2020/Conference | Neural Architecture Search by Learning Action Space for Monte Carlo Tree Search | Neural Architecture Search (NAS) has emerged as a promising technique for automatic neural network design. However, existing NAS approaches often utilize manually designed action space, which is not directly related to the performance metric to be optimized (e.g., accuracy). As a result, using manually designed action space to perform NAS often leads to sample-inefficient explorations of architectures and thus can be sub-optimal. In order to improve sample efficiency, this paper proposes Latent Action Neural Architecture Search (LaNAS) that learns actions to recursively partition the search space into good or bad regions that contain networks with concentrated performance metrics, i.e., low variance. During the search phase, as different architecture search action sequences lead to regions of different performance, the search efficiency can be significantly improved by biasing towards the good regions. On the largest NAS dataset NASBench-101, our experimental results demonstrated that LaNAS is 22x, 14.6x, 12.4x, 6.8x, 16.5x more sample-efficient than Random Search, Regularized Evolution, Monte Carlo Tree Search, Neural Architecture Optimization, and Bayesian Optimization, respectively. When applied to the open domain, LaNAS achieves 98.0% accuracy on CIFAR-10 and 75.0% top1 accuracy on ImageNet in only 803 samples, outperforming SOTA AmoebaNet with 33x fewer samples. | Reject |
ICLR.cc/2023/Conference | Can GNNs Learn Heuristic Information for Link Prediction? | Graph Neural Networks (GNNs) have shown superior performance in Link Prediction (LP). Especially, SEAL and its successors address the LP problem by classifying the subgraphs extracted specifically for candidate links, gaining state-of-the-art results. Nevertheless, we question whether these methods can effectively learn the information equivalent to link heuristics such as Common Neighbors, Katz index, etc. (we refer to such information as heuristic information in this work). We show that link heuristics and GNNs capture different information. Link heuristics usually collect pair-specific information by counting the involved neighbors or paths between two nodes in a candidate link, while GNNs learn node-wise representations through a neighborhood aggregation algorithm in which two nodes in the candidate link do not pay special attention to each other. Our further analysis shows that SEAL-type methods only use a GNN to model the pair-specific subgraphs and also cannot effectively capture heuristic information. To verify our analysis, a straightforward way is to compare the LP performance between existing methods and a model that learns heuristic information independently of the GNN learning. To this end, we present a simple yet light framework ComHG by directly Combining the embeddings of link Heuristics and the representations produced by a GNN. Experiments on OGB LP benchmarks show that ComHG outperforms all top competitors by a large margin, empirically confirming our propositions. Our experimental study also indicates that the contributions of link heuristics and the GNN to LP are sensitive to the graph degree, where the former is powerful on sparse graphs while the latter becomes dominant on dense graphs. | Reject |
ICLR.cc/2019/Conference | Variational Autoencoders for Text Modeling without Weakening the Decoder | Previous work (Bowman et al., 2015; Yang et al., 2017) has found difficulty developing generative models based on variational autoencoders (VAEs) for text. To address the problem of the decoder ignoring information from the encoder (posterior collapse), these previous models weaken the capacity of the decoder to force the model to use information from latent variables. However, this strategy is not ideal as it degrades the quality of generated text and increases hyper-parameters. In this paper, we propose a new VAE for text utilizing a multimodal prior distribution, a modified encoder, and multi-task learning. We show our model can generate well-conditioned sentences without weakening the capacity of the decoder. Also, the multimodal prior distribution improves the interpretability of acquired representations. | Withdrawn |
ICLR.cc/2023/Conference | Holistic Adversarially Robust Pruning | Neural networks can be drastically shrunk in size by removing redundant parameters. While crucial for the deployment on resource-constraint hardware, oftentimes, compression comes with a severe drop in accuracy and lack of adversarial robustness. Despite recent advances, counteracting both aspects has only succeeded for moderate compression rates so far. We propose a novel method, HARP, that copes with aggressive pruning significantly better than prior work. For this, we consider the network holistically. We learn a global compression strategy that optimizes how many parameters (compression rate) and which parameters (scoring connections) to prune specific to each layer individually. Our method fine-tunes an existing model with dynamic regularization, that follows a step-wise incremental function balancing the different objectives. It starts by favoring robustness before shifting focus on reaching the target compression rate and only then handles the objectives equally. The learned compression strategies allow us to maintain the pre-trained model’s natural accuracy and its adversarial robustness for a reduction by 99% of the network’s original size. Moreover, we observe a crucial influence of non-uniform compression across layers. The implementation of HARP is publicly available at https://intellisec.de/research/harp. | Accept: poster |
ICLR.cc/2022/Conference | Towards Model Agnostic Federated Learning Using Knowledge Distillation | Is it possible to design an universal API for federated learning using which an ad-hoc group of data-holders (agents) collaborate with each other and perform federated learning? Such an API would necessarily need to be model-agnostic i.e. make no assumption about the model architecture being used by the agents, and also cannot rely on having representative public data at hand. Knowledge distillation (KD) is the obvious tool of choice to design such protocols. However, surprisingly, we show that most natural KD-based federated learning protocols have poor performance.
To investigate this, we propose a new theoretical framework, Federated Kernel ridge regression, which can capture both model heterogeneity as well as data heterogeneity. Our analysis shows that the degradation is largely due to a fundamental limitation of knowledge distillation under data heterogeneity. We further validate our framework by analyzing and designing new protocols based on KD. Their performance on real world experiments using neural networks, though still unsatisfactory, closely matches our theoretical predictions. | Accept (Poster) |
ICLR.cc/2023/Conference | Federated Learning as Variational Inference: A Scalable Expectation Propagation Approach | The canonical formulation of federated learning treats it as a distributed optimization problem where the model parameters are optimized against a global loss function that decomposes across client loss functions. A recent alternative formulation instead treats federated learning as a distributed inference problem, where the goal is to infer a global posterior from partitioned client data (Al-Shedivat et al., 2021). This paper extends the inference view and describes a variational inference formulation of federated learning where the goal is to find a global variational posterior that well-approximates the true posterior. This naturally motivates an expectation propagation approach to federated learning (FedEP), where approximations to the global posterior are iteratively refined through probabilistic message-passing between the central server and the clients. We conduct an extensive empirical study across various algorithmic considerations and describe practical strategies for scaling up expectation propagation to the modern federated setting. We apply FedEP on standard federated learning benchmarks and find that it outperforms strong baselines in terms of both convergence speed and accuracy. | Accept: poster |
ICLR.cc/2022/Conference | Closed-Loop Control of Additive Manufacturing via Reinforcement Learning | Additive manufacturing suffers from imperfections in hardware control and material consistency. As a result, the deposition of a large range of materials requires on-the-fly adjustment of process parameters. Unfortunately, learning the in-process control is challenging. The deposition parameters are complex and highly coupled, artifacts occur after long time horizons, available simulators lack predictive power, and learning on hardware is intractable. In this work, we demonstrate the feasibility of learning a closed-loop control policy for additive manufacturing. To achieve this goal, we assume that the perception of a deposition device is limited and can capture the process only qualitatively. We leverage this assumption to formulate an efficient numerical model that explicitly includes printing imperfections. We further show that in combination with reinforcement learning, our model can be used to discover control policies that outperform state-of-the-art controllers. Furthermore, the recovered policies have a minimal sim-to-real gap. We showcase this by implementing a first-of-its-kind self-correcting printer. | Reject |
ICLR.cc/2021/Conference | A Deeper Look at Discounting Mismatch in Actor-Critic Algorithms | We investigate the discounting mismatch in actor-critic algorithm implementations from a representation learning perspective. Theoretically, actor-critic algorithms usually have discounting for both actor and critic, i.e., there is a $\gamma^t$ term in the actor update for the transition observed at time $t$ in a trajectory and the critic is a discounted value function. Practitioners, however, usually ignore the discounting ($\gamma^t$) for the actor while using a discounted critic. We investigate this mismatch in two scenarios. In the first scenario, we consider optimizing an undiscounted objective $(\gamma = 1)$ where $\gamma^t$ disappears naturally $(1^t = 1)$. We then propose to interpret the discounting in critic in terms of a bias-variance-representation trade-off and provide supporting empirical results. In the second scenario, we consider optimizing a discounted objective ($\gamma < 1$) and propose to interpret the omission of the discounting in the actor update from an auxiliary task perspective and provide supporting empirical results. | Reject |
ICLR.cc/2023/Conference | Learning Efficient Hybrid Particle-continuum Representations of Non-equilibrium N-body Systems | An important class of multi-scale, non-equilibrium, N-body physical systems deals with an interplay between particle and continuum phenomena. These include hypersonic flow and plasma dynamics, materials science, and astrophysics. Hybrid solvers that combine particle and continuum representations could provide an efficient framework to model these systems. However, the coupling between these two representations has been a key challenge, which is often limited to inaccurate or incomplete prescriptions. In this work, we introduce a method for Learning Hybrid Particle-Continuum (LHPC) models from the data of first-principles particle simulations. LHPC analyzes the local velocity-space particle distribution function and separates it into near-equilibrium (thermal) and far-from-equilibrium (non-thermal) components. The most computationally-intensive particle solver is used to advance the non-thermal particles, whereas a neural network solver is used to efficiently advance the thermal component using a continuum representation. Most importantly, an additional neural network learns the particle-continuum coupling: the dynamical exchange of mass, momentum, and energy between the particle and continuum representations. Training of the different neural network components is done in an integrated manner to ensure global consistency and stability of the LHPC model. We demonstrate our method in an intense laser-plasma interaction problem involving highly nonlinear, far-from-equilibrium dynamics associated with the coupling between electromagnetic fields and multiple particle species. More efficient modeling of these interactions is critical for the design and optimization of compact accelerators for material science and medical applications. Our method achieves an important balance between accuracy and speed: LHPC is 8 times faster than a classical particle solver and achieves up to 6.8-fold reduction of long-term prediction error for key quantities of interest compared to deep-learning baselines using uniform representations. | Reject |
ICLR.cc/2022/Conference | State-Action Joint Regularized Implicit Policy for Offline Reinforcement Learning | Offline reinforcement learning enables learning from a fixed dataset, without further interactions with the environment. The lack of environmental interactions makes the policy training vulnerable to state-action pairs far from the training dataset and prone to missing rewarding actions. For training more effective agents, we propose a framework that supports learning a flexible and well-regularized policy, which consists of a fully implicit policy and a regularization through the state-action visitation frequency induced by the current policy and that induced by the data-collecting behavior policy. We theoretically show the equivalence between policy-matching and state-action-visitation matching, and thus the compatibility of many prior work with our framework. An effective instantiation of our framework through the GAN structure is provided, together with some techniques to explicitly smooth the state-action mapping for robust generalization beyond the static dataset. Extensive experiments and ablation study on the D4RL dataset validate our framework and the effectiveness of our algorithmic designs. | Reject |
ICLR.cc/2022/Conference | Safe Neurosymbolic Learning with Differentiable Symbolic Execution | We study the problem of learning verifiably safe parameters for programs that use neural networks as well as symbolic, human-written code. Such neurosymbolic programs arise in many safety-critical domains. However, because they need not be differentiable, it is hard to learn their parameters using existing gradient-based approaches to safe learning. Our method, Differentiable Symbolic Execution (DSE), samples control flow paths in a program, symbolically constructs worst-case "safety loss" along these paths, and backpropagates the gradients of these losses through program operations using a generalization of the REINFORCE estimator. We evaluate the method on a mix of synthetic tasks and real-world benchmarks. Our experiments show that DSE significantly outperforms the state-of-the-art DiffAI method on these tasks. | Accept (Poster) |