conference
stringclasses 6
values | title
stringlengths 8
176
| abstract
stringlengths 228
5k
| decision
stringclasses 9
values |
|---|---|---|---|
ICLR.cc/2018/Conference | Stabilizing Adversarial Nets with Prediction Methods | Adversarial neural networks solve many important problems in data science, but are notoriously difficult to train. These difficulties come from the fact that optimal weights for adversarial nets correspond to saddle points, and not minimizers, of the loss function. The alternating stochastic gradient methods typically used for such problems do not reliably converge to saddle points, and when convergence does happen it is often highly sensitive to learning rates. We propose a simple modification of stochastic gradient descent that stabilizes adversarial networks. We show, both in theory and practice, that the proposed method reliably converges to saddle points. This makes adversarial networks less likely to "collapse," and enables faster training with larger learning rates. | Accept (Poster) |
ICLR.cc/2020/Conference | Continual Learning with Gated Incremental Memories for Sequential Data Processing | The ability to learn over changing task distributions without forgetting previous knowledge, also known as continual learning, is a key enabler for scalable and trustworthy deployments of adaptive solutions. While the importance of continual learning is largely acknowledged in machine vision and reinforcement learning problems, this is mostly under-documented for sequence processing tasks. This work focuses on characterizing and quantitatively assessing the impact of catastrophic forgetting and task interference when dealing with sequential data in recurrent neural networks. We also introduce a general architecture, named Gated Incremental Memory, for augmenting recurrent models with continual learning skills, whose effectiveness is demonstrated through the benchmarks introduced in this paper. | Reject |
ICLR.cc/2019/Conference | LSH Microbatches for Stochastic Gradients: Value in Rearrangement | Metric embeddings are immensely useful representations of associations between entities (images, users, search queries, words, and more). Embeddings are learned by optimizing a loss objective of the general form of a sum over example associations. Typically, the optimization uses stochastic gradient updates over minibatches of examples that are arranged independently at random. In this work, we propose the use of {\em structured arrangements} through randomized {\em microbatches} of examples that are more likely to include similar ones. We make a principled argument for the properties of our arrangements that accelerate the training and present efficient algorithms to generate microbatches that respect the marginal distribution of training examples. Finally, we observe experimentally that our structured arrangements accelerate training by 3-20\%. Structured arrangements emerge as a powerful and novel performance knob for SGD that is independent and complementary to other SGD hyperparameters and thus is a candidate for wide deployment. | Reject |
ICLR.cc/2021/Conference | Using Synthetic Data to Improve the Long-range Forecasting of Time Series Data | Effective long-range forecasting of time series data remains an unsolved and open problem. One possible approach is to use generative models to improve long-range forecasting, but the challenge then is how to generate high-quality synthetic data. In this paper, we propose a conditional Wasserstein GAN with Gradient and Error Penalty (cWGAN-GEP), aiming to generate accurate synthetic data that preserves the temporal dynamics between the conditioning input and generated data. By using such synthetic data, we develop a long-range forecasting method called Generative Forecasting (GenF). GenF consists of three key components: (i) a cWGAN-GEP based generator, to generate synthetic data for next few time steps. (ii) a predictor which makes long-range predictions based on generated and observed data. (iii) an information theoretic clustering (ITC) algorithm to better train the cWGAN-GEP based generator and the predictor. Our experimental results on three public datasets demonstrate that GenF significantly outperforms a diverse range of state-of-the-art benchmarks and classical approaches. In most cases, we find an improvement of at least 10% over all studied methods. Lastly, we conduct an ablation study to demonstrate the effectiveness of the cWGAN-GEP and the ITC algorithm. | Reject |
ICLR.cc/2020/Conference | X-Forest: Approximate Random Projection Trees for Similarity Measurement | Similarity measurement plays a central role in various data mining and machine learning tasks. Generally, a similarity measurement solution should, in an ideal state, possess the following three properties: accuracy, efficiency and independence from prior knowledge. Yet unfortunately, vital as similarity measurements are, no previous works have addressed all of them. In this paper, we propose X-Forest, consisting of a group of approximate Random Projection Trees, such that all three targets mentioned above are tackled simultaneously. Our key techniques are as follows. First, we introduced RP Trees into the tasks of similarity measurement such that accuracy is improved. In addition, we enforce certain layers in each tree to share identical projection vectors, such that exalted efficiency is achieved. Last but not least, we introduce randomness into partition to eliminate its reliance on prior knowledge. We conduct experiments on three real-world datasets, whose results demonstrate that our model, X-Forest, reaches an efficiency of up to 3.5 times higher than RP Trees with negligible compromising on its accuracy, while also being able to outperform traditional Euclidean distance-based similarity metrics by as much as 20% with respect to clustering tasks. We have released codes in github anonymously so as to meet the demand of reproducibility. | Reject |
ICLR.cc/2022/Conference | Assisted Learning for Organizations with Limited Imbalanced Data | We develop an assisted learning framework for assisting organization-level learners to improve their learning performance with limited and imbalanced data. In particular, learners at the organization level usually have sufficient computation resource, but are subject to stringent collaboration policy and information privacy. Their limited imbalanced data often cause biased inference and sub-optimal decision-making. In our assisted learning framework, an organizational learner purchases assistance service from a service provider and aims to enhance its model performance within a few assistance rounds. We develop effective stochastic training algorithms for assisted deep learning and assisted reinforcement learning. Different from existing distributed algorithms that need to frequently transmit gradients or models, our framework allows the learner to only occasionally share information with the service provider, and still achieve a near-oracle model as if all the data were centralized. | Reject |
ICLR.cc/2022/Conference | On Learning with Fairness Trade-Offs | Previous literature has shown that bias mitigating algorithms were sometimes prone to overfitting and had poor out-of-sample generalisation. This paper is first and foremost concerned with establishing a mathematical framework to tackle the specific issue of generalisation. Throughout this work, we consider fairness trade-offs and objectives mixing statistical loss over the whole sample and fairness penalties on categories (which could stem from different values of protected attributes), encompassing partial de-biasing. We do so by adopting two different but complementary viewpoints: first, we consider a PAC-type setup and derive probabilistic upper bounds involving sample-only information; second, we leverage an asymptotic framework to derive a closed-form limiting distribution for the difference between the empirical trade-off and the true trade-off. While these results provide guarantees for learning fairness metrics across categories, they also point out to the key (but asymmetric) role played by class imbalance. To summarise, learning fairness without having access to enough category-level samples is hard, and a simple numerical experiment shows that it can lead to spurious results. | Reject |
ICLR.cc/2020/Conference | RTFM: Generalising to New Environment Dynamics via Reading | Obtaining policies that can generalise to new environments in reinforcement learning is challenging. In this work, we demonstrate that language understanding via a reading policy learner is a promising vehicle for generalisation to new environments. We propose a grounded policy learning problem, Read to Fight Monsters (RTFM), in which the agent must jointly reason over a language goal, relevant dynamics described in a document, and environment observations. We procedurally generate environment dynamics and corresponding language descriptions of the dynamics, such that agents must read to understand new environment dynamics instead of memorising any particular information. In addition, we propose txt2π, a model that captures three-way interactions between the goal, document, and observations. On RTFM, txt2π generalises to new environments with dynamics not seen during training via reading. Furthermore, our model outperforms baselines such as FiLM and language-conditioned CNNs on RTFM. Through curriculum learning, txt2π produces policies that excel on complex RTFM tasks requiring several reasoning and coreference steps. | Accept (Poster) |
ICLR.cc/2023/Conference | This Looks Like It Rather Than That: ProtoKNN For Similarity-Based Classifiers | Among research on the interpretability of deep learning models, the 'this looks like that' framework with ProtoPNet has attracted significant attention. By combining the strong power of deep learning models with the interpretability of case-based inference, ProtoPNet can achieve high accuracy while keeping its reasoning process interpretable. Many methods based on ProtoPNet have emerged to take advantage of this benefit, but despite their practical usefulness, they run into difficulty when utilizing similarity-based classifiers, e.g., in domains where unknown class samples exist. This is because ProtoPNet and its variants adopt the training process specific to linear classifiers, which allows the prototypes to represent useful image features for class recognition. Due to this difficulty, the effectiveness of similarity-based classifiers (e.g., k-nearest neighbor (KNN)) on the 'this looks like that' framework have not been sufficiently examined. To alleviate this problem, we propose ProtoKNN, an extension of ProtoPNet that adopts KNN classifiers. Extensive experiments on multiple open datasets demonstrate that the proposed method can achieve competitive results with a state-of-the-art method. | Accept: poster |
ICLR.cc/2023/Conference | Multi-View Masked Autoencoders for Visual Control | This paper investigates how to leverage data from multiple cameras to learn representations beneficial for visual control. To this end, we present the Multi-View Masked Autoencoder (MV-MAE), a simple and scalable framework for multi-view representation learning. Our main idea is to mask multiple viewpoints from video frames at random and train a video autoencoder to reconstruct pixels of both masked and unmasked viewpoints. This allows the model to learn representations that capture useful information of the current viewpoint but also the cross-view information from different viewpoints. We evaluate MV-MAE on challenging RLBench visual manipulation tasks by training a reinforcement learning agent on top of frozen representations. Our experiments demonstrate that MV-MAE significantly outperforms other multi-view representation learning approaches. Moreover, we show that the number of cameras can differ between the representation learning phase and the behavior learning phase. By training a single-view control agent on top of multi-view representations from MV-MAE, we achieve 62.3% success rate while the single-view representation learning baseline achieves 42.3%. | Reject |
ICLR.cc/2022/Conference | On Bridging Generic and Personalized Federated Learning for Image Classification | Federated learning is promising for its capability to collaboratively train models with multiple clients without accessing their data, but vulnerable when clients' data distributions diverge from each other. This divergence further leads to a dilemma: "Should we prioritize the learned model's generic performance (for future use at the server) or its personalized performance (for each client)?" These two, seemingly competing goals have divided the community to focus on one or the other, yet in this paper we show that it is possible to approach both at the same time. Concretely, we propose a novel federated learning framework that explicitly decouples a model's dual duties with two prediction tasks. On the one hand, we introduce a family of losses that are robust to non-identical class distributions, enabling clients to train a generic predictor with a consistent objective across them. On the other hand, we formulate the personalized predictor as a lightweight adaptive module that is learned to minimize each client's empirical risk on top of the generic predictor. With this two-loss, two-predictor framework which we name Federated Robust Decoupling (Fed-RoD), the learned model can simultaneously achieve state-of-the-art generic and personalized performance, essentially bridging the two tasks. | Accept (Spotlight) |
ICLR.cc/2023/Conference | Spatio-temporal point processes with deep non-stationary kernels | Point process data are becoming ubiquitous in modern applications, such as social networks, health care, and finance. Despite the powerful expressiveness of the popular recurrent neural network (RNN) models for point process data, they may not successfully capture sophisticated non-stationary dependencies in the data due to their recurrent structures. Another popular type of deep model for point process data is based on representing the influence kernel (rather than the intensity function) by neural networks. We take the latter approach and develop a new deep non-stationary influence kernel that can model non-stationary spatio-temporal point processes. The main idea is to approximate the influence kernel with a novel and general low-rank decomposition, enabling efficient representation through deep neural networks and computational efficiency and better performance. We also take a new approach to maintain the non-negativity constraint of the conditional intensity by introducing a log-barrier penalty. We demonstrate our proposed method's good performance and computational efficiency compared with the state-of-the-art on simulated and real data. | Accept: poster |
ICLR.cc/2023/Conference | Deep Graph-Level Clustering Using Pseudo-Label-Guided Mutual Information Maximization Network | In this work, we study the problem of partitioning a set of graphs into different groups such that the graphs in the same group are similar while the graphs in different groups are dissimilar. This problem was rarely studied previously, although there have been a lot of work on node clustering and graph classification. The problem is challenging because it is difficult to measure the similarity or distance between graphs. One feasible approach is using graph kernels to compute a similarity matrix for the graphs and then performing spectral clustering, but the effectiveness of existing graph kernels in measuring the similarity between graphs is very limited. To solve the problem, we propose a novel method called Deep Graph-Level Clustering (DGLC). DGLC utilizes a graph isomorphism network to learn graph-level representations by maximizing the mutual information between the representations of entire graphs and substructures, under the regularization of a clustering module that ensures discriminative representations via pseudo labels. DGLC achieves graph-level representation learning and graph-level clustering in an end-to-end manner. The experimental results on six benchmark datasets of graphs show that our DGLC has state-of-the-art performance in comparison to many baselines. | Reject |
ICLR.cc/2021/Conference | On the Importance of Looking at the Manifold | Data rarely lies on uniquely Euclidean spaces. Even data typically represented in regular domains, such as images, can have a higher level of relational information, either between data samples or even relations within samples, e.g., how the objects in an image are linked. With this perspective our data points can be enriched by explicitly accounting for this connectivity and analyzing them as a graph. Herein, we analyze various approaches for unsupervised representation learning and investigate the importance of considering topological information and its impact when learning representations. We explore a spectrum of models, ranging from uniquely learning representations based on the isolated features of the nodes (focusing on Variational Autoencoders), to uniquely learning representations based on the topology (using node2vec) passing through models that integrate both node features and topological information in a hybrid fashion. For the latter we use Graph Neural Networks, precisely Deep Graph Infomax (DGI), and an extension of the typical formulation of the VAE where the topological structure is accounted for via an explicit regularization of the loss (Graph-Regularized VAEs, introduced in this work). To extensively investigate these methodologies, we consider a wide variety of data types: synthetic data point clouds, MNIST, citation networks, and chemical reactions. We show that each of the representations learned by these models may have critical importance for further downstream tasks, and that accounting for the topological features can greatly improve the modeling capabilities for certain problems. We further provide a framework to analyze these, and future models under different scenarios and types of data. | Reject |
ICLR.cc/2023/Conference | Schrödinger's FP: Training Neural Networks with Dynamic Floating-Point Containers | We introduce a software-hardware co-design approach to reduce memory traffic and footprint during training with BFloat16 or FP32, in order to boost energy efficiency and execution time performance. Our methods dynamically adjust the size and format of the floating-point containers used to store activations and weights during training. The different value distributions lead us to different approaches for exponents and mantissas. Gecko exploits the favourable exponent distribution with a lossless delta encoding approach to reduce the total exponent footprint by up to 58% in comparison to the FP32 baseline. To contend with the noisy mantissa distributions, we present two lossy methods to eliminate as many as possible least significant bits without affecting accuracy. Quantum Mantissa is a machine learning mantissa compression method that taps onto the gradient descent algorithm to learn the minimal mantissa bitlengths on a per-layer granularity, and obtain up to 92% reduction in total mantissa footprint. Alternatively, BitChop observes changes in the loss function during training to adjust mantissa bitlength network-wide, yielding a reduction of 81% in footprint. Schrödinger's FP implements hardware encoders/decoders that, guided by Gecko/Quantum Mantissa or Gecko/BitChop, transparently encode/decode values when transferring to/from off-chip memory, boosting energy efficiency and reducing execution time. | Reject |
ICLR.cc/2021/Conference | Is Attention Better Than Matrix Decomposition? | As an essential ingredient of modern deep learning, attention mechanism, especially self-attention, plays a vital role in the global correlation discovery. However, is hand-crafted attention irreplaceable when modeling the global context? Our intriguing finding is that self-attention is not better than the matrix decomposition~(MD) model developed 20 years ago regarding the performance and computational cost for encoding the long-distance dependencies. We model the global context issue as a low-rank completion problem and show that its optimization algorithms can help design global information blocks. This paper then proposes a series of Hamburgers, in which we employ the optimization algorithms for solving MDs to factorize the input representations into sub-matrices and reconstruct a low-rank embedding. Hamburgers with different MDs can perform favorably against the popular global context module self-attention when carefully coping with gradients back-propagated through MDs. Comprehensive experiments are conducted in the vision tasks where it is crucial to learn the global context, including semantic segmentation and image generation, demonstrating significant improvements over self-attention and its variants. Code is available at https://github.com/Gsunshine/Enjoy-Hamburger. | Accept (Poster) |
ICLR.cc/2021/Conference | Heteroskedastic and Imbalanced Deep Learning with Adaptive Regularization | Real-world large-scale datasets are heteroskedastic and imbalanced --- labels have varying levels of uncertainty and label distributions are long-tailed. Heteroskedasticity and imbalance challenge deep learning algorithms due to the difficulty of distinguishing among mislabeled, ambiguous, and rare examples. Addressing heteroskedasticity and imbalance simultaneously is under-explored. We propose a data-dependent regularization technique for heteroskedastic datasets that regularizes different regions of the input space differently. Inspired by the theoretical derivation of the optimal regularization strength in a one-dimensional nonparametric classification setting, our approach adaptively regularizes the data points in higher-uncertainty, lower-density regions more heavily. We test our method on several benchmark tasks, including a real-world heteroskedastic and imbalanced dataset, WebVision. Our experiments corroborate our theory and demonstrate a significant improvement over other methods in noise-robust deep learning. | Accept (Poster) |
ICLR.cc/2023/Conference | Reversible Column Networks | We propose a new neural network design paradigm Reversible Column Network (RevCol). The main body of RevCol is composed of multiple copies of subnetworks, named columns respectively, between which multi-level reversible connections are employed. Such architectural scheme attributes RevCol very different behavior from conventional networks: during forward propagation, features in RevCol are learned to be gradually disentangled when passing through each column, whose total information is maintained rather than compressed or discarded as other network does. Our experiments suggest that CNN-style RevCol models can achieve very competitive performances on multiple computer vision tasks such as image classification, object detection and semantic segmentation, especially with large parameter budget and large dataset. For example, after ImageNet-22K pre-training, RevCol-XL obtains 88.2% ImageNet-1K accuracy. Given more pre-training data, our largest model RevCol-H reaches 90.0% on ImageNet-1K, 63.8% AP$_{box}$ on COCO detection minival set, 61.0% mIoU on ADE20k segmentation. To our knowledge, it is the best COCO detection and ADE20k segmentation result among pure (static) CNN models. Moreover, as a general macro architecture fashion, RevCol can also be introduced into transformers or other neural networks, which is demonstrated to improve the performances in both computer vision and NLP tasks.
We release code and models at https://github.com/megvii-research/RevCol | Accept: poster |
ICLR.cc/2022/Conference | Isotropic Contextual Representations through Variational Regularization | Contextual language representations achieve state-of-the-art performance across various natural language processing tasks. However, these representations have been shown to suffer from the degeneration problem, i.e. they occupy a narrow cone in the latent space. This problem can be addressed by enforcing isotropy in the latent space. In analogy to variational autoencoders, we suggest applying a token-level variational loss to a Transformer architecture and introduce the prior distribution's standard deviation as model parameter to optimize isotropy. The encoder-decoder architecture allows for learning interpretable embeddings that can be decoded into text again. Extracted features at sentence-level achieve competitive results on benchmark classification tasks. | Reject |
ICLR.cc/2022/Conference | Maximum Likelihood Estimation for Multimodal Learning with Missing Modality | Multimodal learning has achieved great successes in many scenarios. Compared with unimodal learning, it can effectively combine the information from different modalities to improve the performance of learning tasks. In reality, the multimodal data may have missing modalities due to various reasons, such as sensor failure and data transmission error. In previous works, the information of the modality-missing data has not been well exploited. To address this problem, we propose an efficient approach based on maximum likelihood estimation to incorporate the knowledge in the modality-missing data. Specifically, we design a likelihood function to characterize the conditional distributions of the modality-complete data and the modality-missing data, which is theoretically optimal. Moreover, we develop a generalized form of the softmax function to effectively implement maximum likelihood estimation in an end-to-end manner. Such training strategy guarantees the computability of our algorithm capably. Finally, we conduct a series of experiments on real-world multimodal datasets. Our results demonstrate the effectiveness of the proposed approach, even when 95% of the training data has missing modality. | Reject |
ICLR.cc/2021/Conference | Adaptive Risk Minimization: A Meta-Learning Approach for Tackling Group Shift | A fundamental assumption of most machine learning algorithms is that the training and test data are drawn from the same underlying distribution. However, this assumption is violated in almost all practical applications: machine learning systems are regularly tested under distribution shift, due to temporal correlations, particular end users, or other factors. In this work, we consider the setting where the training data are structured into groups and test time shifts correspond to changes in the group distribution. Prior work has approached this problem by attempting to be robust to all possible test time distributions, which may degrade average performance. In contrast, we propose to use ideas from meta-learning to learn models that are adaptable, such that they can adapt to shift at test time using a batch of unlabeled test points. We acquire such models by learning to adapt to training batches sampled according to different distributions, which simulate structural shifts that may occur at test time. Our primary contribution is to introduce the framework of adaptive risk minimization (ARM), a formalization of this setting that lends itself to meta-learning. We develop meta-learning methods for solving the ARM problem, and compared to a variety of prior methods, these methods provide substantial gains on image classification problems in the presence of shift. | Reject |
ICLR.cc/2018/Conference | TESLA: Task-wise Early Stopping and Loss Aggregation for Dynamic Neural Network Inference | For inference operations in deep neural networks on end devices, it is desirable to deploy a single pre-trained neural network model, which can dynamically scale across a computation range without comprising accuracy. To achieve this goal, Incomplete Dot Product (IDP) has been proposed to use only a subset of terms in dot products during forward propagation. However, there are some limitations, including noticeable performance degradation in operating regions with low computational costs, and essential performance limitations since IDP uses hand-crafted profile coefficients. In this paper, we extend IDP by proposing new training algorithms involving a single profile, which may be trainable or pre-determined, to significantly improve the overall performance, especially in operating regions with low computational costs. Specifically, we propose the Task-wise Early Stopping and Loss Aggregation (TESLA) algorithm, which is showed in our 3-layer multilayer perceptron on MNIST that outperforms the original IDP by 32\% when only 10\% of dot products terms are used and achieves 94.7\% accuracy on average. By introducing trainable profile coefficients, TESLA further improves the accuracy to 95.5\% without specifying coefficients in advance. Besides, TESLA is applied to the VGG-16 model, which achieves 80\% accuracy using only 20\% of dot product terms on CIFAR-10 and also keeps 60\% accuracy using only 30\% of dot product terms on CIFAR-100, but the original IDP performs like a random guess in these two datasets at such low computation costs. Finally, we visualize the learned representations at different dot product percentages by class activation map and show that, by applying TESLA, the learned representations can adapt over a wide range of operation regions. | Reject |
ICLR.cc/2020/Conference | Poly-encoders: Architectures and Pre-training Strategies for Fast and Accurate Multi-sentence Scoring | The use of deep pre-trained transformers has led to remarkable progress in a number of applications (Devlin et al., 2018). For tasks that make pairwise comparisons between sequences, matching a given input with a corresponding label, two approaches are common: Cross-encoders performing full self-attention over the pair and Bi-encoders encoding the pair separately. The former often performs better, but is too slow for practical use. In this work, we develop a new transformer architecture, the Poly-encoder, that learns global rather than token level self-attention features. We perform a detailed comparison of all three approaches, including what pre-training and fine-tuning strategies work best. We show our models achieve state-of-the-art results on four tasks; that Poly-encoders are faster than Cross-encoders and more accurate than Bi-encoders; and that the best results are obtained by pre-training on large datasets similar to the downstream tasks. | Accept (Poster) |
ICLR.cc/2019/Conference | Learning Physics Priors for Deep Reinforcement Learing | While model-based deep reinforcement learning (RL) holds great promise for sample efficiency and generalization, learning an accurate dynamics model is challenging and often requires substantial interactions with the environment. Further, a wide variety of domains have dynamics that share common foundations like the laws of physics, which are rarely exploited by these algorithms. Humans often acquire such physics priors that allow us to easily adapt to the dynamics of any environment. In this work, we propose an approach to learn such physics priors and incorporate them into an RL agent. Our method involves pre-training a frame predictor on raw videos and then using it to initialize the dynamics prediction model on a target task. Our prediction model, SpatialNet, is designed to implicitly capture localized physical phenomena and interactions. We show the value of incorporating this prior through empirical experiments on two different domains – a newly created PhysWorld and games from the Atari benchmark, outperforming competitive approaches and demonstrating effective transfer learning. | Reject |
ICLR.cc/2020/Conference | The Logical Expressiveness of Graph Neural Networks | The ability of graph neural networks (GNNs) for distinguishing nodes in graphs has been recently characterized in terms of the Weisfeiler-Lehman (WL) test for checking graph isomorphism. This characterization, however, does not settle the issue of which Boolean node classifiers (i.e., functions classifying nodes in graphs as true or false) can be expressed by GNNs. We tackle this problem by focusing on Boolean classifiers expressible as formulas in the logic FOC2, a well-studied fragment of first order logic. FOC2 is tightly related to the WL test, and hence to GNNs. We start by studying a popular class of GNNs, which we call AC-GNNs, in which the features of each node in the graph are updated, in successive layers, only in terms of the features of its neighbors. We show that this class of GNNs is too weak to capture all FOC2 classifiers, and provide a syntactic characterization of the largest subclass of FOC2 classifiers that can be captured by AC-GNNs. This subclass coincides with a logic heavily used by the knowledge representation community. We then look at what needs to be added to AC-GNNs for capturing all FOC2 classifiers. We show that it suffices to add readout functions, which allow to update the features of a node not only in terms of its neighbors, but also in terms of a global attribute vector. We call GNNs of this kind ACR-GNNs. We experimentally validate our findings showing that, on synthetic data conforming to FOC2 formulas, AC-GNNs struggle to fit the training data while ACR-GNNs can generalize even to graphs of sizes not seen during training. | Accept (Spotlight) |
ICLR.cc/2023/Conference | Self-supervision through Random Segments with Autoregressive Coding (RandSAC) | Inspired by the success of self-supervised autoregressive representation learning in natural language (GPT and its variants), and advances in recent visual architecture design with Vision Transformers (ViTs), in this paper, we explore the effects various design choices have on the success of applying such training strategies for visual feature learning. Specifically, we introduce a novel strategy that we call Random Segments with Autoregressive Coding (RandSAC). In RandSAC, we group patch representations (image tokens) into hierarchically arranged segments; within each segment, tokens are predicted in parallel, similar to BERT, while across segment predictions are sequential, similar to GPT. We illustrate that randomized serialization of the segments significantly improves the performance and results in distribution over spatially-long (across-segments) and -short (within-segment) predictions which are effective for feature learning. We illustrate the pertinence of these design choices and explore alternatives on a number of datasets (e.g., CIFAR10, ImageNet). While our pre-training strategy works with vanilla Transformer, we also propose a conceptually simple, but highly effective, addition to the decoder that allows learnable skip-connections to encoder feature layers, which further improves the performance. | Accept: poster |
ICLR.cc/2023/Conference | The Plug and Play of Language Models for Text-to-image Generation | Text-to-image (T2I) models enable controllable image generation through user-provided captions. A text encoder is typically used to map captions to a latent space, and it has been shown to be critical for model's performance. However, replacing or upgrading the text encoder in a T2I model is challenging due to the tight bond between the current encoder and the image decoder. It requires training the model from scratch, which can be prohibitively expensive. To address this problem, we introduce a more efficient approach to align a pre-trained language model with the latent space of an existing T2I model. We propose a Model Translation Network (MTN) and a new training objective to align the representation spaces of the two text encoders using only a corpus of unlabeled text. We empirically find that MTN can be trained efficiently and can boost the performance of existing T2I models by upgrading their text encoder. Moreover, we find that MTN can align multilingual language models such as XLM-Roberta, thus allowing existing T2I models to generate high-quality images from captions beyond English. | Reject |
ICLR.cc/2023/Conference | Efficient Attention via Control Variates | Random-feature-based attention (RFA) is an efficient approximation of softmax attention with linear runtime and space complexity. However, the approximation gap between RFA and conventional softmax attention is not well studied. Built upon previous progress of RFA, we characterize this gap through the lens of control variates and show that RFA can be decomposed into a sum of multiple control variate estimators for each element in the sequence. This new framework reveals that exact softmax attention can be recovered from RFA by manipulating each control variate. Besides, it allows us to develop a more flexible form of control variates, resulting in a novel attention mechanism that significantly reduces the approximation gap while maintaining linear complexity. Extensive experiments demonstrate that our model outperforms state-of-the-art efficient attention mechanisms on both vision and language tasks. | Accept: notable-top-5% |
ICLR.cc/2021/Conference | Importance-based Multimodal Autoencoder | Integrating information from multiple modalities (e.g., verbal, acoustic and visual data) into meaningful representations has seen great progress in recent years. However, two challenges are not sufficiently addressed by current approaches: (1) computationally efficient training of multimodal autoencoder networks which are robust in the absence of modalities, and (2) unsupervised learning of important subspaces in each modality which are correlated with other modalities. In this paper we propose the IMA (Importance-based Multimodal Autoencoder) model, a scalable model that learns modality importances and robust multimodal representations through a novel cross-covariance based loss function. We conduct experiments on MNIST-TIDIGITS a multimodal dataset of spoken and image digits,and on IEMOCAP, a multimodal emotion corpus. The IMA model is able to distinguish digits from uncorrelated noise, and word-level importances learnt that correspond to the separation between function and emotional words. The multimodal representations learnt by IMA are also competitive with state-of-the-art baseline approaches on downstream tasks. | Reject |
ICLR.cc/2018/Conference | SpectralNet: Spectral Clustering using Deep Neural Networks | Spectral clustering is a leading and popular technique in unsupervised data analysis. Two of its major limitations are scalability and generalization of the spectral embedding (i.e., out-of-sample-extension). In this paper we introduce a deep learning approach to spectral clustering that overcomes the above shortcomings. Our network, which we call SpectralNet, learns a map that embeds input data points into the eigenspace of their associated graph Laplacian matrix and subsequently clusters them. We train SpectralNet using a procedure that involves constrained stochastic optimization. Stochastic optimization allows it to scale to large datasets, while the constraints, which are implemented using a special purpose output layer, allow us to keep the network output orthogonal. Moreover, the map learned by SpectralNet naturally generalizes the spectral embedding to unseen data points. To further improve the quality of the clustering, we replace the standard pairwise Gaussian affinities with affinities leaned from unlabeled data using a Siamese network. Additional improvement can be achieved by applying the network to code representations produced, e.g., by standard autoencoders. Our end-to-end learning procedure is fully unsupervised. In addition, we apply VC dimension theory to derive a lower bound on the size of SpectralNet. State-of-the-art clustering results are reported for both the MNIST and Reuters datasets.
| Accept (Poster) |
ICLR.cc/2019/Conference | Accelerating first order optimization algorithms | There exist several stochastic optimization algorithms. However in most cases, it is difficult to tell for a particular problem which will be the best optimizer to choose as each of them are good. Thus, we present a simple and intuitive technique, when applied to first order optimization algorithms, is able to improve the speed of convergence and reaches a better minimum for the loss function compared to the original algorithms. The proposed solution modifies the update rule, based on the variation of the direction of the gradient during training. We conducted several tests with Adam and AMSGrad on two different datasets. The preliminary results show that the proposed technique improves the performance of existing optimization algorithms and works well in practice. | Reject |
ICLR.cc/2018/Conference | Stabilizing Gradients for Deep Neural Networks via Efficient SVD Parameterization | Vanishing and exploding gradients are two of the main obstacles in training deep neural networks, especially in capturing long range dependencies in recurrent neural networks (RNNs). In this paper, we present an efficient parametrization of the transition matrix of an RNN that allows us to stabilize the gradients that arise in its training. Specifically, we parameterize the transition matrix by its singular value decomposition (SVD), which allows us to explicitly track and control its singular values. We attain efficiency by using tools that are common in numerical linear algebra, namely Householder reflectors for representing the orthogonal matrices that arise in the SVD. By explicitly controlling the singular values, our proposed svdRNN method allows us to easily solve the exploding gradient problem and we observe that it empirically solves the vanishing gradient issue to a large extent. We note that the SVD parameterization can be used for any rectangular weight matrix, hence it can be easily extended to any deep neural network, such as a multi-layer perceptron. Theoretically, we demonstrate that our parameterization does not lose any expressive power, and show how it potentially makes the optimization process easier. Our extensive experimental results also demonstrate that the proposed framework converges faster, and has good generalization, especially when the depth is large.
| Reject |
ICLR.cc/2023/Conference | $\sigma$Reparam: Stable Transformer Training with Spectral Reparametrization | Training stability is of great importance to Transformers. In this work, we investigate the training dynamics of Transformers by examining the evolution of the attention layers. In particular, we track the "attention entropy" for each attention head during the course of training, which is a proxy of the attention's sharpness. We observe a common, non monotonic evolution of attention entropy across different settings: the attention entropy first quickly decreases in the initial phase of training, followed by quickly increasing, and finally entering a long stable phase. While the exact shape can be affected by hyperparameters such as warmup, initialization, learning rate etc., we found that there is a close correlation between the minima of attention entropy and the model's training stability. To this end, we propose a simple and efficient solution dubbed $\sigma$Reparam, where we reparametrize all linear layers with Spectral Normalization and an additional learned scalar. We provide a lower bound on the attention entropy as a function of the spectral norms of the query and key projections, which suggests that small attention entropy can be obtained with large spectral norms. $\sigma$Reparam decouples the growth rate of a weight matrix's spectral norm from its dimensionality, which we verify empirically. We conduct experiments with $\sigma$Reparam on image classification, image self supervised learning, automatic speech recognition and language modeling tasks. We show that $\sigma$Reparam provides great stability and robustness with respect to the choice of hyperparameters. | Reject |
ICLR.cc/2020/Conference | Detecting Change in Seasonal Pattern via Autoencoder and Temporal Regularization | Change-point detection problem consists of discovering abrupt property changes in the generation process of time-series. Most state-of-the-art models are optimizing the power of a kernel two-sample test, with only a few assumptions on the distribution of the data. Unfortunately, because they presume the samples are distributed i.i.d, they are not able to use information about the seasonality of a time-series. In this paper, we present a novel approach - ATR-CSPD allowing the detection of changes in the seasonal pattern of a time-series. Our method uses an autoencoder together with a temporal regularization, to learn the pattern of each seasonal cycle. Using low dimensional representation of the seasonal patterns, it is possible to accurately and efficiently estimate the existence of a change point using a clustering algorithm. Through experiments on artificial and real-world data sets, we demonstrate the usefulness of the proposed method for several applications. | Reject |
ICLR.cc/2020/Conference | Finding Winning Tickets with Limited (or No) Supervision | The lottery ticket hypothesis argues that neural networks contain sparse subnetworks, which, if appropriately initialized (the winning tickets), are capable of matching the accuracy of the full network when trained in isolation. Empirically made in different contexts, such an observation opens interesting questions about the dynamics of neural network optimization and the importance of their initializations. However, the properties of winning tickets are not well understood, especially the importance of supervision in the generating process. In this paper, we aim to answer the following open questions: can we find winning tickets with few data samples or few labels? can we even obtain good tickets without supervision? Perhaps surprisingly, we provide a positive answer to both, by generating winning tickets with limited access to data, or with self-supervision---thus without using manual annotations---and then demonstrating the transferability of the tickets to challenging classification tasks such as ImageNet.
| Reject |
ICLR.cc/2023/Conference | $\omega$GNNs: Deep Graph Neural Networks Enhanced by Multiple Propagation Operators | Graph Neural Networks (GNNs) are limited in their propagation operators. These operators often contain non-negative elements only and are shared across channels and layers, limiting the expressiveness of GNNs. Moreover, some GNNs suffer from over-smoothing, limiting their depth. On the other hand, Convolutional Neural Networks (CNNs) can learn diverse propagation filters, and phenomena like over-smoothing are typically not apparent in CNNs.
In this paper, we bridge this gap by incorporating trainable channel-wise weighting factors $\omega$ to learn and mix multiple smoothing and sharpening propagation operators at each layer. Our generic method is called $\omega$GNN, and we study two variants: $\omega$GCN and $\omega$GAT.
For $\omega$GCN, we theoretically analyse its behaviour and the impact of $\omega$ on the obtained node features. Our experiments confirm these findings, demonstrating and explaining how both variants do not over-smooth.
Additionally, we experiment with 15 real-world datasets on node- and graph-classification tasks, where our $\omega$GCN and $\omega$GAT perform better or on par with state-of-the-art methods. | Reject |
ICLR.cc/2021/Conference | Hard Masking for Explaining Graph Neural Networks | Graph Neural Networks (GNNs) are a flexible and powerful family of models that build nodes' representations on irregular graph-structured data. This paper focuses on explaining or interpreting the rationale underlying a given prediction of already trained graph neural networks for the node classification task. Existing approaches for interpreting GNNs try to find subsets of important features and nodes by learning a continuous mask. Our objective is to find discrete masks that are arguably more interpretable while minimizing the expected deviation from the underlying model's prediction. We empirically show that our explanations are both more predictive and sparse. Additionally, we find that multiple diverse explanations are possible, which sufficiently explain a prediction. Finally, we analyze the explanations to find the effect of network homophily on the decision-making process of GNNs. | Reject |
ICLR.cc/2021/Conference | Learning Manifold Patch-Based Representations of Man-Made Shapes | Choosing the right representation for geometry is crucial for making 3D models compatible with existing applications. Focusing on piecewise-smooth man-made shapes, we propose a new representation that is usable in conventional CAD modeling pipelines and can also be learned by deep neural networks. We demonstrate its benefits by applying it to the task of sketch-based modeling. Given a raster image, our system infers a set of parametric surfaces that realize the input in 3D. To capture piecewise smooth geometry, we learn a special shape representation: a deformable parametric template composed of Coons patches. Naively training such a system, however, is hampered by non-manifold artifacts in the parametric shapes and by a lack of data. To address this, we introduce loss functions that bias the network to output non-self-intersecting shapes and implement them as part of a fully self-supervised system, automatically generating both shape templates and synthetic training data. We develop a testbed for sketch-based modeling, demonstrate shape interpolation, and provide comparison to related work. | Accept (Poster) |
ICLR.cc/2022/Conference | Graph-Augmented Normalizing Flows for Anomaly Detection of Multiple Time Series | Anomaly detection is a widely studied task for a broad variety of data types; among them, multiple time series appear frequently in applications, including for example, power grids and traffic networks. Detecting anomalies for multiple time series, however, is a challenging subject, owing to the intricate interdependencies among the constituent series. We hypothesize that anomalies occur in low density regions of a distribution and explore the use of normalizing flows for unsupervised anomaly detection, because of their superior quality in density estimation. Moreover, we propose a novel flow model by imposing a Bayesian network among constituent series. A Bayesian network is a directed acyclic graph (DAG) that models causal relationships; it factorizes the joint probability of the series into the product of easy-to-evaluate conditional probabilities. We call such a graph-augmented normalizing flow approach GANF and propose joint estimation of the DAG with flow parameters. We conduct extensive experiments on real-world datasets and demonstrate the effectiveness of GANF for density estimation, anomaly detection, and identification of time series distribution drift. | Accept (Spotlight) |
ICLR.cc/2021/Conference | Flatness is a False Friend | Hessian based measures of flatness, such as the trace, Frobenius and spectral norms, have been argued, used and shown to relate to generalisation. In this paper we demonstrate that, for feed-forward neural networks under the cross-entropy loss, low-loss solutions with large neural network weights have small Hessian based measures of flatness. This implies that solutions obtained without L2 regularisation should be less sharp than those with despite generalising worse. We show this to be true for logistic regression, multi-layer perceptrons, simple convolutional, pre-activated and wide residual networks on the MNIST and CIFAR-$100$ datasets. Furthermore, we show that adaptive optimisation algorithms using iterate averaging, on the VGG-$16$ network and CIFAR-$100$ dataset, achieve superior generalisation to SGD but are $30 \times$ sharper. These theoretical and experimental results further advocate the need to use flatness in conjunction with the weights scale to measure generalisation \citep{neyshabur2017exploring,dziugaite2017computing}. | Reject |
ICLR.cc/2023/Conference | The Role of ImageNet Classes in Fréchet Inception Distance | Fréchet Inception Distance (FID) is the primary metric for ranking models in data-driven generative modeling. While remarkably successful, the metric is known to sometimes disagree with human judgement. We investigate a root cause of these discrepancies, and visualize what FID "looks at" in generated images. We show that the feature space that FID is (typically) computed in is so close to the ImageNet classifications that aligning the histograms of Top-$N$ classifications between sets of generated and real images can reduce FID substantially — without actually improving the quality of results. Thus, we conclude that FID is prone to intentional or accidental distortions. As a practical example of an accidental distortion, we discuss a case where an ImageNet pre-trained FastGAN achieves a FID comparable to StyleGAN2, while being worse in terms of human evaluation. | Accept: notable-top-25% |
ICLR.cc/2020/Conference | Improving Robustness Without Sacrificing Accuracy with Patch Gaussian Augmentation | Deploying machine learning systems in the real world requires both high accuracy on clean data and robustness to naturally occurring corruptions. While architectural advances have led to improved accuracy, building robust models remains challenging, involving major changes in training procedure and datasets. Prior work has argued that there is an inherent trade-off between robustness and accuracy, as exemplified by standard data augmentation techniques such as Cutout, which improves clean accuracy but not robustness, and additive Gaussian noise, which improves robustness but hurts accuracy. We introduce Patch Gaussian, a simple augmentation scheme that adds noise to randomly selected patches in an input image. Models trained with Patch Gaussian achieve state of the art on the CIFAR-10 and ImageNet Common Corruptions benchmarks while also maintaining accuracy on clean data. We find that this augmentation leads to reduced sensitivity to high frequency noise (similar to Gaussian) while retaining the ability to take advantage of relevant high frequency information in the image (similar to Cutout). We show it can be used in conjunction with other regularization methods and data augmentation policies such as AutoAugment. Finally, we find that the idea of restricting perturbations to patches can also be useful in the context of adversarial learning, yielding models without the loss in accuracy that is found with unconstrained adversarial training. | Reject |
ICLR.cc/2021/Conference | Influence Estimation for Generative Adversarial Networks | Identifying harmful instances, whose absence in a training dataset improves model performance, is important for building better machine learning models.
Although previous studies have succeeded in estimating harmful instances under supervised settings, they cannot be trivially extended to generative adversarial networks (GANs).
This is because previous approaches require that (i) the absence of a training instance directly affects the loss value and that (ii) the change in the loss directly measures the harmfulness of the instance for the performance of a model.
In GAN training, however, neither of the requirements is satisfied.
This is because, (i) the generator’s loss is not directly affected by the training instances as they are not part of the generator's training steps, and (ii) the values of GAN's losses normally do not capture the generative performance of a model.
To this end, (i) we propose an influence estimation method that uses the Jacobian of the gradient of the generator's loss with respect to the discriminator’s parameters (and vice versa) to trace how the absence of an instance in the discriminator’s training affects the generator’s parameters, and (ii) we propose a novel evaluation scheme, in which we assess harmfulness of each training instance on the basis of how GAN evaluation metric (e.g., inception score) is expected to change due to the removal of the instance.
We experimentally verified that our influence estimation method correctly inferred the changes in GAN evaluation metrics.
We also demonstrated that the removal of the identified harmful instances effectively improved the model’s generative performance with respect to various GAN evaluation metrics. | Accept (Spotlight) |
ICLR.cc/2020/Conference | Matrix Multilayer Perceptron | Models that output a vector of responses given some inputs, in the form of a conditional mean vector, are at the core of machine learning. This includes neural networks such as the multilayer perceptron (MLP). However, models that output a symmetric positive definite (SPD) matrix of responses given inputs, in the form of a conditional covariance function, are far less studied, especially within the context of neural networks. Here, we introduce a new variant of the MLP, referred to as the matrix MLP, that is specialized at learning SPD matrices. Our construction not only respects the SPD constraint, but also makes explicit use of it. This translates into a model which effectively performs the task of SPD matrix learning even in scenarios where data are scarce. We present an application of the model in heteroscedastic multivariate regression, including convincing performance on six real-world datasets. | Reject |
ICLR.cc/2021/Conference | Image GANs meet Differentiable Rendering for Inverse Graphics and Interpretable 3D Neural Rendering | Differentiable rendering has paved the way to training neural networks to perform “inverse graphics” tasks such as predicting 3D geometry from monocular photographs. To train high performing models, most of the current approaches rely on multi-view imagery which are not readily available in practice. Recent Generative Adversarial Networks (GANs) that synthesize images, in contrast, seem to acquire 3D knowledge implicitly during training: object viewpoints can be manipulated by simply manipulating the latent codes. However, these latent codes often lack further physical interpretation and thus GANs cannot easily be inverted to perform explicit 3D reasoning. In this paper, we aim to extract and disentangle 3D knowledge learned by generative models by utilizing differentiable renderers. Key to our approach is to exploit GANs as a multi-view data generator to train an inverse graphics network using an off-the-shelf differentiable renderer, and the trained inverse graphics network as a teacher to disentangle the GAN's latent code into interpretable 3D properties. The entire architecture is trained iteratively using cycle consistency losses. We show that our approach significantly outperforms state-of-the-art inverse graphics networks trained on existing datasets, both quantitatively and via user studies. We further showcase the disentangled GAN as a controllable 3D “neural renderer", complementing traditional graphics renderers. | Accept (Oral) |
ICLR.cc/2019/Conference | Meta-Learning with Domain Adaptation for Few-Shot Learning under Domain Shift | Few-Shot Learning (learning with limited labeled data) aims to overcome the limitations of traditional machine learning approaches which require thousands of labeled examples to train an effective model. Considered as a hallmark of human intelligence, the community has recently witnessed several contributions on this topic, in particular through meta-learning, where a model learns how to learn an effective model for few-shot learning. The main idea is to acquire prior knowledge from a set of training tasks, which is then used to perform (few-shot) test tasks. Most existing work assumes that both training and test tasks are drawn from the same distribution, and a large amount of labeled data is available in the training tasks. This is a very strong assumption which restricts the usage of meta-learning strategies in the real world where ample training tasks following the same distribution as test tasks may not be available. In this paper, we propose a novel meta-learning paradigm wherein a few-shot learning model is learnt, which simultaneously overcomes domain shift between the train and test tasks via adversarial domain adaptation. We demonstrate the efficacy the proposed method through extensive experiments. | Reject |
ICLR.cc/2023/Conference | Less is More: Rethinking Few-Shot Learning and Recurrent Neural Nets | The statistical supervised learning framework assumes an input-output set with a joint probability distribution that is reliably represented by the training dataset. The learner is then required to output a prediction rule learned from the training dataset's input-output pairs. In this work, we provide meaningful insights into the asymptotic equipartition property (AEP) \citep{Shannon:1948} in the context of machine learning, and illuminate some of its potential ramifications for few-shot learning. We provide theoretical guarantees for reliable learning under the information-theoretic AEP, and for the generalization error with respect to the sample size. We then focus on a highly efficient recurrent neural net (RNN) framework and propose a reduced-entropy algorithm for few-shot learning. We also propose a mathematical intuition for the RNN as an approximation of a sparse coding solver. We verify the applicability, robustness, and computational efficiency of the proposed approach with image deblurring and optical coherence tomography (OCT) speckle suppression. Our experimental results demonstrate significant potential for improving learning models' sample efficiency, generalization, and time complexity, that can therefore be leveraged for practical real-time applications. | Reject |
ICLR.cc/2022/Conference | Logarithmic landscape and power-law escape rate of SGD | Stochastic gradient descent (SGD) undergoes complicated multiplicative noise for the mean-square loss. We use this property of the SGD noise to derive a stochastic differential equation (SDE) with simpler additive noise by performing a random time change. In the SDE, the loss gradient is replaced by the logarithmized loss gradient. By using this formalism, we obtain the escape rate formula from a local minimum, which is determined not by the loss barrier height $\Delta L=L(\theta^s)-L(\theta^*)$ between a minimum $\theta^*$ and a saddle $\theta^s$ but by the logarithmized loss barrier height $\Delta\log L=\log[L(\theta^s)/L(\theta^*)]$. Our escape-rate formula strongly depends on the typical magnitude $h^*$ and the number $n$ of the outlier eigenvalues of the Hessian. This result explains an empirical fact that SGD prefers flat minima with low effective dimensions, which gives an insight into implicit biases of SGD. | Reject |
ICLR.cc/2021/Conference | Improving the Unsupervised Disentangled Representation Learning with VAE Ensemble | Variational Autoencoder (VAE) based frameworks have achieved the state-of-the-art performance on the unsupervised disentangled representation learning. A recent theoretical analysis shows that such success is mainly due to the VAE implementation choices that encourage a PCA-like behavior locally on data samples. Despite this implied model identifiability, the VAE based disentanglement frameworks still face the trade-off between the local orthogonality and data reconstruction. As a result, models with the same architecture and hyperparameter setting can sometime learn entangled representations. To address this challenge, we propose a simple yet effective VAE ensemble framework consisting of multiple VAEs. It is based on the assumption that entangled representations are unique in their own ways, and the disentangled representations are "alike" (similar up to a signed permutation transformation). In the proposed VAE ensemble, each model not only maintains its original objective, but also encodes to and decodes from other models through pair-wise linear transformations between the latent representations. We show both theoretically and experimentally, the VAE ensemble objective encourages the linear transformations connecting the VAEs to be trivial transformations, aligning the latent representations of different models to be "alike". We compare our approach with the state-of-the-art unsupervised disentangled representation learning approaches and show the improved performance. | Reject |
ICLR.cc/2018/Conference | Achieving Strong Regularization for Deep Neural Networks | L1 and L2 regularizers are critical tools in machine learning due to their ability to simplify solutions. However, imposing strong L1 or L2 regularization with gradient descent method easily fails, and this limits the generalization ability of the underlying neural networks. To understand this phenomenon, we investigate how and why training fails for strong regularization. Specifically, we examine how gradients change over time for different regularization strengths and provide an analysis why the gradients diminish so fast. We find that there exists a tolerance level of regularization strength, where the learning completely fails if the regularization strength goes beyond it. We propose a simple but novel method, Delayed Strong Regularization, in order to moderate the tolerance level. Experiment results show that our proposed approach indeed achieves strong regularization for both L1 and L2 regularizers and improves both accuracy and sparsity on public data sets. Our source code is published. | Reject |
ICLR.cc/2019/Conference | Area Attention | Existing attention mechanisms, are mostly item-based in that a model is trained to attend to individual items in a collection (the memory) where each item has a predefined, fixed granularity, e.g., a character or a word. Intuitively, an area in the memory consisting of multiple items can be worth attending to as a whole. We propose area attention: a way to attend to an area of the memory, where each area contains a group of items that are either spatially adjacent when the memory has a 2-dimensional structure, such as images, or temporally adjacent for 1-dimensional memory, such as natural language sentences. Importantly, the size of an area, i.e., the number of items in an area or the level of aggregation, is dynamically determined via learning, which can vary depending on the learned coherence of the adjacent items. By giving the model the option to attend to an area of items, instead of only individual items, a model can attend to information with varying granularity. Area attention can work along multi-head attention for attending to multiple areas in the memory. We evaluate area attention on two tasks: neural machine translation (both character and token-level) and image captioning, and improve upon strong (state-of-the-art) baselines in all the cases. These improvements are obtainable with a basic form of area attention that is parameter free. In addition to proposing the novel concept of area attention, we contribute an efficient way for computing it by leveraging the technique of summed area tables. | Reject |
ICLR.cc/2023/Conference | A UNIFIED VIEW OF FINDING AND TRANSFORMING WINNING LOTTERY TICKETS | While over-parameterized deep neural networks obtain prominent results on various machine learning tasks, their superfluous parameters usually make model training and inference notoriously inefficient. Lottery Ticket Hypothesis (LTH) addresses this issue from a novel perspective: it articulates that there always exist sparse and admirable subnetworks in a randomly initialized dense network, which can be realized by an iterative pruning strategy. Dual Lottery Ticket Hypothesis (DLTH) further investigates sparse network training from a complementary view. Concretely, it introduces a gradually increased regularization term to transform a dense network to an ultra-light subnetwork without sacrificing learning capacity. After revisiting the success of LTH and DLTH, we unify these two research lines by coupling the stability of iterative pruning and the excellent performance of increased regularization, resulting in two new algorithms (UniLTH and UniDLTH) for finding and transforming winning tickets, respectively. Unlike either LTH without regularization or DLTH which applies regularization across the training, our methods first train the network without any regularization force until the model reaches a certain point (i.e., the validation loss does not decrease for several epochs), and then employ increased regularization for information extrusion and iteratively perform magnitude pruning till the end. We theoretically prove that the early stopping mechanism acts analogously as regularization and can help the optimization trajectory stop at a particularly better point in space than regularization. This not only prevent the parameters from being excessively skewed to the training distribution (over-fitting), but also better stimulate the network potential to obtain more powerful subnetworks. Extensive experiments are conducted to show the superiority of our methods in terms of accuracy and sparsity. | Reject |
ICLR.cc/2019/Conference | Correction Networks: Meta-Learning for Zero-Shot Learning | We propose a model that learns to perform zero-shot classification using a meta-learner that is trained to produce a correction to the output of a previously trained learner. The model consists of two modules: a task module that supplies an initial prediction, and a correction module that updates the initial prediction. The task module is the learner and the correction module is the meta-learner. The correction module is trained in an episodic approach whereby many different task modules are trained on various subsets of the total training data, with the rest being used as unseen data for the correction module. The correction module takes as input a representation of the task module's training data so that the predicted correction is a function of the task module's training data. The correction module is trained to update the task module's prediction to be closer to the target value. This approach leads to state-of-the-art performance for zero-shot classification on natural language class descriptions on the CUB and NAB datasets. | Reject |
ICLR.cc/2019/Conference | Differentiable Perturb-and-Parse: Semi-Supervised Parsing with a Structured Variational Autoencoder | Human annotation for syntactic parsing is expensive, and large resources are available only for a fraction of languages. A question we ask is whether one can leverage abundant unlabeled texts to improve syntactic parsers, beyond just using the texts to obtain more generalisable lexical features (i.e. beyond word embeddings). To this end, we propose a novel latent-variable generative model for semi-supervised syntactic dependency parsing. As exact inference is intractable, we introduce a differentiable relaxation to obtain approximate samples and compute gradients with respect to the parser parameters. Our method (Differentiable Perturb-and-Parse) relies on differentiable dynamic programming over stochastically perturbed edge scores. We demonstrate effectiveness of our approach with experiments on English, French and Swedish. | Accept (Poster) |
ICLR.cc/2022/Conference | Learning to Complete Code with Sketches | Code completion is usually cast as a language modelling problem, i.e., continuing an input in a left-to-right fashion. However, in practice, some parts of the completion (e.g., string literals) may be very hard to predict, whereas subsequent parts directly follow from the context.
To handle this, we instead consider the scenario of generating code completions with "holes" inserted in places where a model is uncertain. We develop Grammformer, a Transformer-based model that guides the code generation by the programming language grammar, and compare it to a variety of more standard sequence models.
We train the models on code completion for C# and Python given partial code context. To evaluate models, we consider both ROUGE as well as a new metric RegexAcc that measures success of generating completions matching long outputs with as few holes as possible.
In our experiments, Grammformer generates 10-50% more accurate completions compared to traditional generative models and 37-50% longer sketches compared to sketch-generating baselines trained with similar techniques. | Accept (Poster) |
ICLR.cc/2023/Conference | HyperDeepONet: learning operator with complex target function space using the limited resources via hypernetwork | Fast and accurate predictions for complex physical dynamics are a big challenge across various applications. Real-time prediction on resource-constrained hardware is even more crucial in the real-world problems. The deep operator network (DeepONet) has recently been proposed as a framework for learning nonlinear mappings between function spaces. However, the DeepONet requires many parameters and has a high computational cost when learning operators, particularly those with complex (discontinuous or non-smooth) target functions. In this study, we propose HyperDeepONet, which uses the expressive power of the hypernetwork to enable learning of a complex operator with smaller set of parameters. The DeepONet and its variant models can be thought of as a method of injecting the input function information into the target function. From this perspective, these models can be viewed as a special case of HyperDeepONet. We analyze the complexity of DeepONet and conclude that HyperDeepONet needs relatively lower complexity to obtain the desired accuracy for operator learning. HyperDeepONet was successfully applied to various operator learning problems using low computational resources compared to other benchmarks. | Accept: poster |
ICLR.cc/2023/Conference | On the Importance and Applicability of Pre-Training for Federated Learning | Pre-training is prevalent in nowadays deep learning to improve the learned model's performance. However, in the literature on federated learning (FL), neural networks are mostly initialized with random weights. These attract our interest in conducting a systematic study to explore pre-training for FL. Across multiple visual recognition benchmarks, we found that pre-training can not only improve FL, but also close its accuracy gap to the counterpart centralized learning, especially in the challenging cases of non-IID clients' data. To make our findings applicable to situations where pre-trained models are not directly available, we explore pre-training with synthetic data or even with clients' data in a decentralized manner, and found that they can already improve FL notably. Interestingly, many of the techniques we explore are complementary to each other to further boost the performance, and we view this as a critical result toward scaling up deep FL for real-world applications. We conclude our paper with an attempt to understand the effect of pre-training on FL. We found that pre-training enables the learned global models under different clients' data conditions to converge to the same loss basin, and makes global aggregation in FL more stable. Nevertheless, pre-training seems to not alleviate local model drifting, a fundamental problem in FL under non-IID data. | Accept: poster |
ICLR.cc/2020/Conference | Decoupling Weight Regularization from Batch Size for Model Compression | Conventionally, compression-aware training performs weight compression for every mini-batch to compute the impact of compression on the loss function. In this paper, in order to study when would be the right time to compress weights during optimization steps, we propose a new hyper-parameter called Non-Regularization period or NR period during which weights are not updated for regularization. We first investigate the influence of NR period on regularization using weight decay and weight random noise insertion. Throughout various experiments, we show that stronger weight regularization demands longer NR period (regardless of batch size) to best utilize regularization effects. From our empirical evidence, we argue that weight regularization for every mini-batch allows small weight updates only and limited regularization effects such that there is a need to search for right NR period and weight regularization strength to enhance model accuracy. Consequently, NR period becomes especially crucial for model compression where large weight updates are necessary to increase compression ratio. Using various models, we show that simple weight updates to comply with compression formats along with long NR period is enough to achieve high compression ratio and model accuracy. | Reject |
ICLR.cc/2023/Conference | Diagnosing and exploiting the computational demands of videos games for deep reinforcement learning | Humans learn by interacting with their environments and perceiving the outcomes of their actions. A landmark in artificial intelligence has been the development of deep reinforcement learning (dRL) algorithms capable of doing the same in video games, on par with or better than humans. However, it remains unclear whether the successes of dRL models reflect advances in visual representation learning, the effectiveness of reinforcement learning algorithms at discovering better policies, or both. To address this question, we introduce the Learning Challenge Diagnosticator (LCD), a tool that separately measures the perceptual and reinforcement learning demands of a task. We use LCD to discover a novel taxonomy of challenges in the Procgen benchmark, and demonstrate that these predictions are both highly reliable and can instruct algorithmic development. More broadly, the LCD reveals multiple failure cases that can occur when optimizing dRL algorithms over entire video game benchmarks like Procgen, and provides a pathway towards more efficient progress. | Reject |
ICLR.cc/2020/Conference | Stochastic Neural Physics Predictor | Recently, neural-network based forward dynamics models have been proposed that attempt to learn the dynamics of physical systems in a deterministic way. While near-term motion can be predicted accurately, long-term predictions suffer from accumulating input and prediction errors which can lead to plausible but different trajectories that diverge from the ground truth. A system that predicts distributions of the future physical states for long time horizons based on its uncertainty is thus a promising solution. In this work, we introduce a novel robust Monte Carlo sampling based graph-convolutional dropout method that allows us to sample multiple plausible trajectories for an initial state given a neural-network based forward dynamics predictor. By introducing a new shape preservation loss and training our dynamics model recurrently, we stabilize long-term predictions. We show that our model’s long-term forward dynamics prediction errors on complicated physical interactions of rigid and deformable objects of various shapes are significantly lower than existing strong baselines. Lastly, we demonstrate how generating multiple trajectories with our Monte Carlo dropout method can be used to train model-free reinforcement learning agents faster and to better solutions on simple manipulation tasks. | Reject |
ICLR.cc/2023/Conference | A Curriculum Perspective to Robust Loss Functions | Learning with noisy labels is a fundamental problem in machine learning. Much work has been done in designing loss functions that are theoretically robust against label noise. However, it remains unclear why robust loss functions can underfit and why loss functions deviating from theoretical robustness conditions can appear robust. To elucidate these questions, we show that most robust loss functions differ only in the sample-weighting curriculums they implicitly define. The curriculum perspective enables straightforward analysis of the training dynamics with each loss function, which has not been considered in existing theoretical approaches. We show that underfitting can be attributed to marginal sample weights during training, and noise robustness can be attributed to larger weights for clean samples than noisy samples. With a simple fix to the curriculums, robust loss functions that severely underfit can become competitive with the state-of-the-art. | Reject |
ICLR.cc/2021/Conference | Wasserstein diffusion on graphs with missing attributes | Many real-world graphs are attributed graphs where nodes are associated with non-topological features. While attributes can be missing anywhere in an attributed graph, most of existing node representation learning approaches do not consider such incomplete information.
In this paper, we propose a general non-parametric framework to mitigate this problem. Starting from a decomposition of the attribute matrix, we transform node features into discrete distributions in a lower-dimensional space equipped with the Wasserstein metric. On this Wasserstein space, we propose Wasserstein graph diffusion to smooth the distributional representations of nodes with information from their local neighborhoods. This allows us to reduce the distortion caused by missing attributes and obtain integrated representations expressing information of both topology structures and attributes. We then pull the nodes back to the original space and produce corresponding point representations to facilitate various downstream tasks. To show the power of our representation method, we designed two algorithms based on it for node classification (with missing attributes) and matrix completion respectively, and demonstrate their effectiveness in experiments. | Reject |
ICLR.cc/2023/Conference | Amortised Invariance Learning for Contrastive Self-Supervision | Contrastive self-supervised learning methods famously produce high quality transferable representations by learning invariances to different data augmentations. Invariances established during pre-training can be interpreted as strong inductive biases. However these may or may not be helpful, depending on if they match the invariance requirements of downstream tasks or not. This has led to several attempts to learn task-specific invariances during pre-training, however, these methods are highly compute intensive and tedious to train. We introduce the notion of amortized invariance learning for contrastive self supervision. In the pre-training stage, we parameterize the feature extractor by differentiable invariance hyper-parameters that control the invariances encoded by the representation. Then, for any downstream task, both linear readout and task-specific invariance requirements can be efficiently and effectively learned by gradient-descent. We evaluate the notion of amortized invariances for contrastive learning over two different modalities: vision and audio, on two widely-used contrastive learning methods in vision: SimCLR and MoCo-v2 with popular architectures like ResNets and Vision Transformers, and SimCLR with ResNet-18 for audio. We show that our amortized features provide a reliable way to learn diverse downstream tasks with different invariance requirements, while using a single feature and avoiding task-specific pre-training. This provides an exciting perspective that opens up new horizons in the field of general purpose representation learning. | Accept: poster |
ICLR.cc/2020/Conference | Data augmentation instead of explicit regularization | Modern deep artificial neural networks have achieved impressive results through models with orders of magnitude more parameters than training examples which control overfitting with the help of regularization. Regularization can be implicit, as is the case of stochastic gradient descent and parameter sharing in convolutional layers, or explicit. Explicit regularization techniques, most common forms are weight decay and dropout, have proven successful in terms of improved generalization, but they blindly reduce the effective capacity of the model, introduce sensitive hyper-parameters and require deeper and wider architectures to compensate for the reduced capacity. In contrast, data augmentation techniques exploit domain knowledge to increase the number of training examples and improve generalization without reducing the effective capacity and without introducing model-dependent parameters, since it is applied on the training data. In this paper we systematically contrast data augmentation and explicit regularization on three popular architectures and three data sets. Our results demonstrate that data augmentation alone can achieve the same performance or higher as regularized models and exhibits much higher adaptability to changes in the architecture and the amount of training data. | Reject |
ICLR.cc/2023/Conference | Humanly Certifying Superhuman Classifiers | This paper addresses a key question in current machine learning research: if we believe that a model's predictions might be better than those given by human experts, how can we (humans) verify these beliefs? In some cases, this ``superhuman'' performance is readily demonstrated; for example by defeating top-tier human players in traditional two player games. On the other hand, it can be challenging to evaluate classification models that potentially surpass human performance. Indeed, human annotations are often treated as a ground truth, which implicitly assumes the superiority of the human over any models trained on human annotations. In reality, human annotators are subjective and can make mistakes. Evaluating the performance with respect to a genuine oracle is more objective and reliable, even when querying the oracle is more expensive or sometimes impossible. In this paper, we first raise the challenge of evaluating the performance of both humans and models with respect to an oracle which is $\textit{unobserved}$. We develop a theory for estimating the accuracy compared to the oracle, using only imperfect human annotations for reference. Our analysis provides an executable recipe for detecting and certifying superhuman performance in this setting, which we believe will assist in understanding the stage of current research on classification. We validate the convergence of the bounds and the assumptions of our theory on carefully designed toy experiments with known oracles. Moreover, we demonstrate the utility of our theory by meta-analyzing large-scale natural language processing tasks, for which an oracle does not exist, and show that under our mild assumptions a number of models from recent years have already achieved superhuman performance with high probability---suggesting that our new oracle based performance evaluation metrics are overdue as an alternative to the widely used accuracy metrics that are naively based on imperfect human annotations. | Accept: notable-top-25% |
ICLR.cc/2019/Conference | Learning Multi-Level Hierarchies with Hindsight | Hierarchical agents have the potential to solve sequential decision making tasks with greater sample efficiency than their non-hierarchical counterparts because hierarchical agents can break down tasks into sets of subtasks that only require short sequences of decisions. In order to realize this potential of faster learning, hierarchical agents need to be able to learn their multiple levels of policies in parallel so these simpler subproblems can be solved simultaneously. Yet, learning multiple levels of policies in parallel is hard because it is inherently unstable: changes in a policy at one level of the hierarchy may cause changes in the transition and reward functions at higher levels in the hierarchy, making it difficult to jointly learn multiple levels of policies. In this paper, we introduce a new Hierarchical Reinforcement Learning (HRL) framework, Hierarchical Actor-Critic (HAC), that can overcome the instability issues that arise when agents try to jointly learn multiple levels of policies. The main idea behind HAC is to train each level of the hierarchy independently of the lower levels by training each level as if the lower level policies are already optimal. We demonstrate experimentally in both grid world and simulated robotics domains that our approach can significantly accelerate learning relative to other non-hierarchical and hierarchical methods. Indeed, our framework is the first to successfully learn 3-level hierarchies in parallel in tasks with continuous state and action spaces. | Accept (Poster) |
ICLR.cc/2019/Conference | Learning to Augment Influential Data | Data augmentation is a technique to reduce overfitting and to improve generalization by increasing the number of labeled data samples by performing label preserving transformations; however, it is currently conducted in a trial and error manner. A composition of predefined transformations, such as rotation, scaling and cropping, is performed on training samples, and its effect on performance over test samples can only be empirically evaluated and cannot be predicted. This paper considers an influence function which predicts how generalization is affected by a particular augmented training sample in terms of validation loss. The influence function provides an approximation of the change in validation loss without comparing the performance which includes and excludes the sample in the training process. A differentiable augmentation model that generalizes the conventional composition of predefined transformations is also proposed. The differentiable augmentation model and reformulation of the influence function allow the parameters of the augmented model to be directly updated by backpropagation to minimize the validation loss. The experimental results show that the proposed method provides better generalization over conventional data augmentation methods. | Reject |
ICLR.cc/2020/Conference | Prediction Poisoning: Towards Defenses Against DNN Model Stealing Attacks | High-performance Deep Neural Networks (DNNs) are increasingly deployed in many real-world applications e.g., cloud prediction APIs. Recent advances in model functionality stealing attacks via black-box access (i.e., inputs in, predictions out) threaten the business model of such applications, which require a lot of time, money, and effort to develop. Existing defenses take a passive role against stealing attacks, such as by truncating predicted information. We find such passive defenses ineffective against DNN stealing attacks. In this paper, we propose the first defense which actively perturbs predictions targeted at poisoning the training objective of the attacker. We find our defense effective across a wide range of challenging datasets and DNN model stealing attacks, and additionally outperforms existing defenses. Our defense is the first that can withstand highly accurate model stealing attacks for tens of thousands of queries, amplifying the attacker's error rate up to a factor of 85$\times$ with minimal impact on the utility for benign users. | Accept (Poster) |
ICLR.cc/2019/Conference | FAST OBJECT LOCALIZATION VIA SENSITIVITY ANALYSIS | Deep Convolutional Neural Networks (CNNs) have been repeatedly shown to perform well on image classification tasks, successfully recognizing a broad array of objects when given sufficient training data. Methods for object localization, however, are still in need of substantial improvement. Common approaches to this problem involve the use of a sliding window, sometimes at multiple scales, providing input to a deep CNN trained to classify the contents of the window. In general, these approaches are time consuming, requiring many classification calculations. In this paper, we offer a fundamentally different approach to the localization of recognized objects in images. Our method is predicated on the idea that a deep CNN capable of recognizing an object must implicitly contain knowledge about object location in its connection weights. We provide a simple method to interpret classifier weights in the context of individual classified images. This method involves the calculation of the derivative of network generated activation patterns, such as the activation of output class label units, with regard to each in- put pixel, performing a sensitivity analysis that identifies the pixels that, in a local sense, have the greatest influence on internal representations and object recognition. These derivatives can be efficiently computed using a single backward pass through the deep CNN classifier, producing a sensitivity map of the image. We demonstrate that a simple linear mapping can be learned from sensitivity maps to bounding box coordinates, localizing the recognized object. Our experimental results, using real-world data sets for which ground truth localization information is known, reveal competitive accuracy from our fast technique. | Reject |
ICLR.cc/2021/Conference | Low Complexity Approximate Bayesian Logistic Regression for Sparse Online Learning | Theoretical results show that Bayesian methods can achieve lower bounds on regret for online logistic regression. In practice, however, such techniques may not be feasible especially for very large feature sets. Various approximations that, for huge sparse feature sets, diminish the theoretical advantages, must be used. Often, they apply stochastic gradient methods with hyper-parameters that must be tuned on some surrogate loss, defeating theoretical advantages of Bayesian methods. The surrogate loss, defined to approximate the mixture, requires techniques as Monte Carlo sampling, increasing computations per example. We propose low complexity analytical approximations for sparse online logistic and probit regressions. Unlike variational inference and other methods, our methods use analytical closed forms, substantially lowering computations. Unlike dense solutions,
as Gaussian Mixtures, our methods allow for sparse problems with huge feature sets without increasing complexity. With the analytical closed forms, there is also no need for applying stochastic gradient methods on surrogate losses, and for tuning and balancing learning and regularization hyper-parameters. Empirical results top the performance of the more computationally involved methods. Like such methods, our methods still reveal per feature and per example uncertainty measures.
| Reject |
ICLR.cc/2022/Conference | Task-Induced Representation Learning | In this work, we evaluate the effectiveness of representation learning approaches for decision making in visually complex environments. Representation learning is essential for effective reinforcement learning (RL) from high-dimensional in- puts. Unsupervised representation learning approaches based on reconstruction, prediction or contrastive learning have shown substantial learning efficiency gains. Yet, they have mostly been evaluated in clean laboratory or simulated settings. In contrast, real environments are visually complex and contain substantial amounts of clutter and distractors. Unsupervised representations will learn to model such distractors, potentially impairing the agent’s learning efficiency. In contrast, an alternative class of approaches, which we call task-induced representation learning, leverages task information such as rewards or demonstrations from prior tasks to focus on task-relevant parts of the scene and ignore distractors. We investi- gate the effectiveness of unsupervised and task-induced representation learning approaches on four visually complex environments, from Distracting DMControl to the CARLA driving simulator. For both, RL and imitation learning, we find that representation learning generally improves sample efficiency on unseen tasks even in visually complex scenes and that task-induced representations can double learning efficiency compared to unsupervised alternatives. | Accept (Poster) |
ICLR.cc/2018/Conference | Decoupling the Layers in Residual Networks | We propose a Warped Residual Network (WarpNet) using a parallelizable warp operator for forward and backward propagation to distant layers that trains faster than the original residual neural network. We apply a perturbation theory on residual networks and decouple the interactions between residual units. The resulting warp operator is a first order approximation of the output over multiple layers. The first order perturbation theory exhibits properties such as binomial path lengths and exponential gradient scaling found experimentally by Veit et al (2016).
We demonstrate through an extensive performance study that the proposed network achieves comparable predictive performance to the original residual network with the same number of parameters, while achieving a significant speed-up on the total training time. As WarpNet performs model parallelism in residual network training in which weights are distributed over different GPUs, it offers speed-up and capability to train larger networks compared to original residual networks. | Accept (Poster) |
ICLR.cc/2021/Conference | Learning Private Representations with Focal Entropy | How can we learn a representation with good predictive power while preserving user privacy?
We present an adversarial representation learning method to sanitize sensitive content from the representation in an adversarial fashion.
Specifically, we propose focal entropy - a variant of entropy embedded in an adversarial representation learning setting to leverage privacy sanitization. Focal entropy enforces maximum uncertainty in terms of confusion on the subset of privacy-related similar classes, separated from the dissimilar ones. As such, our proposed sanitization method yields deep sanitization of private features yet is conceptually simple and empirically powerful. We showcase feasibility in terms of classification of facial attributes and identity on the CelebA dataset as well as CIFAR-100. The results suggest that private components can be removed reliably. | Reject |
ICLR.cc/2019/Conference | Exploring Curvature Noise in Large-Batch Stochastic Optimization | Using stochastic gradient descent (SGD) with large batch-sizes to train deep neural networks is an increasingly popular technique. By doing so, one can improve parallelization by scaling to multiple workers (GPUs) and hence leading to significant reductions in training time. Unfortunately, a major drawback is the so-called generalization gap: large-batch training typically leads to a degradation in generalization performance of the model as compared to small-batch training. In this paper, we propose to correct this generalization gap by adding diagonal Fisher curvature noise to large-batch gradient updates. We provide a theoretical analysis of our method in the convex quadratic setting. Our empirical study with state-of-the-art deep learning models shows that our method not only improves the generalization performance in large-batch training but furthermore, does so in a way where the training convergence remains desirable and the training duration is not elongated. We additionally connect our method to recent works on loss surface landscape in the experimental section. | Reject |
ICLR.cc/2022/Conference | Learnability Lock: Authorized Learnability Control Through Adversarial Invertible Transformations | Owing much to the revolution of information technology, recent progress of deep learning benefits incredibly from the vastly enhanced access to data available in various digital formats. Yet those publicly accessible information also raises a fundamental issue concerning Intellectual Property, that is, how to precisely control legal or illegal exploitation of a dataset for training commercial models. To tackle this issue, this paper introduces and investigates a new concept called ''learnability lock'' for securing the process of data authorization. In particular, we propose adversarial invertible transformation, that can be viewed as a mapping from image to image, to encrypt data samples so that they become ''unlearnable'' by machine learning models with negligible loss of visual features. Meanwhile, authorized clients can use a specific key to unlock the learnability of the protected dataset and train models normally. The proposed learnability lock leverages class-wise perturbation that applies a universal transformation function on data samples of the same label. This ensures that the learnability can be easily restored with a simple inverse transformation while remaining difficult to be detected or reverse-engineered. We empirically demonstrate the success and practicability of our method on visual classification tasks. | Accept (Poster) |
ICLR.cc/2023/Conference | FedFA: Federated Feature Augmentation | Federated learning is a distributed paradigm that allows multiple parties to collaboratively train deep models without exchanging the raw data. However, the data distribution among clients is naturally non-i.i.d., which leads to severe degradation of the learnt model. The primary goal of this paper is to develop a robust federated learning algorithm to address feature shift in clients’ samples, which can be caused by various factors, e.g., acquisition differences in medical imaging. To reach this goal, we propose FedFA to tackle federated learning from a dis- tinct perspective of federated feature augmentation. FedFA is based on a major insight that each client’s data distribution can be characterized by statistics (i.e., mean and standard deviation) of latent features; and it is likely to manipulate these local statistics globally, i.e., based on information in the entire federation, to let clients have a better sense of the underlying distribution and therefore alleviate local data bias. Based on this insight, we propose to augment each local feature statistic probabilistically based on a normal distribution, whose mean is the original statistic and variance quantifies the augmentation scope. Key to our approach is the determination of a meaningful Gaussian variance, which is accomplished by taking into account not only biased data of each individual client, but also underlying feature statistics characterized by all participating clients. We offer both theoretical and empirical justifications to verify the effectiveness of FedFA. Our code is available at https://github.com/tfzhou/FedFA.
| Accept: poster |
ICLR.cc/2022/Conference | RL-DARTS: Differentiable Architecture Search for Reinforcement Learning | Recently, Differentiable Architecture Search (DARTS) has become one of the most popular Neural Architecture Search (NAS) methods successfully applied in supervised learning (SL). However, its applications in other domains, in particular for reinforcement learning (RL), has seldom been studied. This is due in part to RL possessing a significantly different optimization paradigm than SL, especially with regards to the notion of replay data, which is continually generated via inference in RL. In this paper, we introduce RL-DARTS, one of the first applications of end-to-end DARTS in RL to search for convolutional cells, applied to the challenging, infinitely procedurally generated Procgen benchmark. We demonstrate that the benefits of DARTS become amplified when applied to RL, namely search efficiency in terms of time and compute, as well as simplicity in integration with complex preexisting RL code via simply replacing the image encoder with a DARTS supernet, compatible with both off-policy and on-policy RL algorithms. At the same time however, we provide one of the first extensive studies of DARTS outside of the standard fixed dataset setting in SL via RL-DARTS. We show that throughout training, the supernet gradually learns better cells, leading to alternative architectures which can be highly competitive against manually designed policies, but also verify previous design choices for RL policies. | Reject |
ICLR.cc/2021/Conference | Neural Topic Model via Optimal Transport | Recently, Neural Topic Models (NTMs) inspired by variational autoencoders have obtained increasingly research interest due to their promising results on text analysis. However, it is usually hard for existing NTMs to achieve good document representation and coherent/diverse topics at the same time. Moreover, they often degrade their performance severely on short documents. The requirement of reparameterisation could also comprise their training quality and model flexibility. To address these shortcomings, we present a new neural topic model via the theory of optimal transport (OT). Specifically, we propose to learn the topic distribution of a document by directly minimising its OT distance to the document's word distributions. Importantly, the cost matrix of the OT distance models the weights between topics and words, which is constructed by the distances between topics and words in an embedding space. Our proposed model can be trained efficiently with a differentiable loss. Extensive experiments show that our framework significantly outperforms the state-of-the-art NTMs on discovering more coherent and diverse topics and deriving better document representations for both regular and short texts. | Accept (Spotlight) |
ICLR.cc/2022/Conference | Bayesian Relational Generative Model for Scalable Multi-modal Learning | The study of complex systems requires the integration of multiple heterogeneous and high-dimensional data types (e.g. multi-omics). However, previous generative approaches for multi-modal inputs suffer from two shortcomings. First, they are not stochastic processes, leading to poor uncertainty estimations over their predictions. This is mostly due to the computationally intensive nature of traditional stochastic processes, such as Gaussian Processes (GPs), that makes their applicability limited in multi-modal learning frameworks. Second, they are not able to effectively approximate the joint posterior distribution of multi-modal data types with various missing patterns. More precisely, their model assumptions result in miscalibrated precisions and/or computational cost of sub-sampling procedure. In this paper, we propose a class of stochastic processes that learns a graph of dependencies between samples across multi-modal data types through adopting priors over the relational structure of the given data modalities. The dependency graph in our method, multi-modal Relational Neural Process (mRNP), not only posits distributions over the functions and naturally enables rapid adaptation to new observations by its predictive distribution, but also makes mRNP scalable to large datasets through mini-batch optimization. We also introduce mixture-of-graphs (MoG) in our model construction and show that it can address the aforementioned limitations in joint posterior approximation. Experiments on both toy regression and classification tasks using real-world datasets demonstrate the potential of mRNP for offering higher prediction accuracies as well as more robust uncertainty estimates compared to existing baselines and state-of-the-art methods. | Reject |
ICLR.cc/2020/Conference | Learning Cluster Structured Sparsity by Reweighting | Recently, the paradigm of unfolding iterative algorithms into finite-length feed-forward neural networks has achieved a great success in the area of sparse recovery. Benefit from available training data, the learned networks have achieved state-of-the-art performance in respect of both speed and accuracy. However, the structure behind sparsity, imposing constraint on the support of sparse signals, is often an essential prior knowledge but seldom considered in the existing networks. In this paper, we aim at bridging this gap. Specifically, exploiting the iterative reweighted $\ell_1$ minimization (IRL1) algorithm, we propose to learn the cluster structured sparsity (CSS) by rewegihting adaptively. In particular, we first unfold the Reweighted Iterative Shrinkage Algorithm (RwISTA) into an end-to-end trainable deep architecture termed as RW-LISTA. Then instead of the element-wise reweighting, the global and local reweighting manner are proposed for the cluster structured sparse learning. Numerical experiments further show the superiority of our algorithm against both classical algorithms and learning-based networks on different tasks. | Reject |
ICLR.cc/2023/Conference | Handling Covariate Shifts in Federated Learning with Generalization Guarantees | Covariate shift across clients is a major challenge for federated learning (FL). This work studies the generalization properties of FL under intra-client and inter-client covariate shifts. To this end, we propose Federated Importance-weighteD Empirical risk Minimization (FIDEM) to optimize a global FL model, along with new variants of density ratio matching methods, aiming to handle covariate shifts. These methods trade off some level of privacy for improving the overall generalization performance. We theoretically show that FIDEM achieves smaller generalization error than classical empirical risk minimization under some certain settings. Experimental results demonstrate the superiority of FIDEM over federated averaging (McMahan et al., 2017) and other baselines, which would open the door to study FL under distribution shifts more systematically.
| Reject |
ICLR.cc/2021/Conference | A Sharp Analysis of Model-based Reinforcement Learning with Self-Play | Model-based algorithms---algorithms that explore the environment through building and utilizing an estimated model---are widely used in reinforcement learning practice and theoretically shown to achieve optimal sample efficiency for single-agent reinforcement learning in Markov Decision Processes (MDPs). However, for multi-agent reinforcement learning in Markov games, the current best known sample complexity for model-based algorithms is rather suboptimal and compares unfavorably against recent model-free approaches. In this paper, we present a sharp analysis of model-based self-play algorithms for multi-agent Markov games. We design an algorithm \emph{Optimistic Nash Value Iteration} (Nash-VI) for two-player zero-sum Markov games that is able to output an $\epsilon$-approximate Nash policy in $\tilde{\mathcal{O}}(H^3SAB/\epsilon^2)$ episodes of game playing, where $S$ is the number of states, $A,B$ are the number of actions for the two players respectively, and $H$ is the horizon length. This significantly improves over the best known model-based guarantee of $\tilde{\mathcal{O}}(H^4S^2AB/\epsilon^2)$, and is the first that matches the information-theoretic lower bound $\Omega(H^3S(A+B)/\epsilon^2)$ except for a $\min\{A,B\}$ factor. In addition, our guarantee compares favorably against the best known model-free algorithm if $\min\{A,B\}=o(H^3)$, and outputs a single Markov policy while existing sample-efficient model-free algorithms output a nested mixture of Markov policies that is in general non-Markov and rather inconvenient to store and execute. We further adapt our analysis to designing a provably efficient task-agnostic algorithm for zero-sum Markov games, and designing the first line of provably sample-efficient algorithms for multi-player general-sum Markov games. | Reject |
ICLR.cc/2023/Conference | Proactive Multi-Camera Collaboration for 3D Human Pose Estimation | This paper presents a multi-agent reinforcement learning (MARL) scheme for proactive Multi-Camera Collaboration in 3D Human Pose Estimation in dynamic human crowds. Traditional fixed-viewpoint multi-camera solutions for human motion capture (MoCap) are limited in capture space and susceptible to dynamic occlusions. Active camera approaches proactively control camera poses to find optimal viewpoints for 3D reconstruction. However, current methods still face challenges with credit assignment and environment dynamics. To address these issues, our proposed method introduces a novel Collaborative Triangulation Contribution Reward (CTCR) that improves convergence and alleviates multi-agent credit assignment issues resulting from using 3D reconstruction accuracy as the shared reward. Additionally, we jointly train our model with multiple world dynamics learning tasks to better capture environment dynamics and encourage anticipatory behaviors for occlusion avoidance. We evaluate our proposed method in four photo-realistic UE4 environments to ensure validity and generalizability. Empirical results show that our method outperforms fixed and active baselines in various scenarios with different numbers of cameras and humans. | Accept: poster |
ICLR.cc/2019/Conference | Adaptive Sample-space & Adaptive Probability coding: a neural-network based approach for compression | We propose Adaptive Sample-space & Adaptive Probability (ASAP) coding, an efficient neural-network based method for lossy data compression.
Our ASAP coding distinguishes itself from the conventional method based on adaptive arithmetic coding in that it models the probability distribution for the quantization process in such a way that one can conduct back-propagation for the quantization width that determines the support of the distribution.
Our ASAP also trains the model with a novel, hyper-parameter free multiplicative loss for the rate-distortion tradeoff.
With our ASAP encoder, we are able to compress the image files in the Kodak dataset to as low as one fifth the size of the JPEG-compressed image without compromising their visual quality, and achieved the state-of-the-art result in terms of MS-SSIM based rate-distortion tradeoff. | Reject |
ICLR.cc/2019/Conference | On the Convergence and Robustness of Batch Normalization | Despite its empirical success, the theoretical underpinnings of the stability, convergence and acceleration properties of batch normalization (BN) remain elusive. In this paper, we attack this problem from a modelling approach, where we perform thorough theoretical analysis on BN applied to simplified model: ordinary least squares (OLS). We discover that gradient descent on OLS with BN has interesting properties, including a scaling law, convergence for arbitrary learning rates for the weights, asymptotic acceleration effects, as well as insensitivity to choice of learning rates. We then demonstrate numerically that these findings are not specific to the OLS problem and hold qualitatively for more complex supervised learning problems. This points to a new direction towards uncovering the mathematical principles that underlies batch normalization. | Reject |
ICLR.cc/2021/Conference | In-N-Out: Pre-Training and Self-Training using Auxiliary Information for Out-of-Distribution Robustness | Consider a prediction setting with few in-distribution labeled examples and many unlabeled examples both in- and out-of-distribution (OOD). The goal is to learn a model which performs well both in-distribution and OOD. In these settings, auxiliary information is often cheaply available for every input. How should we best leverage this auxiliary information for the prediction task? Empirically across three image and time-series datasets, and theoretically in a multi-task linear regression setting, we show that (i) using auxiliary information as input features improves in-distribution error but can hurt OOD error; but (ii) using auxiliary information as outputs of auxiliary pre-training tasks improves OOD error. To get the best of both worlds, we introduce In-N-Out, which first trains a model with auxiliary inputs and uses it to pseudolabel all the in-distribution inputs, then pre-trains a model on OOD auxiliary outputs and fine-tunes this model with the pseudolabels (self-training). We show both theoretically and empirically that In-N-Out outperforms auxiliary inputs or outputs alone on both in-distribution and OOD error. | Accept (Poster) |
ICLR.cc/2023/Conference | Quantifying Memorization Across Neural Language Models | Large language models (LMs) have been shown to memorize parts of their training data, and when prompted appropriately, they will emit the memorized training data verbatim. This is undesirable because memorization violates privacy (exposing user data), degrades utility (repeated easy-to-memorize text is often low quality), and hurts fairness (some texts are memorized over others).
We describe three log-linear relationships that quantify the degree to which LMs emit memorized training data. Memorization significantly grows as we increase (1) the capacity of a model, (2) the number of times an example has been duplicated, and (3) the number of tokens of context used to prompt the model. Surprisingly, we find the situation becomes complicated when generalizing these results across model families. On the whole, we find that memorization in LMs is more prevalent than previously believed and will likely get worse as models continues to scale, at least without active mitigations. | Accept: notable-top-25% |
ICLR.cc/2021/Conference | Black-Box Optimization Revisited: Improving Algorithm Selection Wizards through Massive Benchmarking | Existing studies in black-box optimization for machine learning suffer from low
generalizability, caused by a typically selective choice of problem instances used
for training and testing different optimization algorithms. Among other issues,
this practice promotes overfitting and poor-performing user guidelines. To address
this shortcoming, we propose in this work a benchmark suite, OptimSuite,
which covers a broad range of black-box optimization problems, ranging from
academic benchmarks to real-world applications, from discrete over numerical
to mixed-integer problems, from small to very large-scale problems, from noisy
over dynamic to static problems, etc. We demonstrate the advantages of such a
broad collection by deriving from it Automated Black Box Optimizer (ABBO), a
general-purpose algorithm selection wizard. Using three different types of algorithm
selection techniques, ABBO achieves competitive performance on all
benchmark suites. It significantly outperforms previous state of the art on some of
them, including YABBOB and LSGO. ABBO relies on many high-quality base
components. Its excellent performance is obtained without any task-specific
parametrization. The benchmark collection, the ABBO wizard, its base solvers,
as well as all experimental data are reproducible and open source in OptimSuite. | Reject |
ICLR.cc/2022/Conference | Reinforcement Learning with Efficient Active Feature Acquisition | Solving real-life sequential decision making problems under partial observability involves an exploration-exploitation problem. To be successful, an agent needs to efficiently gather valuable information about the state of the world for making rewarding decisions. However, in real-life, acquiring valuable information is often highly costly, e.g., in the medical domain, information acquisition might correspond to performing a medical test on a patient. Thus it poses a significant challenge for the agent to learn optimal task policy while efficiently reducing the cost for information acquisition. In this paper, we introduce a model-based framework to solve such exploration-exploitation problem during its execution. Key to the success is a sequential variational auto-encoder which could learn high-quality representations over the partially observed/missing features, where such representation learning serves as a prime factor to drive efficient policy training under the cost-sensitive setting. We demonstrate our proposed method could significantly outperform conventional approaches in a control domain as well as using a medical simulator. | Reject |
ICLR.cc/2023/Conference | Transformer needs NMDA receptor nonlinearity for long-term memory | The NMDA receptor (NMDAR) in the hippocampus is essential for learning and memory. We find an interesting resemblance between deep models' nonlinear activation function and the NMDAR's nonlinear dynamics. In light of a recent study that compared the transformer architecture to the formation of hippocampal memory, this paper presents new findings that NMDAR-like nonlinearity may be essential for consolidating short-term working memory into long-term reference memory. We design a navigation task assessing these two memory functions and show that manipulating the activation function (i.e., mimicking the Mg$^{2+}$-gating of NMDAR) disrupts long-term memory formation. Our experimental data suggest that the concept of place cells and reference memory may reside in the feed-forward network layer of transformers and that nonlinearity plays a key role in these processes. Our findings propose that the transformer architecture and hippocampal spatial representation resemble by sharing the overlapping concept of NMDAR-like nonlinearity. | Reject |
ICLR.cc/2023/Conference | Conditional Antibody Design as 3D Equivariant Graph Translation | Antibody design is valuable for therapeutic usage and biological research. Existing deep-learning-based methods encounter several key issues: 1) incomplete context for Complementarity-Determining Regions (CDRs) generation; 2) incapability of capturing the entire 3D geometry of the input structure; 3) inefficient prediction of the CDR sequences in an autoregressive manner. In this paper, we propose Multi-channel Equivariant Attention Network (MEAN) to co-design 1D sequences and 3D structures of CDRs. To be specific, MEAN formulates antibody design as a conditional graph translation problem by importing extra components including the target antigen and the light chain of the antibody. Then, MEAN resorts to E(3)-equivariant message passing along with a proposed attention mechanism to better capture the geometrical correlation between different components. Finally, it outputs both the 1D sequences and 3D structure via a multi-round progressive full-shot scheme, which enjoys more efficiency and precision against previous autoregressive approaches. Our method significantly surpasses state-of-the-art models in sequence and structure modeling, antigen-binding CDR design, and binding affinity optimization. Specifically, the relative improvement to baselines is about 23\% in antigen-binding CDR design and 34\% for affinity optimization. | Accept: notable-top-5% |
ICLR.cc/2021/Conference | A teacher-student framework to distill future trajectories | By learning to predict trajectories of dynamical systems, model-based methods can make extensive use of all observations from past experience. However, due to partial observability, stochasticity, compounding errors, and irrelevant dynamics, training to predict observations explicitly often results in poor models. Model-free techniques try to side-step the problem by learning to predict values directly. While breaking the explicit dependency on future observations can result in strong performance, this usually comes at the cost of low sample efficiency, as the abundant information about the dynamics contained in future observations goes unused. Here we take a step back from both approaches: Instead of hand-designing how trajectories should be incorporated, a teacher network learns to interpret the trajectories and to provide target activations which guide a student model that can only observe the present. The teacher is trained with meta-gradients to maximize the student's performance on a validation set. We show that our approach performs well on tasks that are difficult for model-free and model-based methods, and we study the role of every component through ablation studies. | Accept (Poster) |
ICLR.cc/2023/Conference | Unified neural representation model for physical and conceptual spaces | The spatial processing system of the brain uses grid-like neural representations (grid cells) for supporting vector-based navigation. Experiments also suggest that neural representations for concepts (concept cells) exist in the human brain, and conceptual inference relies on navigation in conceptual spaces. We propose a unified model called ``disentangled successor information (DSI)'' that explains neural representations for both physical and conceptual spaces. DSI generates grid-like representations in a 2-dimensional space that highly resemble those observed in the brain. Moreover, the same model creates concept-specific representations from linguistic inputs, corresponding to concept cells. Mathematically, DSI vectors approximate value functions for navigation and word vectors obtained by word embedding methods, thus enabling both spatial navigation and conceptual inference based on vector-based calculation. Our results suggest that a single principle can explain computation of physical and conceptual spaces in the human brain. | Reject |
ICLR.cc/2022/Conference | Understanding Intrinsic Robustness Using Label Uncertainty | A fundamental question in adversarial machine learning is whether a robust classifier exists for a given task. A line of research has made some progress towards this goal by studying the concentration of measure, but we argue standard concentration fails to fully characterize the intrinsic robustness of a classification problem since it ignores data labels which are essential to any classification task. Building on a novel definition of label uncertainty, we empirically demonstrate that error regions induced by state-of-the-art models tend to have much higher label uncertainty than randomly-selected subsets. This observation motivates us to adapt a concentration estimation algorithm to account for label uncertainty, resulting in more accurate intrinsic robustness measures for benchmark image classification problems. | Accept (Poster) |
ICLR.cc/2021/Conference | Non-Attentive Tacotron: Robust and controllable neural TTS synthesis including unsupervised duration modeling | This paper presents Non-Attentive Tacotron based on the Tacotron 2 text-to-speech model, replacing the attention mechanism with an explicit duration predictor. This improves robustness significantly as measured by unaligned duration ratio and word deletion rate, two metrics introduced in this paper for large-scale robustness evaluation using a pre-trained speech recognition model. With the use of Gaussian upsampling, Non-Attentive Tacotron achieves a 5-scale mean opinion score for naturalness of 4.41, slightly outperforming Tacotron 2. The duration predictor enables both utterance-wide and per-phoneme control of duration at inference time. When accurate target durations are scarce or unavailable in the training data, we propose a method using a fine-grained variational auto-encoder to train the duration predictor in a semi-supervised or unsupervised manner, with results almost as good as supervised training. | Reject |
ICLR.cc/2020/Conference | BRIDGING ADVERSARIAL SAMPLES AND ADVERSARIAL NETWORKS | Generative adversarial networks have achieved remarkable performance on various tasks but suffer from sensitivity to hyper-parameters, training instability, and mode collapse. We find that this is partly due to gradient given by non-robust discriminator containing non-informative adversarial noise, which can hinder generator from catching the pattern of real samples. Inspired by defense against adversarial samples, we introduce adversarial training of discriminator on real samples that does not exist in classic GANs framework to make adversarial training symmetric, which can balance min-max game and make discriminator more robust. Robust discriminator can give more informative gradient with less adversarial noise, which can stabilize training and accelerate convergence. We validate the proposed method on image generation tasks with varied network architectures quantitatively. Experiments show that training stability, perceptual quality, and diversity of generated samples are consistently improved with small additional training computation cost. | Reject |
ICLR.cc/2023/Conference | Adaptive Update Direction Rectification for Unsupervised Continual Learning | Recent works on continual learning have shown that unsupervised continual learning (UCL) methods rival or even beat supervised continual learning methods. However, most UCL methods typically adopt fixed learning strategies with pre-defined objectives and ignore the influence of the constant shift of data distributions on the newer training process. This non-adaptive paradigm tends to achieve sub-optimal performance, since the optimal update direction (to ensure the trade-off between old and new tasks) keeps changing during training over sequential tasks. In this work, we thus propose a novel UCL framework termed AUDR to adaptively rectify the update direction by a policy network (i.e., the Actor) at each training step based on the reward predicted by a value network (i.e., the Critic). Concretely, different from existing Actor-Critic based reinforcement learning works, there are three vital designs that make our AUDR applicable to the UCL setting: (1) A reward function to measure the score/value of the currently selected action, which provides the ground-truth reward to guide the Critic's predictions; (2) An action space for the Actor to select actions (i.e., update directions) according to the reward predicted by the Critic; (3) A multinomial sampling strategy with a lower-bound on the sampling probability of each action, which is designed to improve the variance of the Actor's selected actions for more diversified exploration. Extensive experiments show that our AUDR achieves state-of-the-art results under both the in-dataset and cross-dataset UCL settings. Importantly, our AUDR also shows superior performance when combined with other UCL methods, which suggests that our AUDR is highly extensible and versatile. | Reject |
ICLR.cc/2023/Conference | Efficient Discrete Multi Marginal Optimal Transport Regularization | Optimal transport has emerged as a powerful tool for a variety of problems in machine learning, and it is frequently used to enforce distributional constraints. In this context, existing methods often use either a Wasserstein metric, or else they apply concurrent barycenter approaches when more than two distributions are considered. In this paper, we leverage multi-marginal optimal transport (MMOT), where we take advantage of a procedure that computes a generalized earth mover's distance as a sub-routine. We show that not only is our algorithm computationally more efficient compared to other barycentric-based distance methods, but it has the additional advantage that gradients used for backpropagation can be efficiently computed during the forward pass computation itself, which leads to substantially faster model training. We provide technical details about this new regularization term and its properties, and we present experimental demonstrations of faster runtimes when compared to standard Wasserstein-style methods. Finally, on a range of experiments designed to assess effectiveness at enforcing fairness, we demonstrate our method compares well with alternatives. | Accept: notable-top-25% |
ICLR.cc/2021/Conference | Disentangling Adversarial Robustness in Directions of the Data Manifold | Using generative models (GAN or VAE) to craft adversarial examples, i.e. generative adversarial examples, has received increasing attention in recent years. Previous studies showed that the generative adversarial examples work differently compared to that of the regular adversarial examples in many aspects, such as attack rates, perceptibility, and generalization. But the reasons causing the differences between regular and generative adversarial examples are unclear. In this work, we study the theoretical properties of the attacking mechanisms of the two kinds of adversarial examples in the Gaussian mixture data model case. We prove that adversarial robustness can be disentangled in directions of the data manifold. Specifically, we find that: 1. Regular adversarial examples attack in directions of small variance of the data manifold, while generative adversarial examples attack in directions of large variance. 2. Standard adversarial training increases model robustness by extending the data manifold boundary in directions of small variance, while on the contrary, adversarial training with generative adversarial examples increases model robustness by extending the data manifold boundary directions of large variance. In experiments, we demonstrate that these phenomena also exist on real datasets. Finally, we study the robustness trade-off between generative and regular adversarial examples. We show that the conflict between regular and generative adversarial examples is much smaller than the conflict between regular adversarial examples of different norms. | Reject |
ICLR.cc/2020/Conference | Learning transitional skills with intrinsic motivation | By maximizing an information theoretic objective, a few recent methods empower the agent to explore the environment and learn useful skills without supervision. However, when considering to use multiple consecutive skills to complete a specific task, the transition from one to another cannot guarantee the success of the process due to the evident gap between skills. In this paper, we propose to learn transitional skills (LTS) in addition to creating diverse primitive skills without a reward function. By introducing an extra latent variable for transitional skills, our LTS method discovers both primitive and transitional skills by minimizing the difference of mutual information and the similarity of skills. By considering various simulated robotic tasks, our results demonstrate the effectiveness of LTS on learning both diverse primitive skills and transitional skills, and show its superiority in smooth transition of skills over the state-of-the-art baseline DIAYN. | Reject |