This paper introduces and studies the $k\texttt{-experts}$ problem -- a generalization of the classic Prediction with Expert's Advice (i.e., the $\texttt{Experts}$) problem. Unlike the $\texttt{Experts}$ problem, where the learner chooses exactly one expert, in this problem, the learner selects a subset of $k$ experts from a pool of $N$ experts at each round. The reward obtained by the learner at any round depends on the rewards of the selected experts. The $k\texttt{-experts}$ problem arises in many practical settings, including online ad placements, personalized news recommendations, and paging. Our primary goal is to design an online learning policy having a small regret. In this pursuit, we propose $\texttt{SAGE}$ ($\textbf{Sa}$mpled Hed$\textbf{ge}$) - a framework for designing efficient online learning policies by leveraging statistical sampling techniques. We show that, for many related problems, $\texttt{SAGE}$ improves upon the state-of-the-art bounds for regret and computational complexity. Furthermore, going beyond the notion of regret, we characterize the mistake bounds achievable by online learning policies for a class of stable loss functions. We conclude the paper by establishing a tight regret lower bound for a variant of the $k\texttt{-experts}$ problem and carrying out experiments with standard datasets.
Conditional generative models of high-dimensional images have many applications, but supervision signals from conditions to images can be expensive to acquire. This paper describes Diffusion-Decoding models with Contrastive representations (D2C), a paradigm for training unconditional variational autoencoders (VAEs) for few-shot conditional image generation. D2C uses a learned diffusion-based prior over the latent representations to improve generation and contrastive self-supervised learning to improve representation quality. D2C can adapt to novel generation tasks conditioned on labels or manipulation constraints, by learning from as few as 100 labeled examples. On conditional generation from new labels, D2C achieves superior performance over state-of-the-art VAEs and diffusion models. On conditional image manipulation, D2C generations are two orders of magnitude faster to produce over StyleGAN2 ones and are preferred by 50% - 60% of the human evaluators in a double-blind study.
Data augmentation is often used to enlarge datasets with synthetic samples generated in accordance with the underlying data distribution. To enable a wider range of augmentations, we explore negative data augmentation strategies (NDA)that intentionally create out-of-distribution samples. We show that such negative out-of-distribution samples provide information on the support of the data distribution, and can be leveraged for generative modeling and representation learning. We introduce a new GAN training objective where we use NDA as an additional source of synthetic data for the discriminator. We prove that under suitable conditions, optimizing the resulting objective still recovers the true data distribution but can directly bias the generator towards avoiding samples that lack the desired structure. Empirically, models trained with our method achieve improved conditional/unconditional image generation along with improved anomaly detection capabilities. Further, we incorporate the same negative data augmentation strategy in a contrastive learning framework for self-supervised representation learning on images and videos, achieving improved performance on downstream image classification, object detection, and action recognition tasks. These results suggest that prior knowledge on what does not constitute valid data is an effective form of weak supervision across a range of unsupervised learning tasks.
We consider the classical uncoded caching problem from an online learning point-of-view. A cache of limited storage capacity can hold $C$ files at a time from a large catalog. A user requests an arbitrary file from the catalog at each time slot. Before the file request from the user arrives, a caching policy populates the cache with any $C$ files of its choice. In the case of a cache-hit, the policy receives a unit reward and zero rewards otherwise. In addition to that, there is a cost associated with fetching files to the cache, which we refer to as the switching cost. The objective is to design a caching policy that incurs minimal regret while considering both the rewards due to cache-hits and the switching cost due to the file fetches. The main contribution of this paper is the switching regret analysis of a Follow the Perturbed Leader-based anytime caching policy, which is shown to have an order optimal switching regret. In this pursuit, we improve the best-known switching regret bound for this problem by a factor of $\Theta(\sqrt{C}).$ We conclude the paper by comparing the performance of different popular caching policies using a publicly available trace from a commercial CDN server.
We consider the online $\textsf{Bipartite Caching}$ problem where $n$ users are connected to $m$ caches in the form of a bipartite network. Each of the $m$ caches has a file storage capacity of $C$. There is a library consisting of $N >C$ distinct files. Each user can request any one of the files from the library at each time slot. We allow the file request sequences to be chosen in an adversarial fashion. A user's request at a time slot is satisfied if the requested file is already hosted on at least one of the caches connected to the user at that time slot. Our objective is to design an efficient online caching policy with minimal regret. In this paper, we propose $\textsf{LeadCache,}$ an online caching policy based on the $\textsf{Follow the Perturbed Leader}$ (FTPL) paradigm. We show that $\textsf{LeadCache}$ is regret optimal up to a multiplicative factor of $\tilde{O}(n^{0.375}).$ As a byproduct of our analysis, we design a new linear-time deterministic Pipage rounding procedure for the LP relaxation of a well-known NP-hard combinatorial optimization problem in this area. Our new rounding algorithm substantially improves upon the currently best-known complexity for this problem. Moreover, we show the surprising result that under mild Strong-Law-type assumptions on the file request sequence, the rate of file fetches to the caches approaches to zero under the $\textsf{LeadCache}$ policy. Finally, we derive a tight universal regret lower bound for the $\textsf{Bipartite Caching}$ problem, which critically makes use of results from graph coloring theory and certifies the announced approximation ratio.
Adversarial robust models have been shown to learn more robust and interpretable features than standard trained models. As shown in [\cite{tsipras2018robustness}], such robust models inherit useful interpretable properties where the gradient aligns perceptually well with images, and adding a large targeted adversarial perturbation leads to an image resembling the target class. We perform experiments to show that interpretable and perceptually aligned gradients are present even in models that do not show high robustness to adversarial attacks. Specifically, we perform adversarial training with attack for different max-perturbation bound. Adversarial training with low max-perturbation bound results in models that have interpretable features with only slight drop in performance over clean samples. In this paper, we leverage models with interpretable perceptually-aligned features and show that adversarial training with low max-perturbation bound can improve the performance of models for zero-shot and weakly supervised localization tasks.
In social dilemma situations, individual rationality leads to sub-optimal group outcomes. Several human engagements can be modeled as a sequential (multi-step) social dilemmas. However, in contrast to humans, Deep Reinforcement Learning agents trained to optimize individual rewards in sequential social dilemmas converge to selfish, mutually harmful behavior. We introduce a status-quo loss (SQLoss) that encourages an agent to stick to the status quo, rather than repeatedly changing its policy. We show how agents trained with SQLoss evolve cooperative behavior in several social dilemma matrix games. To work with social dilemma games that have visual input, we propose GameDistill. GameDistill uses self-supervision and clustering to automatically extract cooperative and selfish policies from a social dilemma game. We combine GameDistill and SQLoss to show how agents evolve socially desirable cooperative behavior in the Coin Game.
Social dilemma situations bring out the conflict between individual and group rationality. When individuals act rationally in such situations, the group suffers sub-optimal outcomes. The Iterative Prisoner's Dilemma (IPD) is a two-player game that offers a theoretical framework to model and study such social situations. In the Prisoner's Dilemma, individualistic behavior leads to mutual defection and sub-optimal outcomes. This result is in contrast to what one observes in human groups, where humans often sacrifice individualistic behavior for the good of the collective. It is interesting to study how and why such cooperative and individually irrational behavior emerges in human groups. To this end, recent work models this problem by treating each player as a Deep Reinforcement Learning (RL) agent and evolves cooperative behavioral policies through internal information or reward sharing mechanisms. We propose an approach to evolve cooperative behavior between RL agents playing the IPD game without sharing rewards, internal details (weights, gradients), or a communication channel. We introduce a Status-Quo loss (SQLoss) that incentivizes cooperative behavior by encouraging policy stationarity. We also describe an approach to transform a two-player game (with visual inputs) into its IPD formulation through self-supervised skill discovery (IPDistill).We show how our approach outperforms existing approaches in the Iterative Prisoner's Dilemma and the two-player Coin game.
Interpretability is an emerging area of research in trustworthy machine learning. Safe deployment of machine learning system mandates that the prediction and its explanation be reliable and robust. Recently, it was shown that one could craft perturbations that produce perceptually indistinguishable inputs having the same prediction, yet very different interpretations. We tackle the problem of attributional robustness (i.e. models having robust explanations) by maximizing the alignment between the input image and its saliency map using soft-margin triplet loss. We propose a robust attribution training methodology that beats the state-of-the-art attributional robustness measure by a margin of approximately 6-18% on several standard datasets, ie. SVHN, CIFAR-10 and GTSRB. We further show the utility of the proposed robust model in the domain of weakly supervised object localization and segmentation. Our proposed robust model also achieves a new state-of-the-art object localization accuracy on the CUB-200 dataset.
We propose an unsupervised multi-conditional image generation pipeline: cFineGAN, that can generate an image conditioned on two input images such that the generated image preserves the texture of one and the shape of the other input. To achieve this goal, we extend upon the recently proposed work of FineGAN \citep{singh2018finegan} and make use of standard as well as shape-biased pre-trained ImageNet models. We demonstrate both qualitatively as well as quantitatively the benefit of using the shape-biased network. We present our image generation result across three benchmark datasets- CUB-200-2011, Stanford Dogs and UT Zappos50k.