Unsupervised learning techniques in computer vision often require learning latent representations, such as low-dimensional linear and non-linear subspaces. Noise and outliers in the data can frustrate these approaches by obscuring the latent spaces. Our main goal is deeper understanding and new development of robust approaches for representation learning. We provide a new interpretation for existing robust approaches and present two specific contributions: a new robust PCA approach, which can separate foreground features from dynamic background, and a novel robust spectral clustering method, that can cluster facial images with high accuracy. Both contributions show superior performance to standard methods on real-world test sets.

Active learning has long been a topic of study in machine learning. However, as increasingly complex and opaque models have become standard practice, the process of active learning, too, has become more opaque. There has been little investigation into interpreting what specific trends and patterns an active learning strategy may be exploring. This work expands on the Local Interpretable Model-agnostic Explanations framework (LIME) to provide explanations for active learning recommendations. We demonstrate how LIME can be used to generate locally faithful explanations for an active learning strategy, and how these explanations can be used to understand how different models and datasets explore a problem space over time. In order to quantify the per-subgroup differences in how an active learning strategy queries spatial regions, we introduce a notion of uncertainty bias (based on disparate impact) to measure the discrepancy in the confidence for a model's predictions between one subgroup and another. Using the uncertainty bias measure, we show that our query explanations accurately reflect the subgroup focus of the active learning queries, allowing for an interpretable explanation of what is being learned as points with similar sources of uncertainty have their uncertainty bias resolved. We demonstrate that this technique can be applied to track uncertainty bias over user-defined clusters or automatically generated clusters based on the source of uncertainty.

Two-sample feature selection is the problem of finding features that describe a difference between two probability distributions, which is a ubiquitous problem in both scientific and engineering studies. However, existing methods have limited applicability because of their restrictive assumptions on data distributoins or computational difficulty. In this paper, we resolve these difficulties by formulating the problem as a sparsest $k$-subgraph problem. The proposed method is nonparametric and does not assume any specific parametric models on the data distributions. We show that the proposed method is computationally efficient and does not require any extra computation for model selection. Moreover, we prove that the proposed method provides a consistent estimator of features under mild conditions. Our experimental results show that the proposed method outperforms the current method with regard to both accuracy and computation time.

We study learning problems in the general setup, for arbitrary classes of functions $F$, distributions $X$ and targets $Y$. Because proper learning procedures, i.e., procedures that are only allowed to select functions in $F$, tend to perform poorly unless the problem satisfies some additional structural property (e.g., that $F$ is convex), we consider unrestricted learning procedures, that is, procedures that are free to choose functions outside the given class $F$. We present a new unrestricted procedure that is optimal in a very strong sense: it attains the best possible accuracy/confidence tradeoff for (almost) any triplet $(F,X,Y)$, including in heavy-tailed problems. Moreover, the tradeoff the procedure attains coincides with what one would expect if $F$ were convex, even when $F$ is not; and when $F$ happens to be convex, the procedure is proper; thus, the unrestricted procedure is actually optimal in both realms, for convex classes as a proper procedure and for arbitrary classes as an unrestricted procedure. The notion of optimality we consider is problem specific: our procedure performs with the best accuracy/confidence tradeoff one can hope to achieve for each individual problem. As such, it is a significantly stronger property than the standard `worst-case' notion, in which one considers optimality as the best uniform estimate that holds for a relatively large family of problems. Thanks to the sharp and problem-specific estimates we obtain, classical, worst-case bounds are immediate outcomes of our main result.

While deep learning is remarkably successful on perceptual tasks, it was also shown to be vulnerable to adversarial perturbations of the input. These perturbations denote noise added to the input that was generated specifically to fool the system while being quasi-imperceptible for humans. More severely, there even exist universal perturbations that are input-agnostic but fool the network on the majority of inputs. While recent work has focused on image classification, this work proposes attacks against semantic image segmentation: we present an approach for generating (universal) adversarial perturbations that make the network yield a desired target segmentation as output. We show empirically that there exist barely perceptible universal noise patterns which result in nearly the same predicted segmentation for arbitrary inputs. Furthermore, we also show the existence of universal noise which removes a target class (e.g., all pedestrians) from the segmentation while leaving the segmentation mostly unchanged otherwise.

We consider regret minimization in repeated games with non-convex loss functions. Minimizing the standard notion of regret is computationally intractable. Thus, we define a natural notion of regret which permits efficient optimization and generalizes offline guarantees for convergence to an approximate local optimum. We give gradient-based methods that achieve optimal regret, which in turn guarantee convergence to equilibrium in this framework.

Causality has been recently introduced in databases, to model, characterize, and possibly compute causes for query answers. Connections between QA-causality and consistency-based diagnosis and database repairs (wrt. integrity constraint violations) have already been established. In this work we establish precise connections between QA-causality and both abductive diagnosis and the view-update problem in databases, allowing us to obtain new algorithmic and complexity results for QA-causality. We also obtain new results on the complexity of view-conditioned causality, and investigate the notion of QA-causality in the presence of integrity constraints, obtaining complexity results from a connection with view-conditioned causality. The abduction connection under integrity constraints allows us to obtain algorithmic tools for QA-causality.

The combination of argumentation and probability paves the way to new accounts of qualitative and quantitative uncertainty, thereby offering new theoretical and applicative opportunities. Due to a variety of interests, probabilistic argumentation is approached in the literature with different frameworks, pertaining to structured and abstract argumentation, and with respect to diverse types of uncertainty, in particular the uncertainty on the credibility of the premises, the uncertainty about which arguments to consider, and the uncertainty on the acceptance status of arguments or statements. Towards a general framework for probabilistic argumentation, we investigate a labelling-oriented framework encompassing a basic setting for rule-based argumentation and its (semi-) abstract account, along with diverse types of uncertainty. Our framework provides a systematic treatment of various kinds of uncertainty and of their relationships and allows us to retrieve (by derivation) multiple statements (sometimes assumed) or results from the literature.

You might think, based on the type of research they're doing, that the OpenAI office would be full of gadgets, full of wonder, full of weird experiments. There are no Faraday cages. Well, okay, there is a robot. And it's tucked away in a side room. It's surrounded by cobbled-together protective material so that it doesn't smash into itself if it starts flailing about due to a programming error.

TOLEDO, OHIO – Satellites in space and a robot under Lake Erie's surface are part of a network of scientific tools trying to keep algae toxins out of drinking water supplies in the shallowest of the Great Lakes. It's one of the most wide-ranging freshwater monitoring systems in the U.S., researchers say, and some of its pieces soon will be watching for harmful algae on hundreds of lakes nationwide. Researchers are creating an early warning system using real-time data from satellites that in recent years have tracked algae bloom hotpots such as Florida's Lake Okeechobee and the East Coast's Chesapeake Bay. The plan is to have it in place within two years so that states in the continental U.S. can be alerted to where toxic algae is appearing before they might detect it on the surface, said Blake Schaeffer, a researcher with U.S. Environmental Protection Agency. "You don't have to wait until someone gets sick," said Schaeffer, one of the leaders of the project.