We formally study the problem of classification under adversarial perturbations, both from the learner's perspective, and from the viewpoint of a third-party who aims at certifying the robustness of a given black-box classifier. We analyze a PAC-type framework of semi-supervised learning and identify possibility and impossibility results for proper learning of VC-classes in this setting. We further introduce and study a new setting of black-box certification under limited query budget. We analyze this for various classes of predictors and types of perturbation. We also consider the viewpoint of a black-box adversary that aims at finding adversarial examples, showing that the existence of an adversary with polynomial query complexity implies the existence of a robust learner with small sample complexity.

The study of network formation is pervasive in economics, sociology, and many other fields. In this paper, we model network formation as a ``choice'' that is made by nodes in a network to connect to other nodes. We study these ``choices'' using discrete-choice models, in which an agent chooses between two or more discrete alternatives. One framework for studying network formation is the multinomial logit (MNL) model. We highlight limitations of the MNL model on networks that are constructed from empirical data. We employ the ``repeated choice'' (RC) model to study network formation \cite{TrainRevelt97mixedlogit}. We argue that the RC model overcomes important limitations of the MNL model and is well-suited to study network formation. We also illustrate how to use the RC model to accurately study network formation using both synthetic and real-world networks. Using synthetic networks, we also compare the performance of the MNL model and the RC model; we find that the RC model estimates the data-generation process of our synthetic networks more accurately than the MNL model. We provide examples of qualitatively interesting questions -- the presence of homophily in a teen friendship network and the fact that new patents are more likely to cite older, more cited, and similar patents -- for which the RC model allows us to achieve insights.

Evaluating the reliability of intelligent physical systems against rare catastrophic events poses a huge testing burden for real-world applications. Simulation provides a useful, if not unique, platform to evaluate the extremal risks of these AIenabled systems before their deployments. Importance Sampling (IS), while proven to be powerful for rare-event simulation, faces challenges in handling these systems due to their black-box nature that fundamentally undermines its efficiency guarantee. To overcome this challenge, we propose a framework called Deep Probabilistic Accelerated Evaluation (D-PrAE) to design IS, which leverages rare-event-set learning and and a new notion of efficiency certificate. D-PrAE combines the dominating point method with deep neural network classifiers to achieve superior estimation efficiency. We present theoretical guarantees and demonstrate the empirical effectiveness of D-PrAE via examples on the safety-testing of self-driving algorithms that are beyond the reach of classical variance reduction techniques.

Machine learning models are not static and may need to be retrained on slightly changed datasets, for instance, with the addition or deletion of a set of data points. This has many applications, including privacy, robustness, bias reduction, and uncertainty quantifcation. However, it is expensive to retrain models from scratch. To address this problem, we propose the DeltaGrad algorithm for rapid retraining machine learning models based on information cached during the training phase. We provide both theoretical and empirical support for the effectiveness of DeltaGrad, and show that it compares favorably to the state of the art.

Learning Invariant Representations has been successfully applied for reconciling a source and a target domain for Unsupervised Domain Adaptation. By investigating the robustness of such methods under the prism of the cluster assumption, we bring new evidence that invariance with a low source risk does not guarantee a well-performing target classifier. More precisely, we show that the cluster assumption is violated in the target domain despite being maintained in the source domain, indicating a lack of robustness of the target classifier. To address this problem, we demonstrate the importance of enforcing the cluster assumption in the target domain, named Target Consistency (TC), especially when paired with Class-Level InVariance (CLIV). Our new approach results in a significant improvement, on both image classification and segmentation benchmarks, over state-of-the-art methods based on invariant representations. Importantly, our method is flexible and easy to implement, making it a complementary technique to existing approaches for improving transferability of representations.

The problem of inferring unknown graph edges from numerical data at a graph's nodes appears in many forms across machine learning. We study a version of this problem that arises in the field of landscape genetics, where genetic similarity between populations of organisms living in a heterogeneous landscape is explained by a weighted graph that encodes the ease of dispersal through that landscape. Our main contribution is an efficient algorithm for inverse landscape genetics, which is the task of inferring this graph from measurements of genetic similarity at different locations (graph nodes). We reduced the problem to that of inferring graph edges from noisy measurements of effective resistances between graph nodes, which have been observed to correlate well with genetic similarity. Building on Hoskins et. al., we develop an efficient first-order optimization method for solving this problem. Despite its non-convex nature, extensive experiments on synthetic and real genetic data establish that our method provides fast and reliable convergence, significantly outperforming existing heuristics used in the field.

Several methods have recently been proposed for the Single Image Super-Resolution (SISR) problem. The current methods assume that a single low-resolution image can only yield a single high-resolution image. In addition, all of these methods use low-resolution images that were artificially generated through simple bilinear down-sampling. We argue that, first and foremost, the problem of SISR is an one-to-many mapping problem between the low-resolution and all possible candidate high-resolution images and we address the challenging task of learning how to realistically degrade and down-sample high-resolution images. To circumvent this problem, we propose SR-NAM which utilizes the Non-Adversarial Mapping (NAM) technique. Furthermore, we propose a degradation model that learns how to transform high-resolution images to low-resolution images that resemble realistically taken low-resolution photos. Finally, some qualitative results for the proposed method along with the weaknesses of SR-NAM are included.

The Variational Auto-Encoder (VAE) belongs to a class of models, which we will refer to as deep maximum likelihood models, that uses a deep neural network to learn a maximum likelihood model for some input data. They are perhaps the most simple and efficient deep maximum likelihood model available, and have thus gained popularity in representation learning and generative image modeling. Unfortunately, in my opinion, in some circles the term "VAE" has become somewhat synonymous with "an auto-encoder with stochastic regularization that generates useful or beautiful samples", which has led to various misconceptions about VAEs. In this tutorial, we will return to the probabilistic and information theoretic roots of VAEs, clarify common misconceptions about VAEs, and look at a toy example on 2D data that will illustrate the capabilities and limitations of VAEs. In Section 2, we will give an overview of what is a maximum likelihood model and what a VAE looks like.

We study the effect of norm based regularization on the size of coresets for regression problems. Specifically, given a matrix $ \mathbf{A} \in {\mathbb{R}}^{n \times d}$ with $n\gg d$ and a vector $\mathbf{b} \in \mathbb{R} ^ n $ and $\lambda > 0$, we analyze the size of coresets for regularized versions of regression of the form $\|\mathbf{Ax}-\mathbf{b}\|_p^r + \lambda\|{\mathbf{x}}\|_q^s$ . Prior work has shown that for ridge regression (where $p,q,r,s=2$) we can obtain a coreset that is smaller than the coreset for the unregularized counterpart i.e. least squares regression (Avron et al). We show that when $r \neq s$, no coreset for regularized regression can have size smaller than the optimal coreset of the unregularized version. The well known lasso problem falls under this category and hence does not allow a coreset smaller than the one for least squares regression. We propose a modified version of the lasso problem and obtain for it a coreset of size smaller than the least square regression. We empirically show that the modified version of lasso also induces sparsity in solution, similar to the original lasso. We also obtain smaller coresets for $\ell_p$ regression with $\ell_p$ regularization. We extend our methods to multi response regularized regression. Finally, we empirically demonstrate the coreset performance for the modified lasso and the $\ell_1$ regression with $\ell_1$ regularization.

The framework of variational autoencoders (VAEs) provides a principled method for jointly learning latent-variable models and corresponding inference models. However, the main drawback of this approach is the blurriness of the generated images. Some studies link this effect to the objective function, namely, the (negative) log-likelihood. Here, we propose to enhance VAEs by adding a random variable that is a downscaled version of the original image and still use the log-likelihood function as the learning objective. Further, by providing the downscaled image as an input to the decoder, it can be used in a manner similar to the super-resolution. We present empirically that the proposed approach performs comparably to VAEs in terms of the negative log-likelihood, but it obtains a better FID score in data synthesis.