In this paper, we proposed EWSG, which uses exponentially weighted subsampling of gradients to match the transition kernel of a base MCMC base with full gradient. The goal is better sample quality. Both local variance analysis and global non-asympto…

In this paper we addressed one of the primary challenges facing the deployment of machine learning in safety-critical applications: shifts in environment between training and deployment. To this end, we proposed I-Spec, an end-to-end framework that …

This paper presents a reactive navigation scheme for robots operating in planar workspaces, cluttered with obstacles of familiar geometry but à-priori unknown placement, and completely unknown, but strongly convex and well-separated obstacles. To th…

In this paper, we proposed a method based on graph convolution to model the spatiotemporal relationships of the pedestrians and other objects in the scene. We build our spatiotemporal graph by considering each segmented instance in a frame as a node…

In this work we studied the performance of several methods to select hyperparameters of Lasso-type estimators showing results for the Lasso and the weighted Lasso, which have respectively one or p hyperparameters. We exploited the sparsity of the so…

In this study, we propose a Nyström subspace learning framework (NSL) to ease the application of the Nyström method upon large-scale kernel SVMs. Based on our analysis, the bounds developed for the Nyström method are closely connected to NSL. The ma…

Sham Kakade acknowledges funding from the Washington Research Foundation for Innovation in Data-intensive Discovery, and the NSF Awards CCF-1637360, CCF-1703574, and CCF-1740551.

We have developed a strategy to increase the safety of a DNN-based perception for HAD systems. This strategy provides for mitigating the DNN insufficiencies, which we have divided into five categories and described, which we equate with increasing s…

In this paper, we proposed kernelized factorization matrix completion (KFMC)1, a new method for high rank matrix completion, together with an online version and an out-of-sample extension, which outperform state-of-the-art methods. Our numerics demo…

We introduce Wavesplit, a neural network for speech separation. Our model decomposes the source separation problem in two jointly trained tasks, i.e. our model extracts a representation for each speaker present in the recording and performs separati…