Ruslan Salakhutdinov

TL;DR: This paper proposes feature robust optimal transport (FROT) for high-dimensional data, which jointly solves feature selection and OT problems, and proposes using the FROT algorithm for the layer selection problem in deep neural networks for semantic correspondence.

TL;DR: In this article, a post selection inference (PSI) framework for divergence measure is proposed, which can select a set of statistically significant features that discriminate two distributions, and a novel MMD estimator using the incomplete U-statistics, which has an asymptotically normal distribution under mild assumptions, is also proposed and analyzed theoretically.

TL;DR: A new family of matrix norms, the "local max" norms, are introduced, generalizing existing methods such as the max norm, the trace norm, and the weighted or smoothed weighted trace norms, which have been extensively used in the literature as regularizers for matrix reconstruction problems.

TL;DR: This article propose a method to fuse text-only large language models (LLMs) with pre-trained image encoder and decoder models, by mapping between their embedding spaces.

TL;DR: The authors evaluate several state-of-the-art few-shot methods for NLU tasks and reveal new insights: both the absolute performance and relative gap of the methods were not accurately estimated in prior literature; no single method dominates most tasks with consistent performance; improvements of some methods diminish with a larger pretrained model; and the best combined model performs close to a strong fully-supervised baseline.