Statistics Department Seminar Series: Zachary Lipton, Assistant Professor, Departments of Machine Learning and Tepper school of Business, Carnegie Mellon University

Abstract: In this talk I will discuss distribution shift, both as an obstacle to be overcome to achieve generalization to a target distribution and as a device for establishing a guarantee that we have in fact generalized to a distribution of interest. In the first part, I will discuss the problem of label shift, where the proportion among the labels can shift but the class conditional distributions do not change, including connections to some practical problems and some theoretical results. Then I will discuss a new work in which we alter the distribution of training data in order to establish a generalization guarantee.

This seminar will be livestreamed via Zoom https://umich.zoom.us/j/94350208889. There will be a virtual reception to follow.