### Sidebar

seminars:stat:201022

Statistics Seminar
Department of Mathematical Sciences

 DATE: Thursday, Oct. 22, 2020 1:15pm – 2:15pm zoom meeting Baozhen Wang, Binghamton University Conformal Prediction Under Covariate Shift

Abstract

The authors extend conformal prediction methodology beyond the case of exchangeable data. In particular, they show that a weighted version of conformal prediction can be used to compute distribution-free prediction intervals for problems in which the test and training covariate distributions differ, but the likelihood ratio between these two distributions is known—or, in practice, can be estimated accurately with access to a large set of unlabeled data (test covariate points). Their weighted extension of conformal prediction also applies more generally, to settings in which the data satis&#64257;es a certain weighted notion of exchangeability.