Speaker: Andrew Thomas, Department of Statistics and Actuarial Science, University of Iowa

Title: BCLR: a Flexible, Interpretable Bayesian Changepoint Method Via Logistic Regression

Abstract: In this talk, I will present a novel Bayesian method for multivariate changepoint detection that allows for simultaneous inference on the location of a changepoint and the coefficients of a logistic regression model for distinguishing pre-changepoint data from post-changepoint data. In contrast to many methods for multivariate changepoint detection, the proposed method is applicable to various data types and avoids strict assumptions regarding the distribution of the data and the nature of the change. Additionally, the regression coefficients provide an interpretable description of a change that is potentially complex. For posterior inference, the model admits a simple Gibbs sampling algorithm based on Pólya-gamma data augmentation. I will present conditions under which the proposed method is guaranteed to recover the true underlying changepoint. As a testing ground for our method, we consider the problem of detecting topological changes in time series of noisy images. I will demonstrate that the proposed method—combined with a novel topological feature embedding—exceeds existing methods on both simulated image data and real nanoparticle and solar flare videos. To evince our method’s flexibility, I will also show that our method yields superior performance in a standard covariance change problem. Joint work with Michael Jauch and David S. Matteson.