Machine Learning Seminar Series

Compound sequential change detection in multiple data streams

by

Xiaoou Li
School of Statistics
University of Minnesota

Wednesday, October 21, 2020
3:30–4:30 pm

WE CONSIDER sequential change detection in multiple data streams, where each stream has its own change point. The goal is to maximize the normal operation of the pre-change streams while controlling the proportion of post-change streams among the active streams at all time points. Taking a Bayesian formulation, we develop a compound sequential decision theory framework for this problem and propose an oracle procedure under this framework. We also extend the problem to more general settings and apply the method to the monitoring of item pool quality in educational assessment.

XIAOOU LI is an assistant professor of statistics at the University of Minnesota. She holds a Ph.D. degree in Statistics from Columbia University in 2016. Her research interest includes latent variable models, sequential analysis, psychometrics, and applied probability.