Major Oral: "Stochastic EM and related open problems"
Speaker: Yanjie Zhong, Washington University in Saint Louis
Abstract: Modelling data with missingness or latent variables is an important topic in modern statistics. As a variant of the classical EM algorithm, stochastic EM is a popular approach to deal with problems where missing or latent variables exist because of its efficient and easy-to-implement scheme. In this talk, we discuss some statistical problems related to stochastic EM, including non-asymptotic estimation error, algorithm variability and variable selection.
Hosts: Likai Chen, Todd Kuffner, and Soumendra Lahiri