Seminar: Helen Zhang

Title

Selection of Interaction Effects for Ultra High-Dimensional Data

Speaker

Helen Zhang, University of Arizona

Abstract

For ultra-high dimensional data, the identification of interaction effects is an extremely challenging task. The main difficulties lie in both computational and theoretical aspects. We propose a new model selection framework for interaction selection. The new method is feasible even when the data dimension increases exponentially fast with the sample size. The new methods are featured with fast computational algorithms, rigorous theories and competitive numerical performances. Various examples are presented to illustrate the new methods.