Subhabrata Sen (Harvard University)

Abstract: Multi-modal datasets comprise diverse  features collected on the same entity. For example, one might collect genomic, proteomic and transcriptomic data from the same individual. The goal is to combine these features to improve downstream statistical performance. While multi-modal data is ubiquitous across diverse applications, statistical theory for multi-modal data analysis lies in a nascent state. In this course, we will discuss some recent progress on the rigorous study of these problems. To derive the associated statistical algorithms, we will utilize insights from high-dimensional probability, graphical models and statistical physics.