91´«Ã½

Bi-Co Mathematics Colloquium with Dr. Kyle Stanley

Dr. Kyle Stanley of the University of Pennsylvania/Penn Medicine will give a talk titled, "Functional Factor Analysis of Brain Connectivity.

Abstract:

Functional magnetic resonance imaging (fMRI) allows researchers to study activity within the brain. A common goal of fMRI studies is to examine functional connectivity (FC), or the statistical dependencies among distant brain regions. Many current approaches rely on latent variable models (e.g., factor analysis), which attempt to explain high-dimensional data through a small number of hidden factors. These models have been successful in many settings, but in the context of fMRI they miss the mark: they target the wrong forms of variation and impose exceedingly rigid assumptions. In this talk, I will present a new approach that merges ideas from several areas of mathematics and statistics: factor analysis, functional data analysis (a field which views certain data as arising from smooth curves), and matrix completion (a field concerned with filling in the missing entries of a partially-observed matrix). I will show why traditional latent variable models break down when applied to fMRI, how FFA addresses these challenges, and what we can learn about the brain from FFA.