Spring 2016

Analysis of Algorithms on Dense Matrices using Approximate Message Passing

Monday, May 2, 2016 2:00 pm2:45 pm PDT

Add to Calendar


Calvin Lab

Approximate Message Passing (AMP) is a class of algorithms motivated by statistical physics, that can be applied to a broad family of problems on dense matrices. Remarkably AMP can be analyzed exactly in the large size limit. I will describe the general approach, and illustrate it through applications to specific problems: the hidden clique problem, sparse principal component analysis, non-negative principal component analysis.