Speaker: Songtao Lu, ECpE Graduate Student
Advisor: Zhengdao Wang
Title: Efficient Nonconvex Algorithms and Their Performance Analysis for Some Matrix Factorization Problems
Abstract: Matrix factorization has important applications in data analytics problems such as document clustering, community detection, and image segmentation. In this talk, I will introduce some nonconvex optimization methods for solving matrix factorization problems, such as the alternating direction method of multipliers for symmetric non-negative matrix factorization. Furthermore, I will show the convergence rate of the algorithm and sufficient conditions that guarantee the global and local optimality of the obtained solutions. Also, I will talk about the first order methods of dealing with unconstrained nonconvex problems, focusing on the convergence rate of gradient descent and block coordinate descent to the second-order stationary points (local minimum points).