Randomized Block Coordinate Descent Methods for A Class of Structured Nonlinear Optimization

Time: 2017-07-04 16:49:20

Topic：Randomized Block Coordinate Descent Methods for A Class of Structured Nonlinear Optimization

Speaker：Prof. Zhaosong Lu,  Simon Fraser University

Abstract：Nowadays the optimization problems emerging from some application areas such as machine learning and data mining are typically huge-scale. They have brought tremendous challenge to the traditional first- and second-order methods. In this talk we consider randomized block coordinate descent (RBCD) type of methods for solving these problems, whose iteration cost is typically low. We analyze iteration complexity of some RBCD methods and present some computational results.

Time：2017-06-28 10:30--11:30

Location：School of Mathematical Sciences