Column-wise Block Coordinate Descent Approach for Orthogonal Constrained Optimization Problems
主 题: Column-wise Block Coordinate Descent Approach for Orthogonal Constrained Optimization Problems
报告人: Prof. Xin Liu(Chinese Academy of Sciences)
时 间: 2016-03-31 10:00 - 2016-03-31 11:00
地 点: Room 29, Quan Zhai, BICMR
We propose a column-wise block coordinate descent approach for solving a class of orthogonal constrained optimization problems. This approach combines a Gauss-Sedeil type of iteration with a multiplier symmetrization step to guarantee the stationarity satisfied at any cluttering point. We prove the global convergence of the proposed approach. Preliminary experiments illustrate that the new algorithm performs well and is of great potential.
