An alternating direction method for solving convex nonlinear semidefinite programming problems

Abstract

An alternating direction method is proposed for solving convex semidefinite optimization problems. This method only computes several metric projections at each iteration. Convergence analysis is presented and numerical experiments in solving matrix completion problems are reported.

Keywords:

• alternating direction method
• convex semidefinite programming
• matrix completion
• variational inequalities

AMS Subject Classifications::

• 65L20
• 65K10
• 90C25

Acknowledgements

This research is partially supported by STAR grant at the NUS School of Business, National University of Singapore. The research of Zhang Su is partially supported by the Fundamental Research Funds for the Central Universities under No. NKZXB10089.