Solving semidefinite programming problems via alternating direction methods (Q2493944)

The author proposes an alternating direction method to solve the semi definite programming problem. In this respect the complementary conditions in the primal-dual optimality conditions are reformulated as a projection equation and based on this reformulation at each iterate is needed only to solve a linear system with reduced dimension and make one projection. It is proved that the algorithm converges to a solution of the semidefinite programming problem under weak conditions.