A first-order block-decomposition method for solving two-easy-block structured semidefinite programs (Q892382)

The paper deals with a new first-order block-decomposition method which is proposed to minimize the sum of convex differentiable functions and convex non-smooth functions. More especially, the authors study the performance of a Block-Decomposition (BD) method based on the BD-hybrid proximal extra-gradient. The main contribution of the paper is two-fold: firstly, the authors introduce an adaptive choice of stepsize for performing an extra-gradient step and then they use a scaling factor to balance the blocks. The method is used to solve four broad classes of conic semi-definite programming, and numerical results illustrate its efficiency.