Separable Convex Optimization Research Articles

New Decomposition Method for Nonconvex Separable Optimization Problems

The primal-dual approach has been very effective in decomposing large-scale separable convex optimization problems. However, without convexity, this decomposition technique may necessitate complex solution strategies requiring three levels of optimization.In this paper, we offer a new strategy for nonconvex problems which requires only two levels of optimization. This new strategy is based on a new Lagrangian function, which consists of the standard Lagrangian function and penalty functions. A new optimization method, which makes use of the new Lagrangian function, is presented and it is shown to be locally convergent.