A log-barrier method with Benders decomposition for solving two-stage stochastic linear programs (Q5943079)

The paper deals with two-stage stochastic linear programming problems in which the support of the corresponding probability measure is supposed to be finite, however very large. Evidently, such type of problems can correspond to scenaria approach in stochastic programming problems with a general support; however, then the problem with finite varying support can be considered. The aim of the paper is to propose a new derivative based decomposition solution method which uses the logarithmic barrier to smooth the objective function. The Newton method (or even higher order methods) can be employed to construct the numerical algorithm. In the paper, the global convergence of the solution for the proposed method as well as the convergence of the algorithm in polynomial time are proven. The paper is written in a very understandable way.