Two stage stochastic programming problems with recourse have applications in a wide variety of areas. The principal investigator and his collaborators have developed a parallel scheme that can take advantage of novel computer architecture to compute both exact and approximate values and subgradients of the necessary functions associated with two-stage stochastic programs. Based on that work the focus of this project is on overall algorithms for two stage stochastic programs with recourse. Specifically, they propose to develop decomposition algorithms suited for parallel processing and to integrate the above schemes for parallel computation of values and subgradients of necessary functions into implementations of such decomposition algorithms. Progress from this proposal will result in better algorithms which will provide faster more accurate results in linear programming appearing in industry and government. Real life uncertainties will be taken into account by the stochasticity of the method.
