OPTIMIZATION SEMINAR



Brainstorming Session on Stochastic Linear Programming




Marcus Emsermann and Allen Holder
University of Colorado at Denver
Department of Mathematics

Tuesday, Sep. 16, 1997, 12:00 noon
CU-Denver Bldg., Room 626


ABSTRACT:

Consider the stochastic linear program, min {cx:Ax=b,x >= 0}, where c_i is uniformly distributed on [a_i,b_i]. The use of the analytic center solution and subsequent parametric analysis is proposed to decide which decision variables are most likely to be positive. It is shown that if the measure of the supports of the uniform distributions are sufficiently small, then the analytic center solution yields immediate information about which decision variables are zero. Furthermore, parametric programming techniques are used to attain probabilistic estimates that indicate the most likely positive decision variable.