Scenario-Based Stochastic Programs: Strategies for Deleting Scenarios

The proposed strategies for deleting scenarios are based on postoptimality analysis of the optimal value function with respect to probabilities of the included scenarios. These strategies can be used to reduce the size of the large scenario based problems or of the problems constructed in the course of specific numerical procedures, such as stochastic decomposition or scenario aggregation. A convex nonsmooth optimization problem is replaced by a sequence of line search problems along recursively updated rays. Convergence of the method is proved and applications indicated