On a class of stochastic programs with exponentially many scenarios
We consider a class of stochastic programs whose uncertain data has an exponential number of possible outcomes, where scenarios are affinely parametrized by the vertices of a tractable binary polytope. Under these conditions, we propose a novel formulation that introduces a modest number of additional variables and a class of inequalities that can be efficiently separated. Moreover, when the underlying polytope is the unit hypercube, we present an extended formulation of polynomial size that can be solved directly with off--the--shelf optimization software. We assess the advantages and limitations of our formulation through a computational study.