A new approach to discrete stochastic optimization problems

Solving a discrete stochastic optimization problem involves two efforts: one is to sample and search the design space; and the other is to estimate the performance values of the sampled designs. When the amount of computational resources is limited, we need to know how to balance these two efforts in order to obtain the best result. In this paper, two performance measures which quantify the marginal contribution of the searching process as well as the performance evaluation process are proposed. Using these two measures, we recommend a framework that can dynamically allocate the computational resources to the search process and the performance evaluation process. Two algorithms based on the proposed framework are tested on several scenarios, and the results produced are very promising.