Noisy GA resampling on evolved parameterized policies for stochastic constraint programming

Stochastic Constraint Programming is an extension of Constraint Programming for modeling and solving combinatorial problems which involve uncertainty in real world. Evolved Parameterized Policies (EPP) is the first incomplete approach to stochastic constraint problems which has higher performance rather than other methods, but still seems non-practical for large multi-stage problems due to scenarios exponentially growing. We proposed new resampling method called IDGAS based on Noisy GAs and other Evolutionary Computation algorithms, which aim to ensure the reliability while keeping in high search performance. In experiments on credit portfolio management with multi-stage, it performed more effective than conventional EPP and other resampling methods.