Stochastic optimization using simulated annealing with hypothesis test

Aiming at the non-deterministic property of stochastic optimization problems, a class of simulated annealing (SA) approach with hypothesis test (HT) is proposed for stochastic optimization. By using SA, the probability to be trapped in local minimum can be reduced by employing jumping probability and such behavior can be adjusted by controlling the temperature. Moreover, by using HT, solution performance can be reasonably estimated and solution quality can be identified reliably by hypothesis test, so that the repeated search can be reduced to some extent. The effectiveness of the proposed approach is demonstrated by the simulation results based on both stochastic numerical optimization problems and stochastic flow shop scheduling problems. Meanwhile, the effects of hypothesis test, performance estimation and noise magnitude on searching performance are also studied.