Simulated annealing approach based on hypothesis test for stochastic optimization problems
Da-Zhong Zheng · Kongzhi yu juece · 2004
According to the non-deterministic property of stochastic optimization problems, a class of simulated annealing approach based on hypothesis test is proposed. In the approach, reasonable estimated performance is provided by multiple evaluations, and repeated search is decreased by hypothesis test. The case of being trapped in local minimum can be avoided by jumping probability. The jumping ability can be adjusted by controlling the temperature. The effects of hypothesis test, the performance estimation and the magnitude of noise on the performance of the approach are studied, and the effectiveness and robustness of the proposed approach are demonstrated.