Statistical Surrogate Model based Sampling Criterion for Stochastic Global Optimization of Problems with Constraints

Abstract

Several sequential surrogate model-based global optimization algorithms have been developed to increase the efficiency of commonly used global optimization algorithms as well as ensure the accuracy of optimization. In this paper, a sampling criterion based on statistical surrogate model for stochastic global optimization is proposed to achieve effectively the global optimization of problems with constraints. The proposed method searches for feasible region over design domain and finds several local optima and a global optimum to improve the cost while maintaining the feasibility by using stochastic information of statistical surrogate model. criterion based on statistical surrogate model for stochastic global optimization is proposed to achieve effectively the global optimization of problems with constraints. The proposed method searches for feasible region over design domain and finds several local optima and a global optimum to improve the cost while maintaining the feasibility by using stochastic information of statistical surrogate model.