A level-value estimation method and stochastic implementation for global optimization (Q1949597)

This paper presents a new method, namely the level-value estimation method (LVEM), for solving the continuous global optimization problem. The authors first define the variance function and the mean deviation function. Both functions depend on a level value of the function to be minimized. These functions have some good properties for global optimization. Based on solving the variance equation by the Newton method, the LVEM is proposed and the convergence result is proved. The authors give a brief introduction of importance sampling and the updating mechanism of sample distribution based on the main idea of the cross-entropy method. An estimator of the lower bound of sample size is obtained under some suitable conditions. Finally, an implementable level-value estimation method (ILVEM) is presented along with the convergence result. Some classical global optimization problems are used to test the performance of the ILVEM and numerical results are reported. The numerical results indicate that the proposed method solves global optimization problems efficiently.