Integrative Optimization by RBF Network and Particle Swarm Optimization

摘要

This paper presents a method for the integrative optimization system. Recently, many methods for global optimization have been proposed. The objective of these methods is to find a global minimum of nonconvex function. However, large numbers of function evaluations are required, in general. We utilize the response surface method to approximate function space to reduce the function evaluations. The response surface method is constructed from sampling points. The RBF Network, which is one of the neural networks, is utilized to approximate the function space. Then Particle Swarm Optimization (PSO) is applied to the response surface. The proposed system consists of three parts: (Part 1) generation of the sampling points, (Partrn2) construction of response surface by RBF Network, (Partrn3) optimization by PSO. By iterating these three parts, it is expected that the approximate global minimum of nonconvex function can be obtained with a small number of function evaluations. Through numerical examples, the effectiveness and validity are examined.