Approximate Optimization Using RBF Network and Genetic Range Genetic Algorithms (Proposal of Base Function and Basic Consideration)

Abstract

In this paper, we will propose integrated optimization system that is composed by RBF network as approximation tool and Genetic Range Genetic Algorithms as optimizer. This system has function of optimizing radius of basis function in optimization, indication of next searching points by using basis function and optimization by using Genetic Range Genetic Algorithms that is a step further of Adaptive Range Genetic Algorithms. Key success in approximation by using RBF network lies in giving good data sets. Here, “good data sets” means that they can update accuracy around local optimum points, and also they can give global information of original function's response surface. In that sense, indication function is quite important in this system. For that purpose, we introduce base function which is composed by approximation RBF network. We demonstrated the proposed system to a bench mark problem to show efficiency of the proposed method. In constraint optimization case, it works something like active set strategy, and we can reduce the number of function calls.