RBFネットワークによる逐次近似最適化 : サンプル関数の基礎的検討　

概要

論文の詳細を見る
One of the important issues on the Sequential Approximate Optimization (SAO) is the sampling strategy. The sampling strategy for SAO using the Radial Basis Function (RBF) network is proposed in this paper. The proposed sampling strategy consists of three parts, which are called the density function, the boundary function, and random sampling. In order to add the new sampling points effectively, the density function and the boundary fuction are constructed by the RBF network. The objective of the density function is to find the sparse region in the design variable space and is to add the new sampling points in this region. In the constrained optimization problems, at least, one or more constraints will be active. As the result, it is desirable to add the new sampling points on the constratins. The objective of the boundary function is to add the new sampling points on the boundary. In addition, the random sampling is also introduced to spread the search region. The algorithm of proposed sampling strategy is described in detail. Through the numerical examples, the validity is examined.