강건하고 정확한 방사기저함수를 위한 정규화 파라미터의 최적화

Abstract

This study aims to optimization of regularization parameter in radial basis function to construct robust and accurate meta-model. Regularization parameter is one of the factors that affect accuracy of a radial basis function. The selection of appropriate regularization parameter can improve accuracy of radial basis function. However it is hard to know appropriate regularization parameter in advance because it depends on the problem. In this study, we suggest construction of robust and accurate meta-model by finding the optimal regularization parameter which minimized generalized cross-validation error expressed as a closed form solution.