개선된 최소환산거리 최대화법

Abstract

In order to efficiently optimize the problem involving complex computer codes or computationally expensive simulation, surrogate models are widely used. And because their accuracy significantly depends on sample points, a number of design methods are proposed. One approach is sequential design of experiments which can consider existing information of responses. In earlier research, correlation coefficients of kriging model are introduced as weight parameters to define scaled distance between sample points. However, if existing information is wrong or lacking, new sample points can be misleading. Thus, our goal in this paper is to generate new points robustly by adjusting the weighting parameters derived from correlation coefficients. To verify the performance of the proposed method, several existing design methods are compared by mathematical examples.