Fast Compting Algorithm of Constrained Minimization

Abstract

목적 함수의 1차 도함수를 구하는 과정을 피하기 위하여 패턴무브(Pattern move)를 기본기술로 한 계산방법을 제시하였다. 기본적 구조는 SUMT와 같으나 Pattern move의 시행과정중에 Penalty function의 무게치(weighting value)를 삽입하여 계산과정이 적어지는 이점을 알고리즘을 개발하였다.

In finding out theminimun value of an objective function with constraints, the pattern move is strongly applicable to ill-behavior functions.

In this paper a fast computing algorithm for minimization of an objective function is proposed using pattern move. The basic idea comes from reducing penalty term weighting factor according to one cycle of pattern move.

In finding out theminimun value of an objective function with constraints, the pattern move is strongly applicable to ill-behavior functions.

In this paper a fast computing algorithm for minimization of an objective function is proposed using pattern move. The basic idea comes from reducing penalty term weighting factor according to one cycle of pattern move.