摘要
在本文中,恒假定（H1）:f（x）,gi（x）,1≤i≤m,hj（x）,1≤j≤l为一阶连续可微函数. 上述（NP）问题,若用可行方向法等方法求解时,初始点必须是可行点,且在每一步迭代中,为了得到目标函数值下降而又可行的点。
In this paper, we consider the optimization problem with nonlinear constraints and com-bine linear programming with the penalty function to give an algorithm with arbitrary initialpoint. In every iterative step, we solve the linear programming to get the iterative direction d(x)which is the descending direction of the penalty function. The parameter of the penalty func-tion is given by the simplex multiplier and becomes a constant number when k is sufficientlylarge. Finaly, we give the proof of global convergence of this algorithm.
出处
《数值计算与计算机应用》
CSCD
北大核心
1992年第1期32-38,共7页
Journal on Numerical Methods and Computer Applications
基金
国家自然科学基金
