非线性极大极小问题的一个有效算法

摘要

针对一类非线性约束极大极小问题，利用极大熵方法将其转化为带等式、不等式约束的非线性规划问题，给出了一种梯度投影算法，解决了一般约束的非线性大系统优化问题，该算法初始点可任意；同时证明了该算法的全局收敛性。初步的数值试验表明，对于该类极大极小问题，算法有良好的数值表现。%In this paper, a gradient projection algorithm using maxim-entropy methods is analyzed and a class of max-min problems with nonlinear constraints are changed into nonlinear programming problems with inequality and equality constra. The algorithm which resolves general max-min problems is global convergent. Preliminary numerical experiments show that the proposed algorithm is effective.