本质下确界的最优性条件及其相对熵算法实现  

Optimality condition and cross entropy implementation of essential infimum

作　　者：李惠永 邬冬华[2] 

机构地区：[1]上海大学理学院,上海200444 [2]上海大学管理学院,上海200444

出　　处：《应用数学与计算数学学报》2013年第4期478-490,共13页Communication on Applied Mathematics and Computation

基　　金：上海市自然科学基金资助项目(09ZR1411100);上海市重点学科建设资助项目(S30104)

摘　　要：为了研究带约束的本质下确界优化问题,介绍了m阶偏差积分并研究了它的性质,给出了其最优性条件和概念算法.基于极小化相对熵的技术,提出了一种实现算法,并有效地解决该优化问题.数值算例验证了算法的有效性.To study the constrained optimization with the essential infimum, the ruth moment deviation integral function is introduced, and its properties are ex- amined. Moreover, the optimality condition and the conceptual algorithm are also given to obtain the essential infimum. Based on minimizing the cross entropy, an implementable approach is adopted to solve the optimization problem efficiently. Numerical examples are given to illustrate the effectiveness of the algorithm.

关 键 词：相对熵 约束优化 偏差积分函数 精确不连续罚函数 本质下确界 

分 类 号：O224[理学—运筹学与控制论]