摘要
把基于信息理论的极大熵原理应用到互补问题的一种等价的不动点格式中,构造了一种光滑的迭代算法,给出了迭代算法收敛的充要条件.对于对称单调的互补问题,给出了收敛的充分条件.扩充了原来的梯度投影法要求严格单调的限制,得出了与外梯度投影法类似的收敛性条件.最后,给出了数值算例,包括标准互补问题考题和随机生成的考题,并给出了该光滑迭代算法与外梯度投影法的数值比较.
By using an equivalent fixed point format in the complementarity problems, where the maximum entropy principle of information theory is adopted, a smoothing method of iteration is constructed and, furthermore, the sufficient and necessary conditions of convergence of this iteration algorithm are given. For the complementarity problems with symmetric monotonicity, the corresponding sufficient and necessary conditions of convergence are given, also; so that the original gradient projection method with the restriction of strict monotonicity is extended, resulting in the convergence conditions similar to those of extra-gradient projection method. Finally, illustrative examples of numerical computation are given, including standard examination questions of complementarity problems, randomly generated examination questions, and a comparison between the smoothing iterative algorithm presented and extra-gradient projection method with respect to their numerical results.
出处
《甘肃工业大学学报》
北大核心
2002年第1期102-106,共5页
Journal of Gansu University of Technology
基金
国家基础研究重大项目(G1999032805)
关键词
信息熵
极大熵原理
互补问题
收敛性
凝聚函数
迭代算法
information entropy
maximum entropy principle
complementarity problems
convergency
surrogate functions
