摘要
带约束的极大极小问题是一类不可微优化问题，通常的解法是通过增加约束将其转化为可微优化问题，极大熵方法是一种用光滑函数逼近最大值函数的方法；基于这种方法，给出一种求解带一般约束的极大极小问题的逼近方法，并针对凸规划问题证明了这种方法的收敛性，即当控制参数趋于正无穷时，近似问题的最优解收敛于原问题的最优解。
Approximation methed is one of the effective methOds to complex optimizationproblems. For general constrained minimax problems,the paper presents an approximationmethod──maximum entropy methd, The characteristic of this method is as follows:multi-constrained minimax problem can be approximated by a minimization problem of a smcothfunctinn which subjects to at most two constraints.The practicability of this method depends onthe relation between the optimum solutions of the original problem and of the approximateproblem。 The authors have proved that,for the constrained convex minimax problems,theoptimum solution x_p of the approximate problem converges to the solution of the originalproblem.
出处
《大连理工大学学报》
CAS
CSCD
北大核心
1995年第6期764-769,共6页
Journal of Dalian University of Technology
基金
国家教委博土点基金资助项目
关键词
凸规划
收敛
极大熵
convex programming
convergence/maximum entropy methOd
