用神经元网络的模拟计算方法求解Klee－Minty问题  

Solving the Klee-Minty problem using a simulation computational method of neural network

作　　者：陈剑飞 夏绍玮[1] 

机构地区：[1]清华大学自动化系

出　　处：《清华大学学报（自然科学版）》1995年第4期109-114,共6页Journal of Tsinghua University(Science and Technology)

摘　　要：介绍了一种改进的求解线性规划问题的神经元网络──上ＴＨ神经元网络，分析了该网络稳定点的特性，在此基础上提出关于它的一种比较新颖的模拟计算方法，并用该模拟计算方法求解了著名的Ｋｌｅｅ－Ｍｉｎｔｙ问题。理论分析和实例计算都表明该方法比传统的模拟计算方法欧拉法有效。n improved neural network for linear programming problems is introducedwhose domain is The properties of theneural network's stable point are analysed. Based on the analyses a new simulationcomputational method is proposed, and the method only needs to search a very limitednumber of points which at least satisfies part properties of the networki S stable point forfinding the optimal solution of the solved linear programming problem. Finally, thesimulation computational method is used to solve the famous Klee-Minty problems. It isshown by theoretical analyses and a computational example that the simulationcomputational method is much more efficient than the traditional Euler method.

关 键 词：线性规划 神经网络 模拟计算法 Klee-Minty问题 

分 类 号：O242.1[理学—计算数学]