具有一类不可导多元约束函数的非线性规划神经网络模型

Neural networks for nonliner programming with a kind of nondifferential multivariable constraints

水资源系统中存在一类具有不可导但连续单调递增约束函数的非线性规划问题，针对这类约束函数为一元函数的非线性规划问题建立了基于Ｈｏｐｆｉｅｌｄ 连续模型的人工神经网络模型［１］ ，将神经网络模型扩展到这类约束函数含有多元变量的情况，仿真结果表明了扩展后的模型的有效性．

In water resources system, there exists a kind of nonlinear programming problem whose constraint functions are nondifferential but continuously monotone increasing. An artificial neural network model based on Continuous Hopfield Neural Network is presented by authors for the situation that this kind of constraint funtions are single variable. In this paper, an artificial neural network model is extended to the situation that this kind of constraint function are multivariable. Simulation demonstrates the extended model is effective.