梯度网络平稳点集合的Hopfield稳定性与吸引域

Hopfield-stability of equilibrium point set for gradient neural networks and their attractive regions

给出梯度神经网络模型的电路实现 ,证明了网络平稳点集合的全局 Hopfield稳定性。对各个平稳点吸引域的几何估计得到了相关结果 ,这为网络初始输入的数量与方向判定提供了基本的量度参数准则。

Neural networks for problems of the global optimization were proposed and the global convergence of the equilibrium point set was given.The geometric symptoms of the attractive region for each stationary point were analyzed to give a basic rule for choosing the initial values of the proposed neural networks both in qualities and directions.