延迟离散Hopfield-型网络广义异步收敛性分析  

Convergent Analysis of Discrete Hopfield-Type Neural Networks with Delay in a Generalized Serial Mode

作　　者：邱深山[1] 刘永清[1] 刘明珠[2] 宁殿双 

机构地区：[1]华南理工大学自动化系,广州510641 [2]哈尔滨工业大学数学系,150001 [3]大庆石油管理局采油八厂科技科,163514

出　　处：《系统工程与电子技术》2000年第2期4-6,16,共4页Systems Engineering and Electronics

基　　金：国家自然科学基金!(1 9871 0 1 9);国防预研基金资助课题!(98J4 7 4HT0 1 0 8)

摘　　要：给出了延迟离散Hopfield -型神经网络的收敛性定理。在广义异步运行方式下 ,证明了对称连接权阵 (只要w0 对称 )条件下的收敛性定理 ,推广了已有的延迟离散Hopfield -型神经网络的收敛性结果 ,表明网络收敛滞后于能量函数收敛最多 2n+1步。最后给出了能量函数的极大值点与延迟离散Hopfield -型神经网络的稳定态的关系。In this paper, convergence theorems of discrete Hopfield-type neural networks with delay are obtained. We prove that any discrete Hopfield-type neural networks with delay will converge to stable state when operating in a generalized serial mode, which one of synaptic matrix is a symmetric matrix and generalize convergence theorem in earlier works. It is shown that network will reach a stable state at most 2n+1 steps after energy function associated with the network has converged. Finally we also relate maximum of modified energy function with stable state of neural networks with delay.\;

关 键 词：异步收敛性 HOPFIELD 神经网络 人工智能 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]