延迟离散Hopfield神经网络的一般演化规则(英文)

A Generalized Updating Rule for Discrete Hopfield Neural Network with Delay

从优化计算模型的角度研究延迟离散Ｈｏｐｆｉｅｌｄ网络，提出了它的一般演化规则（ＧＵＲＤ）．在任何演化序列下，证明了ＧＵＲＤ单调收敛到网络的稳定状态，已有的延迟神经网络的演化算法为其特例，同时ＧＵＲＤ的收敛条件已经突破了已有的限制，有更大的应用空间．新的演化规则和收敛条件保证了延迟神经网络收敛到能量函数的局部最大值．

In the paper, the Hopfield neural network with delay is studied from the standpoint of taking it as optimization computational model. A generallied updating rule for the network (GURD) is established. It is proved that in any sequence of updating rule modes , the GURD nonotonously converges to a stable state of the network. All ordinary HNND algorithms are instances of the GURD. It is shown that the convergence conditions of GURD expand into more loose conditions in applications and break through original constrained conditions. New updating nile mode and restrictive conditions can guarantee the network to achieve local maximum of the energy function.