论文部分内容阅读
神经网络中与联想记忆密切相关的两个动力学性质是网络的稳定性和吸引性．针对高阶连接网络，首先讨论了通常所用的Ｈｅｂｂ规则下网络的稳定性和吸引性问题．由于Ｈｅｂｂ规则对于正交或近似正交的原型模式才有较好的联想记忆能力，故文中又针对一般线性无关模式，给出了保证这些模式为网络稳定平衡点的高阶连接权的秩１张量形式，该形式可看作高阶连接的伪逆（投影）规则．通过对网络稳定性和吸引性的分析，得到一些充分条件，这些条件对高阶网的设计和综合具有指导意义
The two dynamical properties that are closely related to associative memory in neural networks are the stability and attractiveness of the network. For high-order connected networks, we first discuss the stability and attractiveness of the network under the Hebb rule. Because Hebb rule has good associative memory ability for orthonormal or approximately orthogonal prototyping patterns, we also give rank 1 of high-order concatenation to ensure that these patterns are stable equilibrium point of network for general linear irrelevant mode Tensor form, this form can be regarded as pseudo-inverse (projection) rules of high-order connections. By analyzing the stability and attractiveness of the network, some sufficient conditions are obtained, which are instructive for the design and synthesis of higher-order networks