摘要
为了降低神经网络的连接复杂度,以利于硬件实现,本文提出了一个非全局连接的神经网络动力学模型,包括网络的能量函数、局域场和学习规则。基本思想是将Hopfield网络分解成若干个子网络,并建立各子网络之间的联系。对此模型的联想记忆过程统计分析表明,与Hopfield网络相比,一次分解后的连接数减少了25%,但网络的存储量和联想能力不变;多次分解后的连接数大大减少了,尽管网络的存储量和联想能力有所下降,但每个连接系数的平均存储能力提高了近1倍。本文最后讨论了高阶连接和多层网络结构的关系.
In order to reduce the complexity of neural network connectivity, a dynamical model for unfully interconnected neural networks, including energy function, local area field and learning rule, is presented. The basic idea is decomposing the Hopfield network into several subnetworks and setting up interconnections between them. The statistical analysis of the associative memory process shows that the number of interconnections after the first decomposition is reduced by 25% compared with that of the Hopfield n...
出处
《东南大学学报（自然科学版）》
EI
CAS
CSCD
1992年第5期1-6,共6页
Journal of Southeast University：Natural Science Edition
关键词
神经网络模型
联想记忆/非全局连接
neural network model
associative memory/unfull interconnection
