摘要
提出了一个具有正交化存储、双层、循环工作的神经网络模型,该模型具有异联想-自联想能力。用蒙特卡洛模拟研究表明,此模型的最大存储模式数M等于每层的神经元数N,而且具有较强的寻址能力。
A model of two-layer neural network with orthogonalized storage is proposed. A system based on the model will have a capability of both auto-associative and hetero-associative memories. Monte Carlo simulation shows that the maximum number of stored key patterns with the model can he up to N(the number of neurons in each layer of the network) and a good content-addressability is achieved.
出处
《红外与毫米波学报》
SCIE
EI
CAS
CSCD
北大核心
1991年第5期371-378,共8页
Journal of Infrared and Millimeter Waves
基金
国家自然科学基金
关键词
正交化存储
神经网络模型
存储模式
neural network, orthogonalized storage, auto-associated and heteroassociated capabilities.
