一种以环路作为存储单元的联想记忆神经网络

A New Associative Memory Neural Network Using Loop as Storage Unit

文章提出了一种新型联想记忆神经网络,每个模式被存储在一个通过网络中所有神经元的环路中,连接包括逻辑状态和一组神经元编号,网络中处理和传递的信号为神经元编号组成的序列,神经元执行一组处理这种序列的符号和逻辑运算;网络记忆容量为2N-2N、完全消除了假模式、同时具有更高的记忆效率和可靠性。

This paper presents a new associative memory neural network,each pattern is stored in a loop which pass through all neurons in the network,a link includes a logic state and a set of neuron numbers,the signals transferred and processed in the network are sequences composed of neuron numbers,each neuron performs a set of symbol and logic operations on such sequence.The network's storage capacity is2 N -2N,it completely eliminates spurious patterns,and has higher storage efficiency and reliability.