一种新型联想记忆神经网络

提出了一种新型联想记忆神经网络,记忆容量接近2^N,完全消除了假模式,具有多模式回忆能力,同时具有较高的记忆效率;模式的各个分量分别存储在网络的N个连接中,这些连接构成一个通过网络中所有神经元的被称为“模式环”的单环回路;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;连接由“连接状态”和“禁止路径”组成,前者用于存储信息,后者用于抑制网络中出现的假模式。

A New Type of Associative Memory Neural Network

This paper presents a new associative memory neural network, which is close to 2N, and has the capability to eliminate spurious patterns almost completely. It can also retrieve a set of patterns at one time and has higher memorizing efficiency. Each pattern is stored in N links, these links form a single loop by the name of "pattern loop", which can pass through every neuron. Signals transferred and processed in the network are sequences consisting of neuron numbers; each neuron performs a set of symbol and logic operation on such sequence. A link is composed of the "link state" and "inhibited path"'; the former is used to store information and the latter to eliminate spurious patterns.