联想记忆离散Hopfield神经网络的样本容量和吸引域

Sample Size and Domain of Attraction of Discrete-Time Hopfield Neural Network for Associative Memory

对用于联想记忆的外积型设计离散Hopfield神经网络,给出了记忆样本容量与样本间Hamming距离以及神经元数目的具体关系,并论证了其在实际应用中的合理性。同时给出了样本的吸引域半径与样本的具体关系。特别地,对于样本正交的情况得到了两个推论。

We studied discrete-time Hopfield neural network based on vector product. As a result, we got the formula of sample size, Hamming distance and the number of neurons. Its rationality in practice was discussed. Furthermore, we got the fomula of the domain of attraction and samples. Particularly, as samples are perpendicular, we got two corollaries.