登入選單
返回Google圖書搜尋
A Hopfield Network with Exponential Storage Ability
註釋A method for storing a large number of exemplars in a Hopfield neural network is presented. Specifically, it is shown how a network of N neurons can store as many as N!/[(N/2)!]2 exemplars. However, this network has poor noise immunity because the Hamming distance between exemplars is very small. Therefore, a method is also presented for altering the original neural network by modifying its transmission matrix to reduce the number of exemplars and obtain better noise immunity.