Design of Hopfield content-addressable memories
X. Zhuang, Yan Huang · 2002
The optimal learning rule for the Hopfield content-addressable memories (CAM) based on three well recognized criteria is designed. After analyzing the real cause of the unsatisfactory performance of the Hebb rule and many other existing learning rules, it is shown that three criteria actually amount to widely expanding the basin of attraction around each desired attractor. For this, a concept called Hamming-stability is introduced. It is found that Hamming-stability for all desired attractors can be reduced to a moderately expansive linear separability condition at each neuron. Thus, Rosenblatt's perceptron learning rule is the correct one for learning Hamming-stability. Computer experiments are conducted, showing that the proposed perceptron Hamming-stability learning rule takes good care of three optimal criteria.>