High Capacity Associative Memory Models – Binary and Bipolar Representation

N. Davey, R. Frank, S. Hunt, R. Adams, and L. Calcraft (UK)


Associative memory, Hopfield Networks, Binary representation, Bipolar representation


Hopfield type associative memory networks usually use a bipolar representation. It is also possible to use a binary 0/1 representation, although in the standard model this lowers performance. This paper reports an empirical investigation into the performance of both binary and bipolar associative memories, trained using the simple perceptron learning rule. Such networks normally have much better performance than the standard model. It is found that the binary networks perform just as well as the bipolar networks, although they take significantly longer to train.

Important Links:

Go Back