Storage capacity of non-monotonic neurons

Non-monotonic neurons have been shown to enhance the storage capacity of auto-associative memories. This work investigates the retrieval capacities of different types of non-monotonic neurons. It is found that storage capacity is maximized when the neuron response is a function with well defined geometrical characteristics. Numerical experiments demonstrate that storage capacity is directly related to the dynamical property of the iterative map, x(i) --> g(Sigma(j) T(ij)x(j)), that describes the network evolution. Maximum capacity is reached when the neuron dynamics are subdivided into two non-overlapping "erratic bands" around points x(i) = +/-1. The capacity improvements are explained in terms of the energy function associated with the network dynamics. (C) 1999 Elsevier Science Ltd. All rights reserved.