RETRIEVAL PROPERTIES OF DILUTED ATTRACTOR NEURAL NETWORKS

We investigate the dependence of the retrieval properties of the pseud o-inverse and optimal attractor neural networks on the fraction of sto red patterns alpha, the temperature T and the margin parameter kappa. Phase diagrams in the full space of parameters are presented in the re gime of extreme dilution, i.e. when the connectivity C satisfies the c ondition C << ln N, where N is the number of neurons. Furthermore, we study analytically the neighbourhood of a stored pattern for both mode ls by calculating the average fraction of unstable sites epsilon in a pattern that differs by d sites from a given stored pattern. This anal ysis may shed light on the properties of the basins of attraction of t he stored patterns.