Attractor Neural Networks with Local Inhibition: from Statistical Physics to a Digitial Programmable Integrated Circuit

Part of Advances in Neural Information Processing Systems 5 (NIPS 1992)

Bibtex Metadata Paper

Authors

E. Pasero, R. Zecchina

Abstract

Networks with local inhibition are shown to have enhanced compu(cid:173) tational performance with respect to the classical Hopfield-like net(cid:173) works. In particular the critical capacity of the network is increased as well as its capability to store correlated patterns. Chaotic dy(cid:173) namic behaviour (exponentially long transients) of the devices in(cid:173) dicates the overloading of the associative memory. An implementa(cid:173) tion based on a programmable logic device is here presented. A 16 neurons circuit is implemented whit a XILINK 4020 device. The peculiarity of this solution is the possibility to change parts of the project (weights, transfer function or the whole architecture) with a simple software download of the configuration into the XILINK chip.