School of Computer Science Intranet
S.B. Furber, W.J. Bainbridge, J.M. Cumpstey and S. Temple
An analysis is presented of a sparse distributed memory (SDM) inspired by that described by Kanerva (1988) but modified to facilitate an implementation based on spiking neurons. The mem- ory presented here employs sparse binary N-of-M codes, unipolar binary synaptic weights and a simple Hebbian learning rule. It is a two-layer network, the first (fixed) layer being similar to the `address decoder' in Jaeckel's (1989) `hyperplane' variant of Kanerva's SDM and the second (writeable) `data store' layer being a correlation matrix memory as first proposed by Willshaw et al (1969). The resulting network is shown to have good storage efficiency and is scalable. The anal- ysis is supported by numerical simulations and gives results that enable the configuration of the memory to be optimised for a range of noiseless and noisy environments.