Abstract: | This project analyses Kanerva's Sparse Distributed Memory (SDM). The SDM has been proposed alongside the work of Marr and of Albus as a possible theory of how the cerebellum functions. However, in its original form it is a static and inflexible system requiring supervised input parameters. this project discusses the biological plausibility of the SDM, and proposes a number of more biologically credible versions. These retain the essential characteristics of the original SDM, whilst providing the memory with a greater scope for plasticity and self-evolution. The new models are not as dependent upon a priori input values, giving them an increased robustness to learn either random or correlated input patterns.
|