BIOLOGICAL PLAUSIBILITY OF SYNAPTIC ASSOCIATIVE MEMORY MODELS

Observations in brains of neuronal networks that subserve associative learning in living organisms have been exceedingly sparse until the pa st decade. Recently, some fundamental biophysical and biochemical prop erties of biological neural networks that demonstrate associative lear ning have been revealed in the marine mollusc, Hermissenda crassicorni s. In mammals, we have localized distributed changes, specific to asso ciative memory, in dendritic regions within biological neural networks . Based on these findings, it has been possible to construct an artifi cial neural network, Dystal (dynamically stable associative learning) that utilizes non-Hebbian learning rules and displays a number of usef ul properties, including self-organization; monotonic convergence; lar ge storage capacity without saturation; computational complexity of O( N); the ability to learn, store, and recall associations among arbitra ry, noisy patterns after four to eight training epochs; a weak depende nce on global parameters; and the ability to intermix training and tes ting as new training information becomes available. The performance of the Dystal network is demonstrated on problems that include face reco gnition and hand-printed Kanji classification. The computational linea rity of Dystal is demonstrated by its performance on a MasPar parallel hardware computer