What Autoassociative Network Models May Tell us about the Neurobiology of Memory

摘要

Most theoretical work on associative memory networks is based on classical models using binary McCulloch-Pitts neurons in discrete time (Willshaw et al. 1969; Hopfield 1982). Over several decades this direction of research was extremely successful in deriving mathematically exact results on how the memory performance of such networks depends on system parameters such as network size, connectivity, sparseness, etc. However, owing to its high level of abstraction, its relevance for the understanding of neurobiological memory circuits is still very unclear.