Biological neural networks: beyond the PDP model

摘要

The biological neural networks (BNNs) of the gravity-sensingendorgans of the inner ear, the vestibular maculas, appear to beorganized in ways more complex than the parallel distributed processing(PDP) systems currently under study in the artificial neural network(ANN) field. The organization and synaptic interactions of theprocessing elements (PEs) have been documented by computer-assisted 3-Dreconstructions of terminal and receptive fields from electronmicrographs of a series of 570 sections and by the mapping of more than300 cells. The macular BNN properties are described. Data were reducedto symbolic representations and compared with features of ANN PEs. Thechief conclusions are that terminal and receptive fields of macular BNNsare not modular, since no two are identical in detail, and that thenetwork changes in threshold properties and in organizational complexityfrom site to site