Adapted and Adaptive Properties in Neural Networks for Visual Pattern Discrimination: A Neurobiological Analysis Toward Neural Engineering

Abstract

A topic of interdisciplinary research in neurobiology and neuroinformatics concerns visual pattern recognition by neuronal networks. Drawing on quantitative studies of visual releasers of prey catching in toads, it can be shown that moving objects are classified based on an evaluation of certain configurational features. The information regarding these features is provided in the manner of parallel distributed processing within a retino-pretectal-tectal interacting network. This processing structure is, to a considerable extent, modifiable and adaptive. Associative and nonassociative learning processes take advantage of loop operations involving various forebrain structures. An artificial neuronal net, applying some principles of the toad's visual system, is tested to promote the dialogue between neurobiology and engineering.