Development and properties of neural networks
Christopher L. Scofield, Leon N. Cooper · Contemporary Physics · 1985
Recent progress in a theory of the development of neural networks and their application to visual cortex is discussed. We find, that under proper assumptions, distributed memories display some of the features of animal memory. We discuss a mechanism, consistent with known neurophysiology, by which these memories develop and modify with experience. This mechanism is applied to the problem of the development of orientation selectivity and ocular dominance in visual cortex. The proposed mechanism of synaptic evolution is based upon the competition of incoming patterns rather than converging afferents. It is found that neurons develop maximal selectivity with respect to their signal environment, and under a variety of rearing conditions develop orientation tuning curves and ocular dominance comparable to that found in experiment. Finally, we describe a model first-order anatomy of visual cortex in which we implement this form of synaptic modification.