Abstract

The biological brain is able to form various feature maps and abstractions of sensory signals. For a characteristic sensory stimulus, a response is obtained from such a map at a location which corresponds to some quantitative feature value of the stimulus. This paper offers an explanation for this ability, using an idealized model of a self-organizing collective system. The model consists of a 2-D array of identical processing elements which receive a set of sensory signals in parallel and change their sensitivity or tuning to these signals, controlled by the input signals and also by the reactions from the neighboring elements. As a result, the various elements in the array become automatically adjusted in a continuous 2-D order. The coordinates on the array then correspond to some feature dimensions that are present in the input signals. For example, such a map may represent optical features, colors, spatial frequencies, acoustical frequencies, phonemes, taxonomic classifications, depending on the signal detectors used and stimuli presented at the input. This paper describes in what conditions a collective system starts to organize itself in this way and also gives several examples of maps already produced in computer simulations.

© 1985 Optical Society of America