Abstract

Massively parallel systems, with millions of simple computing units, are becoming increasingly important in artificial intelligence and computer vision. Applications of such models include local operations like convolutions, cooperative computation as in stereopsis, and perceptual organization as evidenced by ambiguous figures. Massively parallel (connectionist) models appear to be an excellent method for formalizing these problems and computationally similar ones such as the resolution of ambiguous word meanings. One particularly attractive feature of these models is their direct relation to neurobiology.

© 1985 Optical Society of America