Abstract

Optical implementations of neural net models incorporate summing and broadcasting as well as analog multiplication. Traditionally, these operations have been achieved by free space optical propagation and anamorphic optical elements (cylindrical lenses). An alternate approach uses elongated electrodes for photocurrent summation and broadcasting. Such electrooptic devices form the basis of compact optical neural net processors.¹ In this paper we describe the design and fabrication of these special purpose devices and present experimental results obtained with them.

© 1989 Optical Society of America