Abstract

The effort proposed here is aimed at demonstrating the use of existing optical and electrooptical components in implementing adaptive neural network systems. Specifically, we propose a system which implements the outer product model^(1,2) of auto- or hetero-association. Both learning and association operation can be executed using this system which is based on the use of 1-D striped-electrode fast input modulators (based on the PRIMO technology⁽³⁾) coupled with a time integrating photoactivated liquid crystal light valve.

© 1989 Optical Society of America