Abstract

An anti-Hebbian local learning algorithm for two-layer optical neural networks is introduced. With this learning rule, the weight update for a certain connection depends only on the input and output of that connection and a global, scalar error signal. Therefore the backpropagation of error signals through the network, as required by the commonly used back error propagation algorithm, is avoided. It still guarantees, however, that the synaptic weights are updated in the error descent direction. With the apparent advantage of simpler optical implementation this learning rule is also shown by simulations to be computationally effective.

© 1992 Optical Society of America

Design and analysis of fixed planar holographic interconnects for optical neural networks

Paul E. Keller and Arthur F. Gmitro
Appl. Opt. 31(26) 5517-5526 (1992)

Multilayer optical learning networks

Kelvin Wagner and Demetri Psaltis
Appl. Opt. 26(23) 5061-5076 (1987)

Trainable optically programmed neural network

Richard G. Stearns
Appl. Opt. 31(29) 6230-6239 (1992)

Optical neural network with bipolar neural states

Xu-Ming Wang and Guo-Guang Mu
Appl. Opt. 31(23) 4712-4719 (1992)

Optical learning neural network with a Pockels readout optical modulator

Masahiko Mori, Yutaka Yagai, Toyohiko Yatagai, and Masanobu Watanabe
Appl. Opt. 37(14) 2852-2857 (1998)