Abstract

An optical implementation of a second-order, translation-invariant neural network algorithm is described. The active devices are two arrays of integrated asymmetric Fabry–Perot modulators. The device and the array characteristics are given. The arrays are used, with passive optics, to provide optical input, matrix multiplication, and weighted interconnects in a cascaded optical system. The detailed parameters of the system are discussed, and the operation is assessed in terms of noise tolerance. Translation invariance is confirmed. Finally, the relative merits of the devices and algorithm are considered.

© 1994 Optical Society of America

Optical implementation of a translation-invariant second-order neural network for multiple-pattern classification

Sunao Kakizaki, Paul Horan, Akira Arimoto, Hiroshi Sako, Atsushi Saito, and Fumio Kugiya
Appl. Opt. 33(35) 8270-8280 (1994)

Optical implementation of a second-order translation-invariant network algorithm

Paul Horan, Andrew Jennings, Brian Kelly, and John Hegarty
Appl. Opt. 32(8) 1311-1321 (1993)

High-capacity neural networks on nonideal hardware

Leonard Neiberg and David Casasent
Appl. Opt. 33(32) 7665-7675 (1994)

Reversal-input superposing technique for all-optical neural networks

Yoshio Hayasaki, Ichiro Tohyama, Toyohiko Yatagai, Masahiko Mori, and Satoshi Ishihara
Appl. Opt. 33(8) 1477-1484 (1994)

Image identification system based on an optical broadcast neural network and a pulse coupled neural network preprocessor stage

Horacio Lamela and Marta Ruiz-Llata
Appl. Opt. 47(10) B52-B63 (2008)