Abstract

A figure of merit, the probability a bit is correct after update, is used to evaluate the performance of randomly coded outer-product associative memory models. Networks with bipolar binary states and nonzero diagonal connections are shown to yield the best performance with respect to this figure of merit. A surprising result is that an all-positive network, one with binary states and positive connections, is superior to a standard Hopfield style network with binary states and bipolar connections. A prescription for the optimal threshold point for the all-positive network is given.

© 1989 Optical Society of America

Hopfield model associative memory with nonzero-diagonal terms in memory matrix

Gene R. Gindi, Arthur F. Gmitro, and K. Parthasarathy
Appl. Opt. 27(1) 129-134 (1988)

Holographic associative memory based on adaptive learning including outer-product learning

Ho Hyung Suh and Sang Soo Lee
Appl. Opt. 31(2) 199-204 (1992)

Projection method formulations of Hopfield-type associative memory neural networks

M. Ibrahim Sezan, Henry Stark, and Shu-Jen Yeh
Appl. Opt. 29(17) 2616-2622 (1990)

Optical inner-product implementation of neural networks models

P. Lalanne, P. Chavel, and J. Taboury
Appl. Opt. 28(2) 377-385 (1989)

Optical associatron: a simple model for optical associative memory

Masatoshi Ishikawa, Naohisa Mukohzaka, Haruyoshi Toyoda, and Yoshiji Suzuki
Appl. Opt. 28(2) 291-301 (1989)