An optoelectronic architecture for multilayer learning in a single photorefractive crystal
(1990) In Neural Computation 2(1). p.25-34- Abstract
- We propose a simple architecture for implementing supervised neural network models optically with photorefractive technology. The architecture is very versatile: a wide range of supervised learning algorithms can be implemented including mean-field-theory, backpropagation, and Kanerva-style networks. Our architecture is based on a single crystal with spatial multiplexing rather than the more commonly used angular multiplexing. It handles hidden units and places no restrictions on connectivity. Associated with spatial multiplexing are certain physical phenomena, rescattering and beam depletion, which tend to degrade the matrix multiplications. Detailed simulations including beam absorption and grating decay show that the supervised learning... (More)
- We propose a simple architecture for implementing supervised neural network models optically with photorefractive technology. The architecture is very versatile: a wide range of supervised learning algorithms can be implemented including mean-field-theory, backpropagation, and Kanerva-style networks. Our architecture is based on a single crystal with spatial multiplexing rather than the more commonly used angular multiplexing. It handles hidden units and places no restrictions on connectivity. Associated with spatial multiplexing are certain physical phenomena, rescattering and beam depletion, which tend to degrade the matrix multiplications. Detailed simulations including beam absorption and grating decay show that the supervised learning algorithms (slightly modified) compensate for these degradations. (Less)
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/record/4e7193d1-3960-4355-bd56-a3151419d7ce
- author
- Peterson, Carsten LU ; Redfield, Stephen ; Keeler, James D. and Hartman, Eric
- organization
- publishing date
- 1990
- type
- Contribution to journal
- publication status
- published
- subject
- in
- Neural Computation
- volume
- 2
- issue
- 1
- pages
- 9 pages
- publisher
- MIT Press
- ISSN
- 1530-888X
- DOI
- 10.1162/neco.1990.2.1.25
- language
- English
- LU publication?
- yes
- id
- 4e7193d1-3960-4355-bd56-a3151419d7ce
- date added to LUP
- 2024-12-11 09:11:07
- date last changed
- 2025-04-04 13:56:51
@article{4e7193d1-3960-4355-bd56-a3151419d7ce,
abstract = {{We propose a simple architecture for implementing supervised neural network models optically with photorefractive technology. The architecture is very versatile: a wide range of supervised learning algorithms can be implemented including mean-field-theory, backpropagation, and Kanerva-style networks. Our architecture is based on a single crystal with spatial multiplexing rather than the more commonly used angular multiplexing. It handles hidden units and places no restrictions on connectivity. Associated with spatial multiplexing are certain physical phenomena, rescattering and beam depletion, which tend to degrade the matrix multiplications. Detailed simulations including beam absorption and grating decay show that the supervised learning algorithms (slightly modified) compensate for these degradations.}},
author = {{Peterson, Carsten and Redfield, Stephen and Keeler, James D. and Hartman, Eric}},
issn = {{1530-888X}},
language = {{eng}},
number = {{1}},
pages = {{25--34}},
publisher = {{MIT Press}},
series = {{Neural Computation}},
title = {{An optoelectronic architecture for multilayer learning in a single photorefractive crystal}},
url = {{http://dx.doi.org/10.1162/neco.1990.2.1.25}},
doi = {{10.1162/neco.1990.2.1.25}},
volume = {{2}},
year = {{1990}},
}