An optoelectronic architecture for multilayer learning in a single photorefractive crystal

Summary, in English

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.

• Författare: Carsten Peterson; Stephen Redfield; James D. Keeler; Eric Hartman
• Avdelning/ar: Beräkningsvetenskap för hälsa och miljö; Centrum för miljö- och klimatvetenskap (CEC); Department of Astronomy and Theoretical Physics - Has been reorganised
• Publiceringsår: 1990
• Språk: Engelska
• Sidor: 25-34
• Publikation/Tidskrift/Serie: Neural Computation
• Volym: 2
• Issue: 1
• Länkar: http://dx.doi.org/10.1162/neco.1990.2.1.25
• Dokumenttyp: Artikel i tidskrift
• Förlag: MIT Press
• Ämne: Bioinformatics and Systems Biology
• Aktiv: Published
• Forskningsgrupp: Computational Science for Health and Environment
• ISBN/ISSN/Övrigt: ISSN: 1530-888X