Abstract

Photorefractive materials are attractive for use as high density distributed storage media due to their dynamic response and high diffraction efficiency. Real-time holography is of particular interest in the implementation of optical neural networks. One of the key attractions of neural processors is their ability to “learn” input-output mappings by adapting distributed connections between processors. In an optical neural system, this adaptation typically involves updating holographic interconnections. Photorefractive crystals are useful for this application because photorefractive holograms may be updated continuously with no degradation in the holographic response.

© 1989 Optical Society of America