OPTICAL LEARNING NEURAL-NETWORK USING PHOTOREFRACTIVE WAVE-GUIDES

A model of an optical neural network with learning ability is proposed . We numerically evaluate the learning ability of the proposed network by using parameters determined by experiments. Adaptive connections b etween artificial neurons are implemented using photorefractive (PR) w aveguides that can be optically modified by guided beams. The network consists of three layers and has bipolar weights within the limited ra nge. The bipolar weight is encoded as the difference between optical p ower transmittances of signal beams in two channels of the PR waveguid es. The adaptivity of the transmittance of PR waveguide is experimenta lly evaluated and is incorporated into the proposed network simulated in a computer. The proposed net-work is trained by a simplified local learning algorithm. Numerical results showed that the proposed three-l ayered network with six hidden neurons can solve the exclusive-or prob lem.