GAIN AND EXPOSURE SCHEDULING TO COMPENSATE FOR PHOTOREFRACTIVE NEURAL-NETWORK WEIGHT DECAY

A gain and exposure schedule that theoretically eliminates the effect of photorefractive weight decay for the general class of outer-product neural-network learning algorithms (e.g., backpropagation, Widrow-Hof f, perceptron) is presented. This schedule compensates for photorefrac tive diffraction-efficiency decay by iteratively increasing the spatia l-light-modulator transfer function gain and decreasing the weight-upd ate exposure time. Simulation results for the scheduling procedure, as applied to backpropagation learning for the exclusive-OR problem, sho w improved learning performance compared with results for networks tra ined without scheduling.