MODIFIED 2-DIMENSIONAL HAMMING NEURAL-NETWORK AND ITS OPTICAL IMPLEMENTATION USING DAMMANN GRATINGS

A 2-D neural network and its optical implementation are presented. the neural network consists of two interconnection layers. The first laye r is a modified Hamming net layer that calculates the Hamming distance between the input and each of the stored patterns. Instead of perform ing a winner-take-all operation, thresholding is performed within each hidden neuron. The second layer is a mapping network for pattern asso ciation. The presented neural network has fewer interconnections and a higher storage capacity than a fully interconnected Hopfield network. In comparison with muitilayer perceptrons, it has advantages such as easy and direct training and unipolar binary interconnection weights a nd therefore is more suitable for optical implementation with currentl y available optoelectronic devices. The neural network can be easily r econfigured when the training set needs to be updated or extra trainin g patterns need to be added into the training set, The attraction basi ns of stored patterns can be adjusted, and the input fault tolerance c an be enhanced locally. An optical system employing Dammann gratings h as been used in the preliminary experiments. Experimental results veri fied the feasibility of the neural network model as well as its optica l implementation. (C) 1996 Society of Photo-Optical Instrumentation En gineers.