64-CHANNEL CORRELATOR IMPLEMENTING A KOHONEN-LIKE NEURAL-NETWORK FOR HANDWRITTEN-DIGIT RECOGNITION

We present an optical implementation of an improved version of the Koh onen map neural network applied to the recognition of handwritten digi ts taken from a postal code database. Improvements result from the int roduction of supervision during the learning stage, a technique that a lso simplifies the map layer labeling. The experimental implementation is based on a frequency-multiplexed raster computer-generated hologra m used to realize the required N-4 interconnection. The setup is shown to be equivalent to a 64-channel correlator. Computer simulations are used to study various detection and classification procedures. The re sults of the optical experiments, obtained with binary phase computer generated holograms, are presented and shown to be in excellent agreem ent with the simulations. (C) 1996 Optical Society of America