Memory Capacity of a Novel Optical Neural Net Architecture

A new associative memory neural network which can be constructed using optical matched filters is described. It has three layers, the centre one being iterative with its weights set prior to training. The other two layers are feedforward nets and the weights are set during training. The best choice of central layer weights, or in optical terms, of pairs of images associated in a hologram is considered. The stored images or codes are selected carefully form an orthogonal set using a novel algorithm. This enables the net to have a high memory capacity equal to half the umber of neurons with a low probability of error.