Discrete all-positive multilayer perceptrons for optical implementation

All-optical multilayer perceptrons differ in various ways from the ideal neural network model. Examples are the use of nonideal activation functions, which are truncated, asymmetric, and have a nonstandard gain; restriction of the network parameters to non-negative values, and the limited accuracy of the weights. A backpropagation-based learning rule is presented that compensates for these nonidealities and enables the implementation of all-optical multilayer perceptrons where learning occurs under computer control. The good performance of this learning rule, even when using a small number of weight levels, is illustrated by a series of computer simulations incorporating the nonidealities.