An optical outer product processor and its applications

In this thesis, we present a quadratic polynomial processor utilizing polarization encoding. The polynomial is evaluated using an outer product of the input with itself followed by a generalized inner product with a coefficient matrix. Analytical and experimental results on this polynomial processor will be presented. Four applications have been investigated for this processor, and results and design criteria for each of these applications are also presented. The applications include the quadratic perception neural network, linear and quadratic Hopfield associative memories, Walsh and Haar Transforms and optical logic. A breakeven point for using this parallel polynomial processor versus a serial microcomputer is established in terms of the size of a problem. Finally, some suggestions for future work in this research area are made.