Computational Learning by an Optical Thin-Film Model

This paper describes a computational learning model inspired by the technology of optical thin-film multilayers from the field of optics. It can be used as an alternative to the widely used neural network models. With the thicknesses of thin-film layers serving as adjustable “weights” for the computation, the optical thin-film model is capable of approximating virtually any kind of nonlinear mapping. The thin-film multilayer structure and its computational algorithm are presented. A pattern recognition problem is used as an example to demonstrate the model’s learning capability.