Abstract

Apart from any applications we may find for them, neural networks in and of themselves are fascinating systems. Their ability to learn, self-organize, and generalize has that seductive quality which challenges us to scrutinize and understand them. The match between optical systems and the requisite properties of neural networks seems so well made that many of us believe that optics, or optoelectronics, is an excellent means to implement them. Moreover, the mathematics of some neural network models have striking similarities to the physics of certain nonlinear optical systems, so perhaps we can also gain insight into neural networks by taking advantage of the analogies.

© 1989 Optical Society of America

PDF Article