Design of fiber optic adaptive filters and neural networks

Abstract

There is a tremendous need for high speed optical signal processing devices for working with guided lightwave signals. Various adaptive filters and Adaline neural networks can be fabricated by combining an optical transversal filter based on tapped fiber optic delay lines and an analog electronic feedback. Utilizing the high speed of optical device and parallel interconnection properties of light these optoelectronic adaptive systems can be faster and more efficient. The learning rates and accuracy of such optical adaptive filters are affected by optical round-off errors and the speed of operation of the spatial light modulator and electronic components. By analyzing the sensitivity of optical adaptive filters, a set of design guidelines is derived.