Abstract

Optical approaches to vector-matrix multiplication have been proposed for 25 years.¹ However the development of a versatile, accurate implementation has been lagging. The free space architecture used by Goodman, et al.² suffered the disadvantage of a fixed photographic matrix mask. Some of these disadvantages were eliminated through the use of acousto-optic² or electro-optic³ devices. The accuracy of free space systems is critically dependent on the quality of large optics. Alignment, stray light, and temperature fluctuations are difficult to control. The use of fiber optics using a lattice structure approach for vector-matrix multiplication was introduced in 1985.⁴ These operations were limited to Toeplitz matrices, and mechanical adjustments were necessary to change the matrix elements. Other approaches have included the use of photorefractive media⁵ and optical implementations of special arithmetics.⁶

© 1989 Optical Society of America