Optical neural net for matrix inversion

Abstract
A novel optical neural net for performing matrix inversion is presented. This algorithm does not suffer from accumulative round-off error. Various optical architectures are introduced, and the circumstances in which each of them is advantageous are investigated.