An injection laser neural network
Jean J. H. B. Schleipen, S. Colak, Evert C. Mos, Coen T. H. F. Liedenbaum · 2002
In this study an optical laser neural network is described, using longitudinal modes of an injection laser as individual neurons. Neural action is provided by external optical feedback into the active layer of the laser. Sigmoidal thresholding of weighted inputs is governed by injection laser physics under conditions of optical feedback. Training of the network is done by supervised stochastic learning algorithms. Ultimate network response times are estimated to exceed the 10 giga connections per second (GCPS) range.