Photonic delay systems as machine learning implementations
Nonlinear photonic delay systems present interesting implementation platforms for machine learning models. They can be extremely fast, offer great degrees of parallelism and potentially consume far less power than digital processors. So far they have been successfully employed for signal processing using the Reservoir Computing paradigm. In this paper we show that their range of applicability can be greatly extended if we use gradient descent with backpropagation through time on a model of the system to optimize the input encoding of such systems. We perform physical experiments that demonstrate that the obtained input encodings work well in reality, and we show that optimized systems perform significantly better than the common Reservoir Computing approach. The results presented here demonstrate that common gradient descent techniques from machine learning may well be applicable on physical neuro-inspired analog computers. ; P.B., M.H. and J.D. acknowledge support by the interuniversity attraction pole (IAP) Photonics@be of the Belgian Science Policy Office, the ERC NaResCo Starting grant and the European Union Seventh Framework Programme under grant agreement no. 604102 (Human Brain Project). M.C.S. and I.F. acknowledge support by MINECO (Spain), Comunitat Autónoma de les Illes Balears, FEDER, and the European Commission under Projects TEC2012-36335 (TRIPHOP), and Grups Competitius. M.H. and I.F. acknowledge support from the Universitat de les Illes Balears for an Invited Young Researcher Grant. ; Peer Reviewed