High-Speed Neuromorphic Computing Using Spin-Controlled VCSELs

Reservoir computing (RC) is a neuromorphic computing paradigm that exploits the transient dynamics of a nonlinear system (the reservoir) to map input data unto a high dimensional state space. Different nonlinear systems have been proposed to serve as reservoirs, with photonic systems being particularly well-suited for RC, due to their inherent parallelism and high speed [1] . One of such photonic implementations is based on a semiconductor laser subjected to optical feedback. The high dimensional state space of this scheme is distributed temporally and can be accessed by sampling the output power along the delay line.