Optical neuromorphic computing based on chaotic frequency combs in nonlinear microresonators

In this work, we present an unique implementation of delay line-free reservoir computing based on state-of-the-art photonic technologies, which exploits chaotic optical frequency comb formation in an optical microresonator as the nonlinear reservoir. Our solution leverages the high resonator quality (Q)-factor both for memory and for enhancing high-dimensional nonlinear mapping of input symbols. We numerically demonstrate the accurate prediction of about one thousand symbols in chaotic time series without the need of dedicated optimization for specific tasks. Our results will enable design of optical neuromorphic computing architectures combining on-chip integrability, low footprint, high speed, and low power consumption.