Reservoir computing with signal-mixing cavities
Floris Laporte, Joni Dambre, Peter Bienstman · 2017
In an age where we get swamped by big data, new machine learning techniques for efficient high-speed data processing become more important than ever. One of these techniques, known as reservoir computing, is specifically designed for processing time-dependent data. We propose some new ideas for implementing such a reservoir computer on a silicon photonics chip for low-power and high-bandwidth optical communication applications. Our simulations show that this photonic reservoir can for example be used in pattern recognition tasks such as header recognition.