Reservoir Computing Using Multiple Lasers With Feedback on a Photonic Integrated Circuit

Abstract

We propose a scheme for reservoir computing using multiple semiconductor lasers with optical feedback arranged in parallel on a photonic integrated circuit, and we investigate the performance of reservoir computing numerically. The virtual nodes are obtained from the temporal waveforms of the outputs of the parallel reservoir lasers. We test the chaotic time-series prediction task, memory capacity, and nonlinear channel equalization task to investigate the performance of reservoir computing. We found that our scheme using multiple lasers outperforms that using a single laser with multiple delay times. Large memory capacity can also be obtained for the multiple lasers. Finally, we investigate the effect of parameter mismatch of the multiple lasers on reservoir computing performance.