Prediction of Chaotic Intensities and Dynamics for an Optically Injected Semiconductor Laser Using Reservoir Computing

Prediction of continuous intensity time series and the underlying dynamical behaviors are investigated for chaotic optically injected semiconductor lasers. An intensity time series of about 20ns is first simulated by a rate-equation model, and is then utilized for training of a reservoir network. Upon careful optimization of the reservoir parameters, prediction of the continuous intensity time series is successfully demonstrated for over 0.8ns. Moreover, the dynamical behaviors of the chaotic laser are also successfully reproduced, including the microwave power spectrum, probability density function, autocorrelation function, and chaotic attractors.