Reservoir computing system based on an optoelectronic oscillator with double delay feedback

A time-delayed reservoir computing (RC) system based on an optoelectronic oscillator with double delay feedback is proposed. On the basis of a conventional RC system based on an optoelectronic oscillator, an optical delay feedback path is introduced, through which the output light of the system is fed back to the Mach-Zehnder modulator (MZM). A word error rate of 0 for signal waveform classification task and a normalized mean square error of 0.0097 for Santa Fe chaotic time series prediction task are obtained. The simulation results verify the prominent classification and prediction performances of the proposed RC system. The system is compact in structure and easy to implement.