Time-delayed reservoir computing based on an optically pumped spin VCSEL for high-speed processing

We report on the first numerical implementation of photonic reservoir computing (RC) based on an optically pumped spin vertical-cavity surface-emitting laser (spin VCSEL) with optical feedback and injection. The proposed RC aims at both fast, single task processing and parallel tasks processing, benefiting from feasible tunability and multiplexing of the left and right circularly polarized modes. We evaluate its prediction and classification abilities through two benchmarks, i.e., a Santa Fe time series prediction task and a waveform recognition task. In particular, both the influence of external and internal parameters on the prediction and classification performance is systematically analyzed. The numerical results show that the proposed RC based on a spin VCSEL has remarkable prediction and classification abilities over wider parameter ranges due to the feasible adjustment of the pump intensity and polarization as compared to conventional VCSELs. Most importantly, because of its intrinsic fast response, the spin VCSEL-based RC system is capable of enhancing the information processing rate by significantly reducing the allowable feedback delay time and virtual node interval, reaching 20 Gbps for single task processing and 10 Gbps for parallel tasks processing, respectively. Such a spin VCSEL-based RC system has a potential to achieve high-speed information processing and lower power consumption.