High performance reservoir computing system based on VCSELs with variable polarization information injection
Yanting Liu, Guang-Qiong Xia, Qiupin Wang, Xulin Gao, Zheng-Mao Wu · 2023
In this work, we have demonstrated that, by introducing gradient boost technology, the parallel information processing performance can be improved in a reservoir computing (RC) system based on vertical-cavity surface-emitting lasers (VCSELs) under variable polarization information injection (VPII). For the VPII scheme, the information is added into X-polarization component (X-PC) and Y-polarization component (Y-PC) of VCSEL simultaneously, and outputs from two PCs of VCSEL are utilized to construct the node matrix. Although such an information injection method can double the processing rate, the system performance needs to be further improved. In this work, the gradient boost technology is adopted to enhance the system performance, for which an extra VCSEL-based reservoir is added to correct the residual error of the original reservoir. Via parallel processing Santa-Fe chaotic time series prediction task (Stask) and waveform recognition task (Wtask), the effectiveness of adopting gradient boost technology to improve the performance of VCSEL-based RC is confirmed.