Complex-Valued Reservoir Computing with Enhanced Sparsity for Nonlinear Equalization in Optical Fiber Transmission Systems
Takumi Yamamoto, Kai Ikuta, Yuta Ito, Takahide Sakamoto, Naokatsu Yamamoto, Moriya Nakamura · 2024
Through numerical simulations and experiments, this research investigated the performance of a complex-valued reservoir-computing-based (CVRC) nonlinear equalizer, designed to mitigate fiber-optic nonlinearity. The CVRC was optimized by introducing sparsity not only in the reservoir layer but also in the input and output layers. The sparse connectivity of the input and output layers could be reduced to about 50% without significant performance-degradation, when the number of reservoir units was 200.