Linear shallow neural network to accelerate transmitter dispersion eye closure quaternary (TDECQ) assessment
摘要
We have demonstrated a data-driven TDECQ assessment scheme based on L-SNN. In comparison with existing DL-based schemes, the proposed L-SNN can achieve the lowest computation complexity with only 210 multiplications. The MAE of the L-SNN scheme for 25 and 50 Gbaud PAM-4 optical signals is experimentally verified to be 0.13 and 0.15 dB, respectively, over the TDECQ range of 1.5–4.0 dB, which has reached the accuracy threshold of 0.25 dB recommended by the IEEE standard.