Fast-convergence Physics-informed Correlation-enhanced Neural Network in DML-DD Link for Analog RoF Fronthaul

Inhalt:
We propose a physics-informed correlation-enhanced neural network (CorrNet) for composite second-order distortion compensation in C-band DML-DD link. For 10-GHz 64-QAM signal A-RoF transmission over 15-km SSMF, CorrNet achieves 1-dB ROP sensitivity improvement compared with Volterra-based feedforward equalizer, and 72.7% epochs reduction compared with fully-connected NN.