Demonstration of spectrum narrowing mitigation based on recurrent neural networks for ultra-dense WDM networks

Traversing each WSS in ultra-dense WDM networks narrows the signal spectra. Simulations and experiments demonstrate, for the first time to our knowledge, spectrum narrowing mitigation based on RNN. Numerical simulations show that the RNN-based demodulation with impairment-aware optical path control significantly enlarges the transmission distance. Transmission experiments in the extended C-band successfully confirm an extension of the transmissible distance of 16QAM signals by over 500 km.