Machine learning-based Raman amplifier design

A multi-layer neural network is employed to learn the mapping between Raman gain profile and pump powers and wavelengths. The learned model predicts with high-accuracy, low-latency and low-complexity the pumping setup for any gain profile.