Machine learning enabled Raman amplifiers

Ultra-wideband (UWB) optical communication systems, envision to operate in O+E+S+C+l band, are a viable solution to cope with the network’s exponential traffic growth [1] . One of the main challenges to provide beyond C-band transmission is a lack of optical amplifiers. Since the erbium-doped fiber amplifiers (EDFAs) are limited to C and L bands only, new technologies will have to be explored to cover the remaining bands. Some examples of amplifiers able to provide amplification beyond C–band are: bismuth doped fibre amplifiers (BDFA) [2] , semiconductor optical amplifiers, (SOAs) [3] and Raman amplifiers (RAs) [4] . Compared to the solutions based on BDFA and SOA, optical amplifiers based RAs offer a higher degree of commercial maturity [5] . Most importantly, RA amplifiers can provide gain in any band provided a proper allocation of pump powers and wavelength