Intelligent gain flattening in wavelength and space domain for FMF Raman amplification by machine learning based inverse design

Abstract

We propose a machine learning based approach to design few-mode DRAs by using neural networks to optimize the pump wavelengths, powers and mode content in order to obtain flat gain spectrum with low mode-dependent gain (MDG). Based on the proposed intelligent inverse design method, amplification optimization for the random fiber laser based two-mode DRA can be achieved with gain flatness of 1.0 dB and MDG of 0.6 dB at 14.5 dB on-off gain level. For backward pumping four-mode DRA, gain flatness of 0.46 dB and MDG of 0.3 dB can be achieved at 12.5 dB on-off gain.

Model-Aware Deep Learning Method for Raman Amplification in Few-Mode Fibers

“…For all these reasons this unsupervised method is expected to be more useful in self-adaptive networks. This method is validated on different 4-mode fibers using a counterpropagating scheme with various numbers of Raman pumps, up to 8, predicting the required pump powers and wavelengths to generate gain profiles on the C+L band with average gain and tilt in the interval from 5 dB to 15 dB and from −1.425 dB to 1.425 dB, respectively; results show MDG and gain flatness comparable to those reported in [16], but on a larger bandwidth and quantifying the advantage of higher number of pumps also in terms of the reached root-mean-square error (RMSE).…”

“…For the purpose of RA they can consequently be treated as a unique mode [25], [26]. Conversely, linear mode coupling between different mode groups is weak and will be neglected here, like in [16], [25], [26].…”

“…This work does not use neither modelaveraging nor fine-tuning algorithms; therefore, memory and time requirements for both the training and the inference phase are substantially cut. For the 4-mode FMF, [16] showed encouraging results in terms of MDG and gain flatness; nevertheless, the analysis is limited to the C band only.…”

“…In the context of MDM, a similar approach to design flat gain profiles both for 2-mode and 4-mode fibers has been demostrated in [16]. This work does not use neither modelaveraging nor fine-tuning algorithms; therefore, memory and time requirements for both the training and the inference phase are substantially cut.…”

“…This aspect is also more problematic when increasing the amplification bandwidth or the number of modes and pumps, as the dimensionality of the space to explore also increases. The choice of parameters for the generation of the dataset is also critical for the effectiveness of the methods presented in [15], [16]. For example, the powers and wavelengths of the pumps are selected a priori, which requires preliminary supervision and that can finally mean that the trained NNs might not able to predict the optimal pump parameters.…”