Inverse System Design Using Machine Learning: The Raman Amplifier Case

Abstract

 Users may download and print one copy of any publication from the public portal for the purpose of private study or research.  You may not further distribute the material or use it for any profit-making activity or commercial gain  You may freely distribute the URL identifying the publication in the public portal If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.

Experimental demonstration of arbitrary Raman gain–profile designs using machine learning

“…During the training stage, a multi-layer neural network, NN bw (·), is employed to learn the (inverse) mapping between the gain profiles, G and pump powers P. Thereafter, if the error is not acceptable a fine optimization based on gradient descent is applied to fine-adjust the pump powers and reduce the test error. The gradient descent uses another multi-layer neural network, NN f w (·), representing the (forward) mapping between P and G [9].…”

“…Recently, a machine learning (ML)-based framework for Raman amplifier design has been proposed [8,9]. The advantage of the proposed framework in [8,9] is that it can provide an ultra-fast and low-complexity pump powers and wavelengths allocation for the design of arbitrary Raman gain profiles.…”

“…The accuracy of the ML-based framework in predicting the pump powers to provide a set of arbitrary gain profiles is investigated by comparing the target gain profiles with the profiles experimentally obtained when applying the pump powers predicted by the framework. A vast statistically analysis of several thousand cases is conducted and error distributions are estimated showing that the ML-based framework proposed in [8,9] can effectively design Raman amplifier with limited errors.…”

“…Even though experimental results are shown only for the C-band in this paper, because of limitation in the available Raman pump wavelengths, we believe that main conclusions can be extended to larger bandwidth, i.e. C+L, as we have shown by simulation in [9]. Fig.1 shows the experimental setup used to generate the data-set for the ML-based Raman amplifier design.…”

Summary:

“…The talk was based on a recent survey [1] and a recent tutorial, [2] published by the speaker. Following the short introduction, she presented the results of the group's recent activities in the area of phase noise characterization for lasers and frequency combs [3,4], Raman amplifier inverse design and modelling [5][6][7], and auto-encoders for optical communication systems [8][9][10]. Additionally, to have a comprehensive view of the main applications of machine learning in optical communications and the current state-of-the-art, she discussed the key point of a few relevant surveys on the topic [11][12][13][14].…”