Modelling self-similar parabolic pulses in optical fibres with a neural network

Sonia Boscolo,John M Dudley,Christophe Finot

doi:10.1016/j.rio.2021.100066

Abstract

We expand our previous analysis of nonlinear pulse shaping in optical fibres using machine learning [Opt. Laser Technol., 131 (2020) 106439] to the case of pulse propagation in the presence of gain/loss, with a special focus on the generation of self-similar parabolic pulses. We use a supervised feedforward neural network paradigm to solve the direct and inverse problems relating to the pulse shaping, bypassing the need for direct numerical solution of the governing propagation model.

Highlights

Machine learning is transforming the scientific landscape, with the use of advanced algorithmic tools in data analysis yielding new insights into many areas of fundamental and applied science [1]
In our previous work [24], we presented a solution to this problem using a supervised machine-learning model based on a feedforward neural network (NN) to solve both the direct and inverse problems relating to pulse shaping in a passive, lossless fibre, bypassing the need for numerical solution of the governing propagation model
We extend the use of our model-free method and show that a feedforward NN can excellently predict the behaviour of nonlinear pulse shaping in the presence of distributed gain or loss

Summary

Introduction

Machine learning is transforming the scientific landscape, with the use of advanced algorithmic tools in data analysis yielding new insights into many areas of fundamental and applied science [1]. From a control and feedback perspective, various machine-learning strategies have been deployed to design and optimise mode-locked fibre lasers [5,6,7,8,9], optimise optical supercontinuum sources [10], analyse the complex nonlinear dynamics occurring during the buildup of supercontinuum [11,12,13], and control pulse shaping [14] In parallel with these developments, pulse shaping based on nonlinear propagation effects in optical fibres has developed into a remarkable tool to tailor the spectral and temporal content of light signals [15, 16], leading to the generation of a large variety of optical waveforms such as ultra-short compressed pulses [17], parabolic- [18], triangular- [19, 20] and rectangular- [21] profiled pulses. The network is able to infer the key characteristics of parabolic pulses with remarkably high accuracy, and to successfully handle variations of the pulse parameters over more than two orders of magnitude

Governing propagation model and neural network

Model-free modelling of nonlinear pulse shaping

Findings

Conclusion