Machine Learning for Ultrafast Nonlinear Fiber Photonics

The interplay of nonlinear and dispersive effects in guided photonics leads to complex and rich spectro-temporal dynamics. To better exploit these phenomena, machine learning now offers new and powerful tools. We present several approaches developed in ultrafast fiber photonics.

First, evolutionary algorithms—especially genetic algorithms—prove highly effective for experimental optimization in demanding systems, such as the exploration of novel regimes in fiber cavities: frequency-locked breathing solitons, noisy pulse trains, or rogue wave generation. Neural networks offer a second route. Once trained, they can predict temporal and spectral profiles with high accuracy, greatly reducing computational load—by over three orders of magnitude in some cases—while also solving inverse problems efficiently. Lastly, machine learning improves our understanding of nonlinear dynamics: classification, dynamic mode decomposition, data-driven dominant balance analysis, and sparse modeling techniques like SINDy enable the extraction of meaningful physical models from experimental or numerical data.

These methods are revolutionizing nonlinear photonics, expanding the accessible parameter space, enhancing physical interpretability, and enabling faster, smarter exploration and control of complex optical systems.

What You Will Learn:
• The basics of several machine-learning approaches applied to ultrafast nonlinear photonics
• Several examples of nonlinear dynamics in optical fibers

Who Should Attend:
• PhD students
• Young/early career researchers

About the Presenter: Christophe Finot from University of Bourgogne Europe, Dijon, France

Christophe FINOT is professor of Physics at the University Bourgogne-Europe (Dijon, France). Born in France in 1978, he graduated from the Institut d’Optique Graduate School (Paris-Saclay University in 2002), and received a PhD in Physics from the University of Bourgogne in 2005 before spending a year at the Optoelectronics Research Center, in Southampton (UK). Appointed associate professor in 2006, he was promoted to professor in 2010 and is currently vice-dean of the Faculty of Sciences and Technologies. He was elected junior member of the Institut Universitaire de France in 2017.

His main research areas dedicated to ultrafast photonics in fiber concern nonlinear optical shaping, all-optical information processing, extreme events, analogies between diffractive and dispersive optics as well as fiber lasers. More recently, he has taken an interest in the implementation of machine-learning approaches. He has co-authored more than 200 publications in international journals.