Paper 14092-82
Paper 14092-82
Information embedding and spectral correlations in extreme learning machines based on fiber supercontinuum generation
On demand | Presented live 14 April 2026 
Abstract
Photonic-based Extreme Learning Machines (ELMs) offer promising opportunities for fast and energy-efficient computing. Here we use the recently implemented generalized nonlinear Schrödinger equation simulation framework to systematically study fiber-based ELM performance exploring the impact of dispersion, nonlinearity, encoding, and different noise sources. By analyzing both numerical simulation and experimental data, we relate classification accuracy to correlations between output supercontinuum spectra and embedded information with a specific focus on physics governing the complex nonlinear dynamics in optical fibers. Our results highlight the rich parameter space determining fiber-based ELM performance and provide insight into the limits, optimization strategies, and practical implementation of these systems.
Presenter
Andrei V. Ermolaev
Université Marie et Louis Pasteur, FEMTO-ST Institute, CNRS UMR 6174 (France)
Andrei V. Ermolaev is a postdoctoral researcher at the FEMTO-ST Institute (CNRS, Université Marie et Louis Pasteur, Besançon, France), where he is actively working on the implementation of an extreme learning machine based on optical fiber propagation with a specific focus on physics governing the complex nonlinear dynamics of supercontinuum generation process. His research interests include nonlinear and ultrafast fiber optics, neuromorphic photonics, numerical and analytical modeling of complex light-matter interactions, and data-driven machine learning techniques. Before joining the FEMTO-ST Institute, he worked at the Corning Scientific Center on rigorous modeling of vector light scattering from rough dielectric surfaces for industrial applications, and previously conducted theoretical research on the resonant interaction of light with atomic vapor confined in nanometer-thick cells.
Also in the session