Comparison of Metaheuristic Algorithms for Supercontinuum Control

Machine learning algorithms are increasingly used in ultrafast photonics, enabling new functionalities and enhancing dynamical control [1]. The generation of a broadband supercontinuum (SC) involves complex nonlinear phenomena including soliton fission, dispersive wave generation and Raman frequency shift, which can be particularly difficult to control and optimize. In this context, metaheuristic algorithms offer an efficient and systematic approach compared to typical trial and error or numerical simulations, and it was recently shown that genetic algorithms (GA) can be efficiently used for the control of SC generation [2].