Data-driven vector soliton solutions of coupled nonlinear Schrödinger equation using a deep learning algorithm

Abstract In this work, we explore a deep learning algorithm for vector solitons of the coupled nonlinear Schrodinger equation (CNLSE). Based on the original physics-informed neural networks (PINN), we propose a pre-fixed multi-stage training algorithm by combining the ideas of error measurement, multi-stage training and adaptive weights. The result of numerical simulation demonstrates that the improved algorithm not only can recover different dynamical behaviors of solitons in the coupled equation but also has better approximation ability and faster convergence rate.