Physics-Informed Neural Network Modeling of Soliton Pulses in Optical Communication Systems

The nonlinear Schródinger equation which models the pulse propagation in an optical fiber is solved using a physics-informed neural network for the case of soliton propagation. The prediction accuracy, measured against the exact solution (computed using the Runge-Kutta method), is found to be 2.223×10
⁻³
.