Abstract:Reverse design of highly GeO2-doped silica optical fibers with broadband and flat dispersion profiles is proposed using a neural network combined with a particle swarm optimization algorithm. Firstly, the neural network model designed to predict optical fiber dispersion is trained with an appropriate choice of hyperparameters, achieving a root mean square error (RMSE) of 1.15?10-6 on the test dataset, with a determination coefficient (R-squared) of 0.999. Secondly, the neural network is combined with the particle swarm optimization algorithm for the inverse design of dispersion-flattened optical fibers. To expand the search space and avoid particles getting trapped in local optimal solutions, the particle swarm optimization algorithm incorporates adaptive inertia weight updating and a simulated annealing algorithm. Finally, by using a proper fitness function, the designed fibers exhibit flat group velocity dispersion (GVD) profiles at 1400-2400 nm, where the GVD fluctuations and minimum absolute GVD values are below 18 and 7 ps/nm?km, respectively.