Prediction of soliton molecule evolution in passively mode-locked fiber laser using gated recurrent unit network

The formation and evolution of soliton molecules in a passively mode-locked fiber laser system is an extremely complex nonlinear dynamic process. In this paper, the gated recurrent unit (GRU) neural network model was adopted to predict the complex dynamic evolution process of soliton molecules in passively mode-locked fiber laser. The model accurately predicts the dynamic evolution of soliton molecules with different separations (10.15 ps, 26.07 ps, and 54.3 ps), as well as that of three-pulse soliton molecules, and effectively captures the multistability and bifurcation behavior of the system. Its prediction results are highly consistent with the nonlinear Schrödinger equation using the split-step Fourier numerical method, which greatly reduces the calculation cost. For all test samples in this study, the root mean square error is less than 0.05. Furthermore, it is shown that the model based on GRU achieves superior prediction accuracy and high computational efficiency compared with the conventional long short-term memory network model.