All-normal dispersion supercontinuum evolution prediction using a hybrid convolutional neural network-gated recurrent unit network

In this study, we propose a hybrid network integrating a convolutional neural network (CNN) and a gated recurrent unit (GRU) for the prediction of all-normal dispersion (ANDi) supercontinuum (SC). To evaluate the proposed model, we design a double-clad fiber (DCF) as the target nonlinear optical device. This DCF exhibits zero dispersion at 4905 and 4929 nm and achieves an SC bandwidth of 3000 nm under 6000 W and 200 fs pumping. Trained by a dataset generated by this DCF, the proposed model achieves a mean squared error of 0.0054 and a mean absolute error of 0.027. Meanwhile, an ablation study, k-fold cross-validation, and generalization verification are further conducted to validate the model’s capacity. The results demonstrate that the hybrid CNN-GRU model can achieve superior SC prediction performance with fewer computational resources and less time consumption, and the proposed model can facilitate the design and optimization of nonlinear optical devices.