Fig. 1. Overview of deep learning methods. (a) Application of deep learning methods in modeling nonlinear pulse propagation; (b) workflow of deep learning methods; (c) common neural network structures

Fig. 2. Deep learning applications in predicting the nonlinear ultrashort pulse propagation in optical fiber. (a) Predicting the ultrashort pulse propagation under GVD and SPM^[2]; (b) predicting the higher-order soliton propagation^[4]; (c) predicting the supercontinuum generation^[47]; (d) predicting the signal pulse propagation in optical fiber parametric amplification system^[48]

Fig. 3. Deep learning applications in solving the inverse problems of the ultrashort laser pulse propagation in optical fiber. (a) Reconstruction of ultrashort pulse based on the nonlinear spectral changes induced by SPM^[57]; (b) physics-based deep learning method for the reconstruction of the initial pulse^[59]; (c) predicting typical parameters of single soliton and soliton molecule^[62]; (d) estimation of M² factor for few-mode fibers^[64]