Fig. 1. The structure of a coherent optical transmission simulation system

Fig. 2. The processing of dataset construction

Fig. 3. Classification standards of DL-based optical fiber transmission system channel waveform modeling schemes. (a) Long-haul transmission mode; (b) deep learning model structure; (c) incorporating physical prior knowledge

Fig. 4. Framework diagram of BiLSTM-based fiber channel model^[41]

Fig. 5. Structure of CGAN^[42]

Fig. 6. Fourier neural operator framework^[75]

Fig. 7. Process of modeling nonlinear dynamics in fiber optics using PINN^[57]. (a) Temporal-spatial 2D region and exact evolution data generated by SSFM as a reference; (b) discrete-points sampling for PINN; (c) the PINN structure constructed for solving the NLSE in fiber

Fig. 8. Coherent WDM system and FDD modeling scheme details^[67]. (a) Schematic of WDM simulation system; (b) transmission processes of FDD-based and SSFM-based optical fiber channel models; (c) generation process for the training dataset; (d) FDD-BiLSTM structure for nonlinearity modeling

Fig. 9. Optical fiber channel modeling frameworks based on the Seq2Seq-FDD scheme^[81]. (a) The structure and parameters of optical fiber transmission WDM simulation systems and feature decoupled distributed schemes; (b) the structures of Bi-LSTM employed in Vanilla-FDD and Seq2Seq-FDD; (c) Seq2seq framework implemented by transfer learning processing

Fig. 10. Optical fiber communication simulation system based on S-SSFM-FNO^[82]

Fig. 11. Comparison of Q-factor of signals modeled using SSFM and DL-based optical fiber channel models under different launch powers, after processing with CDC, 10 StPS DBP, and 50 StSP DBP algorithms