Fig. 1. System performance of 400G and 800G systems under different numbers of channels

Fig. 2. Structure of DBP^[15]

Fig. 3. Structures of various DBP-based improved algorithms. (a) Filtering DBP algorithm^[17]; (b) iDBP algorithm^[21]; (c) DBP algorithm based on CB-ESSFM^[22]; (d) WDBP algorithm^[19]; (e) MDBP algorithm^[18]; (f) ADBP algorithm^[23]

Fig. 4. The block diagram of transmitter-side nonlinear pre-compensation based on perturbation theory^[26]

Fig. 5. Structures of NLEE^[36] and FD-VSNE^[42]

Fig. 6. DSP Implementation of VNLE^[45]. (a) Traditional parallel structure; (b) C1-VNLE; (c) C2-VNLE

Fig. 7. Schematic diagram of LDBP principle^[47]

Fig. 8. LDBP structures with different architectures. (a) LDBP structure combined with distributed PMD compensation^[50]; (b) DU-DBP-NPCC structure^[51]; (c) PA-LDBP algorithm^[52]; (d) perturbation-based LDBP for SCM systems^[53]; (e) M-LDBP and DDLMS hybrid architecture for WDM systems^[54]; (f) SCM-LDBP^[55]

Fig. 9. Neural network structure for nonlinear compensation based on perturbation method^[58]

Fig. 10. Neural network equalizer structures. (a) Photonic neural network^[80]; (b) ResNet^[81]; (c) CVNN^[82]; (d) RC-SATCN^[83]

Fig. 11. Equalizer structure based on center-oriented long short-term memory network^[89]

Fig. 12. Equalizer structures based on neural network operators. (a) FNO^[102]; (b) F-FNO^[103]; (c) MiKNO^[104]; (d) PINO^[105]

Fig. 13. Autocorrelation coefficient of polarization crosstalk and phase noise caused by XPM on X-polarization after 600 km transmission^[115]

Fig. 14. 3-channel LDBP structure^[121]

Fig. 15. Flowchart of NPCC algorithm^[125]

Fig. 16. Flowchart of DF-NPCC algorithm^[126]