@inproceedings{7a742e4da8014213b46fc7ba15b48a6e,
title = "SSFM-Assisted Residual Temporal Convolutional Network for WDM Optical Fiber Channel Modeling",
abstract = "This paper proposes an SSFM-assisted residual temporal convolutional network for fast WDM optical fiber channel modeling. Instead of directly learning the complete fiber channel response, the proposed method first uses a linear SSFM operator to model chromatic dispersion and attenuation, and then employs a residual TCN to learn the nonlinear discrepancy between the linear propagation output and the Full-SSFM reference. Simulation results show that the proposed SSFM-TCN achieves a BER of 1.35 {\texttimes} 103 and an NMSE of -31.2 dB at 1040 km, close to the Full-SSFM reference, while reducing the runtime from 12.84 s to 2.31 s per batch. These results demonstrate its potential as an efficient surrogate model for WDM transmission simulation.",
keywords = "SSFM, TCN, WDM, fiber nonlinearity, optical fiber channel modeling, residual learning",
author = "Xin Zhang and Yongjun Wang and Hongyuan Li and Dingxiang Shen and Wang, \{Shuai Hang\} and Qi Zhang and Xiangjun Xin",
note = "Publisher Copyright: {\textcopyright} 2026 IEEE.; 24th International Conference on Optical Communications and Networks, ICOCN 2026 ; Conference date: 20-07-2026 Through 23-07-2026",
year = "2026",
doi = "10.1109/ICOCN70692.2026.11656081",
language = "English",
series = "2026 24th International Conference on Optical Communications and Networks, ICOCN 2026",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
booktitle = "2026 24th International Conference on Optical Communications and Networks, ICOCN 2026",
address = "United States",
}