Self-Supervised Neural Mutual Information Estimator for Probabilistic Shaping Signals in Fiber-Optic Systems

  • Yifan Cai
  • , Qinghua Tian*
  • , Zuxian Li
  • , Fangxu Yang
  • , Sitong Zhou
  • , Feng Tian
  • , Qi Zhang
  • , Xiangjun Xin
  • *Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

A self-supervised neural mutual information estimator for probabilistic shaping (PS) signals is proposed, achieving high accuracy in additive white Gaussian noise (AWGN) channels (|ΔMI|<0.01) and fiber-optic channels (|ΔMI|<0.03) under varying conditions, outperforming existing estimators.

Original languageEnglish
Title of host publication2025 23rd International Conference on Optical Communications and Networks, ICOCN 2025
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9798331548759
DOIs
Publication statusPublished - 2025
Externally publishedYes
Event23rd International Conference on Optical Communications and Networks, ICOCN 2025 - Zhangjiajie, China
Duration: 28 Jul 202531 Jul 2025

Publication series

Name2025 23rd International Conference on Optical Communications and Networks, ICOCN 2025

Conference

Conference23rd International Conference on Optical Communications and Networks, ICOCN 2025
Country/TerritoryChina
CityZhangjiajie
Period28/07/2531/07/25

Keywords

  • deep learning
  • mutual information
  • optical fiber communication
  • probabilistic shaping

Fingerprint

Dive into the research topics of 'Self-Supervised Neural Mutual Information Estimator for Probabilistic Shaping Signals in Fiber-Optic Systems'. Together they form a unique fingerprint.

Cite this