TR2018-047
Fiber Nonlinearity Equalization with Multi-Label Deep Learning Scalable to High-Order DP-QAM
-
- "Fiber Nonlinearity Equalization with Multi-Label Deep Learning Scalable to High-Order DP-QAM", Signal Processing in Photonic Communications (SPPCom), DOI: 10.1364/SPPCOM.2018.SpM4G.1, July 2018.BibTeX TR2018-047 PDF
- @inproceedings{Koike-Akino2018jul3,
- author = {Koike-Akino, Toshiaki and Millar, David S. and Parsons, Kieran and Kojima, Keisuke},
- title = {Fiber Nonlinearity Equalization with Multi-Label Deep Learning Scalable to High-Order DP-QAM},
- booktitle = {Signal Processing in Photonic Communications (SPPCom)},
- year = 2018,
- month = jul,
- doi = {10.1364/SPPCOM.2018.SpM4G.1},
- url = {https://www.merl.com/publications/TR2018-047}
- }
,
- "Fiber Nonlinearity Equalization with Multi-Label Deep Learning Scalable to High-Order DP-QAM", Signal Processing in Photonic Communications (SPPCom), DOI: 10.1364/SPPCOM.2018.SpM4G.1, July 2018.
-
MERL Contacts:
-
Research Areas:
Abstract:
We use deep neural network (DNN) to compensate for Kerr-induced nonlinearity in fiber-optic communications. The proposed DNN is scalable to high-order modulations by employing multi-label classification, achieving greater than 1.2 dB gain in nonlinear regimes
Related News & Events
-
NEWS MERL researcher presented an invited talk and 2 papers at Advanced Photonics Congress 2018 Date: July 2, 2018 - July 5, 2018
Where: Advanced Photonics Congress 2018
MERL Contacts: Toshiaki Koike-Akino; Kieran Parsons; Ye Wang
Research Areas: Communications, Signal ProcessingBrief- Three papers from the Optical Communication team were presented at Advanced Photonics Congress, held at ETH Switzerland from 2-5 July 2018. One of the papers was an invited talk of MERL's recent advancement in high-speed reliable coded modulation schemes based on polar coding. The other papers are related to fiber nonlinearity mitigation techniques based on pulse-shaping filter optimization and deep neural networks.