Optimization of recurrent neural network-based pre-distorter for coherent optical transmitter via stochastic orthogonal decomposition

Moshe Tzur, Gil Paryanti, Dan Sadot

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

Abstract

A gradual training paradigm for Recurrent Neural Network-based pre-distorter is proposed. Stochastic decomposition is used to separate nonlinearity and quantization noise features. Performance improvement of more than 6dB is presented.

Original languageEnglish
Title of host publicationSignal Processing in Photonic Communications, SPPCom 2020
PublisherOSA - The Optical Society
ISBN (Print)9781943580798
StatePublished - 1 Jan 2020
EventSignal Processing in Photonic Communications, SPPCom 2020 - Washington, United States
Duration: 13 Jul 202016 Jul 2020

Publication series

NameOptics InfoBase Conference Papers
VolumePart F191-SPPCom 2020

Conference

ConferenceSignal Processing in Photonic Communications, SPPCom 2020
Country/TerritoryUnited States
CityWashington
Period13/07/2016/07/20

ASJC Scopus subject areas

  • Electronic, Optical and Magnetic Materials
  • Mechanics of Materials

Fingerprint

Dive into the research topics of 'Optimization of recurrent neural network-based pre-distorter for coherent optical transmitter via stochastic orthogonal decomposition'. Together they form a unique fingerprint.

Cite this