Showing 1–2 of 2 results for author: Mrabet, H

A Survey of Applied Machine Learning Techniques for Optical OFDM based Networks

Authors: Hichem Mrabet, Elias Giaccoumidis, Iyad Dayoub

Abstract: In this survey, we analyze the newest machine learning (ML) techniques for optical orthogonal frequency division multiplexing (O-OFDM)-based optical communications. ML has been proposed to mitigate channel and transceiver imperfections. For instance, ML can improve the signal quality under low modulation extinction ratio or can tackle both determinist and stochastic-induced nonlinearities such as… ▽ More In this survey, we analyze the newest machine learning (ML) techniques for optical orthogonal frequency division multiplexing (O-OFDM)-based optical communications. ML has been proposed to mitigate channel and transceiver imperfections. For instance, ML can improve the signal quality under low modulation extinction ratio or can tackle both determinist and stochastic-induced nonlinearities such as parametric noise amplification in long-haul transmission. The proposed ML algorithms for O-OFDM can in particularly tackle inter-subcarrier nonlinear effects such as four-wave mixing and cross-phase modulation. In essence, these ML techniques could be beneficial for any multi-carrier approach (e.g. filter bank modulation). Supervised and unsupervised ML techniques are analyzed in terms of both O-OFDM transmission performance and computational complexity for potential real-time implementation. We indicate the strict conditions under which a ML algorithm should perform classification, regression or clustering. The survey also discusses open research issues and future directions towards the ML implementation. △ Less

Submitted 7 May, 2021; originally announced May 2021.

Comments: 26 pages, 4 figures and 4 tables

Fiber-induced nonlinearity compensation in coherent optical systems by affinity propagation soft-clustering

Authors: Elias Giacoumidis, **long Wei, Ivan Aldaya, Christian Sanchez, Hichem Mrabet, Liam P. Barry

Abstract: There is a tremendous need to maximize signal capacity without sacrificing energy consumption and complexity in optical fiber communications due to the rise of network traffic demand and critical latency-aware services such as telemedicine and the Internet-of-Things. In this work, the first system-agnostic and training-data-free fiber nonlinearity compensator (NLC) using affinity propagation (AP)… ▽ More There is a tremendous need to maximize signal capacity without sacrificing energy consumption and complexity in optical fiber communications due to the rise of network traffic demand and critical latency-aware services such as telemedicine and the Internet-of-Things. In this work, the first system-agnostic and training-data-free fiber nonlinearity compensator (NLC) using affinity propagation (AP) clustering is experimentally demonstrated. It is shown that compared to linear equalization, AP can increase the signal quality-factor up to about 5 dB in 16-quadrature-amplitude-modulated coherent optical systems by effectively tackling deterministic and stochastic nonlinearities. AP outperforms benchmark clustering algorithms and complex deterministic schemes such as K-means, fuzzy-logic, digital-back propagation and Volterra-based NLC, offering a power margin extension of up to 4 dB. △ Less

Submitted 21 December, 2018; v1 submitted 12 December, 2018; originally announced December 2018.