Deep Learning of Geometric Constellation Shaping including Fiber Nonlinearities

TL;DR

This paper introduces a novel geometric shaping method using unsupervised machine learning to optimize optical fiber communication constellations, effectively mitigating nonlinear effects and improving data transmission efficiency.

Contribution

It presents a new unsupervised learning-based geometric shaping approach that accounts for fiber nonlinearities, outperforming traditional methods.

Findings

Achieves up to 0.13 bit/4D gain in simulated fiber channels

Mitigates fiber nonlinear effects through learned constellation design

Demonstrates effectiveness with a simplified fiber channel model

Abstract

A new geometric shaping method is proposed, leveraging unsupervised machine learning to optimize the constellation design. The learned constellation mitigates nonlinear effects with gains up to 0.13 bit/4D when trained with a simplified fiber channel model.