Neural Probabilistic Shaping: Joint Distribution Learning for Optical Fiber Communications

TL;DR

This paper introduces an autoregressive learning method for probabilistic shaping in optical fiber communications, achieving significant information rate gains by jointly learning symbol distributions for nonlinear channels.

Contribution

It proposes a novel end-to-end learning approach that jointly models symbol distributions, improving information rates over traditional marginal distribution methods in fiber-optic systems.

Findings

Achieves 0.3 bits/2D information rate gain

Effective joint distribution learning for nonlinear channels

Improves over optimized marginal distribution methods

Abstract

We present an autoregressive end-to-end learning approach for probabilistic shaping on nonlinear fiber channels. Our proposed scheme learns the joint symbol distribution and provides a 0.3-bits/2D achievable information rate gain over an optimized marginal distribution for dual-polarized 64-QAM transmission over a single-span 205 km link.