Graph Neural Network Pooling for BCH Channel Decoding in Software Defined Satellites

This paper delves into the transformative impact of Deep Learning (DL) techniques on decoding tasks at the physical layer onboard regenerative software-defined satellites, reshaping traditional error correction methodologies. Specifically, we focus on the integration of Graph Neural Networks (GNNs)...

Full description

Saved in:
Bibliographic Details
Main Authors: Varadarajulu, Swetha, Baeza, Victor Monzon, Querol, Jorge, Mendonca, Marcele O. K., Chatzinotas, Symeon
Format: Conference Proceeding
Language:English
Subjects:
BCH codes
channel decoding
Conferences
Decoding
GNN
Graph neural networks
Physical layer
Program processors
Satellites
Software
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This paper delves into the transformative impact of Deep Learning (DL) techniques on decoding tasks at the physical layer onboard regenerative software-defined satellites, reshaping traditional error correction methodologies. Specifically, we focus on the integration of Graph Neural Networks (GNNs) for channel decoding, which offers a fresh perspective by adeptly handling graph-structured data and effectively modelling intricate interference and channel dependencies. The study systematically explores the potential performance tradeoffs that arise from modifying the graph structure. Furthermore, we extend our investigation by implementing the message-passing algorithm with GNN, employing a topk pooling method following pick, prune, and link optimization strategies. This strategic approach aims to mitigate computational complexity and minimize latency, by 30 to 35 % which is particularly advantageous for decoding BCH codes. This advancement promises to enhance the efficiency of communication systems sianificantly,
ISSN:2694-2941
DOI:10.1109/ICCWorkshops59551.2024.10615375