Joint Constellation and Labeling Design for Gaussian Interference Channels via Autoencoders

Razavi Pour, Seyed Reza; Vaezi, Mojtaba

Permalink

https://hdl.handle.net/2142/130298 Copy

Description

Title

Joint Constellation and Labeling Design for Gaussian Interference Channels via Autoencoders

Author(s)

• Razavi Pour, Seyed Reza
• Vaezi, Mojtaba

Issue Date

2025-09-17

Keyword(s)

• Interference channel
• Deep learning
• Constellation design

Abstract

We propose a learning-based modulation design and symbol labeling scheme for the two-user interference channel (IC) with arbitrary channel gains. To this end, we develop autoencoders (AEs) with trainable mappers and demappers, jointly optimized in an end-to-end manner. We show that initialization of the transmitter constellations plays a key role in achieving stable convergence and improving performance. Our analysis shows that the proposed AEs can learn effective modulation schemes across a wide range of interference levels and signal-to-noise ratios by jointly performing bitwise clustering through suitable labeling and maximizing the minimum symbol distances to reduce the impact of interference in the IC. We further incorporate low-density parity-check (LDPC) channel codes and demonstrate that the proposed AE-IC with LDPC outperforms existing methods in terms of bit error rates.

Publisher

Allerton Conference on Communication, Control, and Computing

Series/Report Name or Number

2025 61st Allerton Conference on Communication, Control, and Computing Proceedings

ISSN

2836-4503

Type of Resource

Text

Genre of Resource

Conference Paper/Presentation

Language

eng

Handle URL

https://hdl.handle.net/2142/130298&&

Copyright and License Information

Copyright 2025 is held by Seyed Reza Razavi Pour and Mojtaba Vaezi.

Owning Collections