Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Learning to Interfere in Non-Orthogonal Multiple-Access Joint Source-Channel Coding

Selim F. Yilmaz, Can Karamanli, Deniz Gunduz

TL;DR

This work tackles multi-user wireless image transmission over a MAC with non-orthogonal access by proposing DeepJSCC-PNOMA, a joint source-channel coding framework that uses a shared encoder/decoder with user-specific projections to enable scalable, interference-exploiting transmissions. A progressive fine-tuning strategy doubles the number of users at each stage while preserving initial performance via orthogonal projections, enabling scaling to at least 16 users with minimal parameter overhead. Empirical results across CIFAR-10, TinyImagenet, Cityscapes, and Kodak demonstrate that DeepJSCC-PNOMA surpasses point-to-point DeepJSCC, TDMA, and various NOMA-based baselines in PSNR, MS-SSIM, and LPIPS under AWGN and Rayleigh channels, and it maintains robustness and fairness across users. The approach combines a multi-view autoencoder with CDMA-inspired projections, offering a practical, scalable path toward efficient multi-user wireless transmission with joint source-channel coding and semantic-awareness.

Abstract

We consider multiple transmitters aiming to communicate their source signals (e.g., images) over a multiple access channel (MAC). Conventional communication systems minimize interference by orthogonally allocating resources (time and/or bandwidth) among users, which limits their capacity. We introduce a machine learning (ML)-aided wireless image transmission method that merges compression and channel coding using a multi-view autoencoder, which allows the transmitters to use all the available channel resources simultaneously, resulting in a non-orthogonal multiple access (NOMA) scheme. The receiver must recover all the images from the received superposed signal, while also associating each image with its transmitter. Traditional ML models deal with individual samples, whereas our model allows signals from different users to interfere in order to leverage gains from NOMA under limited bandwidth and power constraints. We introduce a progressive fine-tuning algorithm that doubles the number of users at each iteration, maintaining initial performance with orthogonalized user-specific projections, which is then improved through fine-tuning steps. Remarkably, our method scales up to 16 users and beyond, with only a 0.6% increase in the number of trainable parameters compared to a single-user model, significantly enhancing recovered image quality and outperforming existing NOMA-based methods over a wide range of datasets, metrics, and channel conditions. Our approach paves the way for more efficient and robust multi-user communication systems, leveraging innovative ML components and strategies.

Learning to Interfere in Non-Orthogonal Multiple-Access Joint Source-Channel Coding

TL;DR

This work tackles multi-user wireless image transmission over a MAC with non-orthogonal access by proposing DeepJSCC-PNOMA, a joint source-channel coding framework that uses a shared encoder/decoder with user-specific projections to enable scalable, interference-exploiting transmissions. A progressive fine-tuning strategy doubles the number of users at each stage while preserving initial performance via orthogonal projections, enabling scaling to at least 16 users with minimal parameter overhead. Empirical results across CIFAR-10, TinyImagenet, Cityscapes, and Kodak demonstrate that DeepJSCC-PNOMA surpasses point-to-point DeepJSCC, TDMA, and various NOMA-based baselines in PSNR, MS-SSIM, and LPIPS under AWGN and Rayleigh channels, and it maintains robustness and fairness across users. The approach combines a multi-view autoencoder with CDMA-inspired projections, offering a practical, scalable path toward efficient multi-user wireless transmission with joint source-channel coding and semantic-awareness.

Abstract

We consider multiple transmitters aiming to communicate their source signals (e.g., images) over a multiple access channel (MAC). Conventional communication systems minimize interference by orthogonally allocating resources (time and/or bandwidth) among users, which limits their capacity. We introduce a machine learning (ML)-aided wireless image transmission method that merges compression and channel coding using a multi-view autoencoder, which allows the transmitters to use all the available channel resources simultaneously, resulting in a non-orthogonal multiple access (NOMA) scheme. The receiver must recover all the images from the received superposed signal, while also associating each image with its transmitter. Traditional ML models deal with individual samples, whereas our model allows signals from different users to interfere in order to leverage gains from NOMA under limited bandwidth and power constraints. We introduce a progressive fine-tuning algorithm that doubles the number of users at each iteration, maintaining initial performance with orthogonalized user-specific projections, which is then improved through fine-tuning steps. Remarkably, our method scales up to 16 users and beyond, with only a 0.6% increase in the number of trainable parameters compared to a single-user model, significantly enhancing recovered image quality and outperforming existing NOMA-based methods over a wide range of datasets, metrics, and channel conditions. Our approach paves the way for more efficient and robust multi-user communication systems, leveraging innovative ML components and strategies.

Paper Structure

This paper contains 38 sections, 9 equations, 19 figures, 4 tables, 4 algorithms.

Table of Contents

  1. Introduction
  2. Contributions
  3. Organization of This Article
  4. Related Work
  5. Non-Orthogonal Multiple Access (NOMA)
  6. Related Deep Learning Paradigms
  7. Problem Definition
  8. Methodology
  9. Point-to-Point DeepJSCC Architecture
  10. Parameter Sharing Between Users via Novel User-Specific Projections
  11. Training Procedure and Construction of Training Samples
  12. Novel Progressive Fine-tuning Method for Doubling the Number of Users
  13. Numerical Results
  14. Datasets
  15. Implementation Details
  16. ...and 23 more sections

Figures (19)

  • Figure 1: Extension of DeepJSCC to two users via .
  • Figure 2: Overall architecture of our DeepJSCC-PNOMA method.
  • Figure 3: Encoder architecture, user-specific projection layer and power normalization layer of the introduced method.
  • Figure 4: User-specific projection layer and decoder architecture of the introduced method.
  • Figure 5: Comparison with point-to-point DeepJSCC and separation-based methods.
  • ...and 14 more figures

Theorems & Definitions (8)

  • Remark 1: Extensibility
  • Remark 2: CDMA Analogy
  • Remark 3: Scalability
  • Remark 4: Shape Adaptivity
  • Remark 5: Parallelization
  • Remark 6: Comparing Methods with Different Number of Users
  • Definition 1: Fairness Objective
  • Definition 2: Comprehensiveness Objective