Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Classification-Oriented Semantic Wireless Communications

Emrecan Kutay, Aylin Yener

TL;DR

The paper addresses the challenge of transmitting semantic meaning for classification tasks over wireless channels under latency and data constraints. It compares memory-based semantic processing (semantic quantization and semantic compression) with a learning-based vector quantized autoencoder (VQ-VAE) approach, both built on pre-trained semantic extractors (SBERT/CLIP), and introduces system time efficiency $\eta_T(\tilde{T})$ to quantify time saved by avoiding training. Key findings show memory-based methods achieve higher system time efficiency and comparable accuracy to semantic-agnostic baselines, while the learning-based VQ-AE is more sensitive to training data, with results validated on text (AG's News) and image (STL-10) modalities over AWGN and Rayleigh fading channels. The work demonstrates that leveraging past realizations as codebooks enables fast, interpretable, and resource-efficient semantic communications suitable for dynamic wireless environments, with potential benefits for 6G-era task-oriented links.

Abstract

We propose semantic communication over wireless channels for various modalities, e.g., text and images, in a task-oriented communications setup where the task is classification. We present two approaches based on memory and learning. Both approaches rely on a pre-trained neural network to extract semantic information but differ in codebook construction. In the memory-based approach, we use semantic quantization and compression models, leveraging past source realizations as a codebook to eliminate the need for further training. In the learning-based approach, we use a semantic vector quantized autoencoder model that learns a codebook from scratch. Both are followed by a channel coder in order to reliably convey semantic information to the receiver (classifier) through the wireless medium. In addition to classification accuracy, we define system time efficiency as a new performance metric. Our results demonstrate that the proposed memory-based approach outperforms its learning-based counterpart with respect to system time efficiency while offering comparable accuracy to semantic agnostic conventional baselines.

Classification-Oriented Semantic Wireless Communications

TL;DR

The paper addresses the challenge of transmitting semantic meaning for classification tasks over wireless channels under latency and data constraints. It compares memory-based semantic processing (semantic quantization and semantic compression) with a learning-based vector quantized autoencoder (VQ-VAE) approach, both built on pre-trained semantic extractors (SBERT/CLIP), and introduces system time efficiency to quantify time saved by avoiding training. Key findings show memory-based methods achieve higher system time efficiency and comparable accuracy to semantic-agnostic baselines, while the learning-based VQ-AE is more sensitive to training data, with results validated on text (AG's News) and image (STL-10) modalities over AWGN and Rayleigh fading channels. The work demonstrates that leveraging past realizations as codebooks enables fast, interpretable, and resource-efficient semantic communications suitable for dynamic wireless environments, with potential benefits for 6G-era task-oriented links.

Abstract

We propose semantic communication over wireless channels for various modalities, e.g., text and images, in a task-oriented communications setup where the task is classification. We present two approaches based on memory and learning. Both approaches rely on a pre-trained neural network to extract semantic information but differ in codebook construction. In the memory-based approach, we use semantic quantization and compression models, leveraging past source realizations as a codebook to eliminate the need for further training. In the learning-based approach, we use a semantic vector quantized autoencoder model that learns a codebook from scratch. Both are followed by a channel coder in order to reliably convey semantic information to the receiver (classifier) through the wireless medium. In addition to classification accuracy, we define system time efficiency as a new performance metric. Our results demonstrate that the proposed memory-based approach outperforms its learning-based counterpart with respect to system time efficiency while offering comparable accuracy to semantic agnostic conventional baselines.
Paper Structure (10 sections, 5 equations, 5 figures, 2 tables)

This paper contains 10 sections, 5 equations, 5 figures, 2 tables.

Table of Contents

  1. INTRODUCTION
  2. SYSTEM MODEL
  3. Classification Block
  4. PROPOSED MODELS
  5. Semantic Quantization
  6. Semantic Compression
  7. Semantic Vector Quantized Autoencoder
  8. Performance Metric
  9. RESULTS
  10. CONCLUSION

Figures (5)

  • Figure 1: The model of memory-based semantic quantization.
  • Figure 2: The model of memory-based semantic compression.
  • Figure 3: The model of learning-based semantic VQ AE.
  • Figure 4: System time efficiency results obtained for the AG's News dataset over AWGN and Rayleigh fading channels with a time budget of 100 seconds.
  • Figure 5: System time efficiency results obtained for the STL 10 dataset over AWGN and Rayleigh fading channels with a time budget of 100 seconds.