Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Personalized Federated Learning for Generative AI-Assisted Semantic Communications

Yubo Peng, Feibo Jiang, Li Dong, Kezhi Wang, Kun Yang

TL;DR

This work proposes a GAI-assisted SC (GSC) model deployed between MUs and the Base Station (BS), and introduces Personalized Semantic Federated Learning (PSFL), which incorporates a novel Personalized Local Distillation (PLD) and Adaptive Global Pruning (AGP).

Abstract

Semantic Communication (SC) focuses on transmitting only the semantic information rather than the raw data. This approach offers an efficient solution to the issue of spectrum resource utilization caused by the various intelligent applications on Mobile Users (MUs). Generative Artificial Intelligence (GAI) models have recently exhibited remarkable content generation and signal processing capabilities, presenting new opportunities for enhancing SC. Therefore, we propose a GAI-assisted SC (GSC) model deployed between MUs and the Base Station (BS). Then, to train the GSC model using the local data of MUs while ensuring privacy and accommodating heterogeneous requirements of MUs, we introduce Personalized Semantic Federated Learning (PSFL). This approach incorporates a novel Personalized Local Distillation (PLD) and Adaptive Global Pruning (AGP). In PLD, each MU selects a personalized GSC model as a mentor tailored to its local resources and a unified Convolutional Neural Networks (CNN)-based SC (CSC) model as a student. This mentor model is then distilled into the student model for global aggregation. In AGP, we perform network pruning on the aggregated global model according to real-time communication environments, reducing communication energy. Finally, numerical results demonstrate the feasibility and efficiency of the proposed PSFL scheme.

Personalized Federated Learning for Generative AI-Assisted Semantic Communications

TL;DR

This work proposes a GAI-assisted SC (GSC) model deployed between MUs and the Base Station (BS), and introduces Personalized Semantic Federated Learning (PSFL), which incorporates a novel Personalized Local Distillation (PLD) and Adaptive Global Pruning (AGP).

Abstract

Semantic Communication (SC) focuses on transmitting only the semantic information rather than the raw data. This approach offers an efficient solution to the issue of spectrum resource utilization caused by the various intelligent applications on Mobile Users (MUs). Generative Artificial Intelligence (GAI) models have recently exhibited remarkable content generation and signal processing capabilities, presenting new opportunities for enhancing SC. Therefore, we propose a GAI-assisted SC (GSC) model deployed between MUs and the Base Station (BS). Then, to train the GSC model using the local data of MUs while ensuring privacy and accommodating heterogeneous requirements of MUs, we introduce Personalized Semantic Federated Learning (PSFL). This approach incorporates a novel Personalized Local Distillation (PLD) and Adaptive Global Pruning (AGP). In PLD, each MU selects a personalized GSC model as a mentor tailored to its local resources and a unified Convolutional Neural Networks (CNN)-based SC (CSC) model as a student. This mentor model is then distilled into the student model for global aggregation. In AGP, we perform network pruning on the aggregated global model according to real-time communication environments, reducing communication energy. Finally, numerical results demonstrate the feasibility and efficiency of the proposed PSFL scheme.
Paper Structure (25 sections, 23 equations, 13 figures, 1 table, 2 algorithms)

This paper contains 25 sections, 23 equations, 13 figures, 1 table, 2 algorithms.

Table of Contents

  1. Introduction
  2. System Model
  3. GSC Model
  4. Transmitter
  5. Physical channel
  6. Receiver
  7. FL Model
  8. Communication Model
  9. Personalized Semantic Federated Learning for GSC Model
  10. PLD for Local Training
  11. Distill knowledge from hard labels
  12. Distill knowledge from soft labels
  13. Distill knowledge from the semantic information
  14. Update mentor and student models
  15. AGP for Global Aggregation
  16. ...and 10 more sections

Figures (13)

  • Figure 1: The illustration that MUs communicate with BS using SC.
  • Figure 2: The illustration of image transmission utilizing the proposed GSC model.
  • Figure 3: The illustration of the proposed PSFL.
  • Figure 4: Loss versus iteration under student and mentor models on datasets (a) MNIST, (b) Fashion-MNIST, (c) CIFAR-10, and (d) CIFAR-100.
  • Figure 5: Accuracy versus iteration under student and mentor models on datasets (a) MNIST, (b) Fashion-MNIST, (c) CIFAR-10, and (d) CIFAR-100.
  • ...and 8 more figures