Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

GAI-Enabled Explainable Personalized Federated Semi-Supervised Learning

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

TL;DR

This work tackles three persistent challenges in federated learning over mobile users: label scarcity, data non-IID-ness, and lack of explainability. It proposes XPFL, a framework that combines GAI-enabled semi-supervised local training (GFed) with explainability-driven global aggregation (XFed). GFed uses a Vision Transformer-based Generative Autoencoder (GAE) to learn from unlabeled data and distills its knowledge into the local FL models, while a cosine-distance-based personalization in aggregation preserves local characteristics. XFed provides explainability by fitting a Decision Tree (DT) to the local model and visualizing the global update with t-SNE, augmented by a QoX metric to quantify explainability, with simulations showing improved performance and transparency under label scarcity and non-IID settings.

Abstract

Federated learning (FL) is a commonly distributed algorithm for mobile users (MUs) training artificial intelligence (AI) models, however, several challenges arise when applying FL to real-world scenarios, such as label scarcity, non-IID data, and unexplainability. As a result, we propose an explainable personalized FL framework, called XPFL. First, we introduce a generative AI (GAI) assisted personalized federated semi-supervised learning, called GFed. Particularly, in local training, we utilize a GAI model to learn from large unlabeled data and apply knowledge distillation-based semi-supervised learning to train the local FL model using the knowledge acquired from the GAI model. In global aggregation, we obtain the new local FL model by fusing the local and global FL models in specific proportions, allowing each local model to incorporate knowledge from others while preserving its personalized characteristics. Second, we propose an explainable AI mechanism for FL, named XFed. Specifically, in local training, we apply a decision tree to match the input and output of the local FL model. In global aggregation, we utilize t-distributed stochastic neighbor embedding (t-SNE) to visualize the local models before and after aggregation. Finally, simulation results validate the effectiveness of the proposed XPFL framework.

GAI-Enabled Explainable Personalized Federated Semi-Supervised Learning

TL;DR

This work tackles three persistent challenges in federated learning over mobile users: label scarcity, data non-IID-ness, and lack of explainability. It proposes XPFL, a framework that combines GAI-enabled semi-supervised local training (GFed) with explainability-driven global aggregation (XFed). GFed uses a Vision Transformer-based Generative Autoencoder (GAE) to learn from unlabeled data and distills its knowledge into the local FL models, while a cosine-distance-based personalization in aggregation preserves local characteristics. XFed provides explainability by fitting a Decision Tree (DT) to the local model and visualizing the global update with t-SNE, augmented by a QoX metric to quantify explainability, with simulations showing improved performance and transparency under label scarcity and non-IID settings.

Abstract

Federated learning (FL) is a commonly distributed algorithm for mobile users (MUs) training artificial intelligence (AI) models, however, several challenges arise when applying FL to real-world scenarios, such as label scarcity, non-IID data, and unexplainability. As a result, we propose an explainable personalized FL framework, called XPFL. First, we introduce a generative AI (GAI) assisted personalized federated semi-supervised learning, called GFed. Particularly, in local training, we utilize a GAI model to learn from large unlabeled data and apply knowledge distillation-based semi-supervised learning to train the local FL model using the knowledge acquired from the GAI model. In global aggregation, we obtain the new local FL model by fusing the local and global FL models in specific proportions, allowing each local model to incorporate knowledge from others while preserving its personalized characteristics. Second, we propose an explainable AI mechanism for FL, named XFed. Specifically, in local training, we apply a decision tree to match the input and output of the local FL model. In global aggregation, we utilize t-distributed stochastic neighbor embedding (t-SNE) to visualize the local models before and after aggregation. Finally, simulation results validate the effectiveness of the proposed XPFL framework.

Paper Structure

This paper contains 27 sections, 31 equations, 10 figures, 5 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. FL for Label Scarcity
  4. FL for Non-IID data
  5. Explainability in FL
  6. System Model
  7. FL Model
  8. Methods
  9. Overview of the XPFL Framework
  10. GAI-assisted semi-supervised learning
  11. Explainable local model
  12. Personalized global aggregation
  13. Visual global aggregation
  14. The Proposed GFed Algorithm
  15. GAI-assisted semi-supervised learning
  16. ...and 12 more sections

Figures (10)

  • Figure 1: The illustration of the considered system model.
  • Figure 2: The illustration of the proposed XPFL framework.
  • Figure 3: The illustration of GAE. (a) The unsupervised learning is based on GAE. (b) The core architecture of ViT.
  • Figure 4: Image transmission results on the MNIST dataset. (a) Visual comparison of image transmission quality. (b) Quantitative comparison of image transmission quality.
  • Figure 5: Image transmission results on the Fashion-MNIST dataset. (a) Visual comparison of image transmission quality. (b) Quantitative comparison of image transmission quality.
  • ...and 5 more figures