Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Navigating High-Degree Heterogeneity: Federated Learning in Aerial and Space Networks

Fan Dong, Henry Leung, Steve Drew

TL;DR

The results reveal that the heterogeneity challenge is more pronounced in ASNs-based federated learning and that prevailing algorithms often fail to effectively address high levels of heterogeneity.

Abstract

Federated learning offers a compelling solution to the challenges of networking and data privacy within aerial and space networks by utilizing vast private edge data and computing capabilities accessible through drones, balloons, and satellites. While current research has focused on optimizing the learning process, computing efficiency, and minimizing communication overhead, the heterogeneity issue and class imbalance remain a significant barrier to rapid model convergence. In this paper, we explore the influence of heterogeneity on class imbalance, which diminishes performance in Aerial and Space Networks (ASNs)-based federated learning. We illustrate the correlation between heterogeneity and class imbalance within grouped data and show how constraints such as battery life exacerbate the class imbalance challenge. Our findings indicate that ASNs-based FL faces heightened class imbalance issues even with similar levels of heterogeneity compared to other scenarios. Finally, we analyze the impact of varying degrees of heterogeneity on FL training and evaluate the efficacy of current state-of-the-art algorithms under these conditions. Our results reveal that the heterogeneity challenge is more pronounced in ASNs-based federated learning and that prevailing algorithms often fail to effectively address high levels of heterogeneity.

Navigating High-Degree Heterogeneity: Federated Learning in Aerial and Space Networks

TL;DR

The results reveal that the heterogeneity challenge is more pronounced in ASNs-based federated learning and that prevailing algorithms often fail to effectively address high levels of heterogeneity.

Abstract

Federated learning offers a compelling solution to the challenges of networking and data privacy within aerial and space networks by utilizing vast private edge data and computing capabilities accessible through drones, balloons, and satellites. While current research has focused on optimizing the learning process, computing efficiency, and minimizing communication overhead, the heterogeneity issue and class imbalance remain a significant barrier to rapid model convergence. In this paper, we explore the influence of heterogeneity on class imbalance, which diminishes performance in Aerial and Space Networks (ASNs)-based federated learning. We illustrate the correlation between heterogeneity and class imbalance within grouped data and show how constraints such as battery life exacerbate the class imbalance challenge. Our findings indicate that ASNs-based FL faces heightened class imbalance issues even with similar levels of heterogeneity compared to other scenarios. Finally, we analyze the impact of varying degrees of heterogeneity on FL training and evaluate the efficacy of current state-of-the-art algorithms under these conditions. Our results reveal that the heterogeneity challenge is more pronounced in ASNs-based federated learning and that prevailing algorithms often fail to effectively address high levels of heterogeneity.

Paper Structure

This paper contains 14 sections, 9 figures, 1 table.

Table of Contents

  1. Introduction
  2. Statistical Heterogeneity
  3. System Heterogeneity
  4. Communication Heterogeneity
  5. Label Distribution Skew
  6. Related Work
  7. Heterogeneity and Imbalance
  8. How Battery Life Constraint Aggravates the Imbalance Issue
  9. Selecting fewer devices
  10. Selecting devices from a smaller pool
  11. How Degree of Heterogeneity Affects FL Training
  12. FedAvg Performance Under Different Heterogeneity Degrees
  13. State-of-the-art FL Algorithms Performance Under Different Heterogeneity Degrees
  14. Conclusion

Figures (9)

  • Figure 1: Different devices would possess various data concerning their active locations.
  • Figure 2: Only select devices with enough battery percentage.
  • Figure 3: Grouped dataset (10 devices selected out of 100 devices) imbalance degree ($\alpha$) with various heterogeneity degrees under Dirichlet distribution. The smaller $\alpha$ is, the more heterogeneous the distribution among devices will be.
  • Figure 4: Grouped dataset imbalance degree with various number of clients select.
  • Figure 5: Imbalance degree under different pool sizes.
  • ...and 4 more figures