Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

QuantFL: Sustainable Federated Learning for Edge IoT via Pre-Trained Model Quantisation

Charuka Herath, Yogachandran Rahulamathavan, Varuna De Silva, Sangarapillai Lambotharan

Abstract

Federated Learning (FL) enables privacy-preserving intelligence on Internet of Things (IoT) devices but incurs a significant carbon footprint due to the high energy cost of frequent uplink transmission. While pre-trained models are increasingly available on edge devices, their potential to reduce the energy overhead of fine-tuning remains underexplored. In this work, we propose QuantFL, a sustainable FL framework that leverages pre-trained initialisation to enable aggressive, computationally lightweight quantisation. We demonstrate that pre-training naturally concentrates update statistics, allowing us to use memory-efficient bucket quantisation without the energy-intensive overhead of complex error-feedback mechanisms. On MNIST and CIFAR-100, QuantFL reduces total communication by 40\% ($\simeq40\%$ total-bit reduction with full-precision downlink; $\geq80\%$ on uplink or when downlink is quantised) while matching or exceeding uncompressed baselines under strict bandwidth budgets; BU attains 89.00\% (MNIST) and 66.89\% (CIFAR-100) test accuracy with orders of magnitude fewer bits. We also account for uplink and downlink costs and provide ablations on quantisation levels and initialisation. QuantFL delivers a practical, "green" recipe for scalable training on battery-constrained IoT networks.

QuantFL: Sustainable Federated Learning for Edge IoT via Pre-Trained Model Quantisation

Abstract

Federated Learning (FL) enables privacy-preserving intelligence on Internet of Things (IoT) devices but incurs a significant carbon footprint due to the high energy cost of frequent uplink transmission. While pre-trained models are increasingly available on edge devices, their potential to reduce the energy overhead of fine-tuning remains underexplored. In this work, we propose QuantFL, a sustainable FL framework that leverages pre-trained initialisation to enable aggressive, computationally lightweight quantisation. We demonstrate that pre-training naturally concentrates update statistics, allowing us to use memory-efficient bucket quantisation without the energy-intensive overhead of complex error-feedback mechanisms. On MNIST and CIFAR-100, QuantFL reduces total communication by 40\% ( total-bit reduction with full-precision downlink; on uplink or when downlink is quantised) while matching or exceeding uncompressed baselines under strict bandwidth budgets; BU attains 89.00\% (MNIST) and 66.89\% (CIFAR-100) test accuracy with orders of magnitude fewer bits. We also account for uplink and downlink costs and provide ablations on quantisation levels and initialisation. QuantFL delivers a practical, "green" recipe for scalable training on battery-constrained IoT networks.
Paper Structure (17 sections, 10 equations, 4 figures, 3 tables, 1 algorithm)

This paper contains 17 sections, 10 equations, 4 figures, 3 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Motivation
  3. Contributions
  4. Related Work
  5. System Model and Problem Formulation
  6. Communication Cost Model
  7. Problem Formulation
  8. Methodology: QuantFL Framework
  9. Bucketed Scalar Quantisation (Per-Layer)
  10. Quantiser Instantiations
  11. Codebook Refresh and Bit Budget
  12. Implementation Details
  13. Convergence and Efficiency (Qualitative)
  14. Results and Discussion
  15. Experimental setup
  16. ...and 2 more sections

Figures (4)

  • Figure 1: QuantFL pipeline. Each client quantises its per-layer update using BU (equal-width) or BQ (equal-mass) buckets and sends only indices; boundaries are refreshed infrequently. The server applies mid-point decoding and aggregates. Pre-training makes updates narrowly distributed, enabling aggressive compression with little error.
  • Figure 2: Update concentration with pre-training vs. training from scratch (simulated illustration). Pre-trained updates exhibit substantially smaller dynamic range R and variance, and reduced kurtosis (shorter tails), enabling lower quantisation error at fixed L.
  • Figure 3: Training loss and test accuracy curves for MNIST and CIFAR-100 under different quantisation methods. Baseline FedAvg is a non-quantised and non-pre-trained setting.
  • Figure 4: Test accuracy vs. total bits per round (per client; uplink+downlink; log x-axis) on MNIST and CIFAR-100. Downlink is full-precision in all points unless stated.