Authors: Tinghao Zhang, Kwok-Yan Lam, Jun Zhao

Published on: February 04, 2024

Impact Score: 8.22

Arxiv code: Arxiv:2402.02506

Summary

• What is new: Improves Hierarchical Federated Learning (HFL) by introducing a deep reinforcement learning-based approach for device scheduling and an improved K-Center algorithm for IoT device assignment to edge servers.
• Why this is important: Network congestion in Federated Learning (FL) for IoT due to simultaneous participation of a large number of devices in the training process.
• What the research proposes: A new device scheduling methodology combined with an improved K-Center algorithm and a deep reinforcement learning approach for effective IoT device assignment to edge servers in HFL scenarios.
• Results: Scheduling 50% of IoT devices achieves model convergence with lower time delay and energy consumption. Scheduling 30% of devices significantly reduces energy and communication overhead with little to no loss in model accuracy.

Technical Details

Technological frameworks used: Federated Learning, Hierarchical Federated Learning

Models used: Deep Reinforcement Learning-based model, K-Center algorithm

Data used: IoT device data

Potential Impact

IoT providers, telecom companies, and firms leveraging IoT within operations could greatly benefit; also, energy and network service providers may see changes in demand patterns.

Want to implement this idea in a business?

We have generated a startup concept here: EffiFluent IoT Solutions.