# Uplink Scheduling in Federated Learning: an Importance-Aware Approach via Graph Representation Learning

**Authors:** Marco Skocaj, Pedro Enrique Iturria Rivera, Roberto Verdone, Melike Erol-Kantarci  
**arXiv:** [2301.11903](https://arxiv.org/abs/2301.11903)  
**Date:** 2023-01-27  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/2301.11903v1/

## Key points

- Federated Learning (FL) is a promising framework for distributed training of AI-based services, applications, and network procedures in 6G.
- One major challenge affecting the performance and efficiency of 6G wireless FL systems is the massive scheduling of user devices over resource-constrained channels.
- Marco Skocaj et al. propose a novel approach to client scheduling in FL applications by leveraging Unsupervised Graph Representation Learning (UGRL).
- Their proposed solution introduces a relational inductive bias in the scheduling process and does not require the collection of training feedback information from client devices, unlike state-of-the-art importance-aware mechanisms.
- They present an energy-efficient and importance-aware metric for client scheduling that considers scenarios where nodes exhibit spatial relations.
- The authors evaluate their proposed solution against baseline scheduling algorithms based on recently proposed metrics in the literature.
- The results show that their approach can achieve an average gain of up to 10% in model accuracy and up to 17 times in energy efficiency compared to state-of-the-art importance-aware policies.
- This demonstrates the effectiveness of their approach towards addressing the challenges posed by massive scheduling of user devices over resource-constrained channels.
- Skocaj et al.'s work highlights how UGRL can be leveraged to address complex problems such as uplink scheduling in FL applications while improving energy efficiency and model accuracy without requiring feedback information from clients.
- Their findings could have significant implications for future research into improving wireless FL systems' performance and efficiency under resource constraints.

## Summary

The emergence of Federated Learning (FL) has provided a promising framework for distributed training of AI-based services, applications, and network procedures in 6G. However, one of the major challenges affecting the performance and efficiency of 6G wireless FL systems is the massive scheduling of user devices over resource-constrained channels. To tackle this challenge, Marco Skocaj et al. propose a novel approach to client scheduling in FL applications by leveraging Unsupervised Graph Representation Learning (UGRL). Their proposed solution introduces a relational inductive bias in the scheduling process and does not require the collection of training feedback information from client devices, unlike state-of-the-art importance-aware mechanisms. Specifically, they present an energy-efficient and importance-aware metric for client scheduling that considers scenarios where nodes exhibit spatial relations. The authors evaluate their proposed solution against baseline scheduling algorithms based on recently proposed metrics in the literature. The results show that their approach can achieve an average gain of up to 10% in model accuracy and up to 17 times in energy efficiency compared to state-of-the-art importance-aware policies. This demonstrates the effectiveness of their approach towards addressing the challenges posed by massive scheduling of user devices over resource-constrained channels. Overall, Skocaj et al. 's work highlights how UGRL can be leveraged to address complex problems such as uplink scheduling in FL applications while improving energy efficiency and model accuracy without requiring feedback information from clients. Their findings could have significant implications for future research into improving wireless FL systems' performance and efficiency under resource constraints.