Research on Solving Communication Instability and Non-IID

Lan Wang

2024

Abstract

Conventional machine learning (ML) methods for load forecasting rely on a central server for ML training. However, this approach has drawbacks as it necessitates transmitting all data collected by diverse devices to the central server. This process poses risks to privacy and security, strains the communication network, and demands significant centralized computing resources. In contrast, federated learning (FL) allows multiple parties to collaboratively train ML models without sharing their local data. An inherent challenge in FL is addressing the diversity in the distribution of local data across participating parties. Despite numerous studies aimed at overcoming this challenge, existing approaches often fall short in achieving satisfactory performance, particularly when dealing with image datasets and deep learning models. Model-contrastive Federated Learning (MOON) presents a straightforward and effective FL framework. MOON's core concept involves leveraging the similarity between model representations to refine individual local training, essentially conducting comparative learning at the model level. Extensive experiments demonstrate that MOON outperforms the most advanced FL algorithms across various image classification tasks.

Paper Citation

in Harvard Style

Wang L. (2024). Research on Solving Communication Instability and Non-IID. In Proceedings of the 1st International Conference on Data Science and Engineering - Volume 1: ICDSE; ISBN 978-989-758-690-3, SciTePress, pages 168-171. DOI: 10.5220/0012835800004547

in Bibtex Style

@conference{icdse24,
author={Lan Wang},
title={Research on Solving Communication Instability and Non-IID},
booktitle={Proceedings of the 1st International Conference on Data Science and Engineering - Volume 1: ICDSE},
year={2024},
pages={168-171},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0012835800004547},
isbn={978-989-758-690-3},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 1st International Conference on Data Science and Engineering - Volume 1: ICDSE
TI - Research on Solving Communication Instability and Non-IID
SN - 978-989-758-690-3
AU - Wang L.
PY - 2024
SP - 168
EP - 171
DO - 10.5220/0012835800004547
PB - SciTePress