Performance Comparison and Analysis Between MOON and FedProx in Image Classification

Xuanzhou Shi

2024

Abstract

Traditional machine learning algorithms mostly follow a centralized training paradigm, which poses challenges to data security and privacy, and restricts the in-depth application of artificial intelligence models in many fields. As a great way to train models in siloed data, federated learning has emerged in recent years and attracted much attention from the industry and academic community. Though much effort has been devoted, challenges persist within federated learning frameworks when addressing data that is identically distributed yet lacks independence. This paper focuses on introducing the latest research to solve this problem and quantitively discusses their performance on various datasets, aiming to provide the decision-making basis for algorithm selection. Specifically, three representative federated learning algorithms are first introduced, including their design ideas and key steps. This paper further analyzes the advantages and disadvantages of these methods by comparing the performance in accuracy, communication efficiency, different numbers of local epochs, and different heterogeneous environments. Extensive results show that MOON is superior to FedProx in all aspects of the experiment, showing the superiority of MOON in image classification.

Paper Citation

in Harvard Style

Shi X. (2024). Performance Comparison and Analysis Between MOON and FedProx in Image Classification. In Proceedings of the 1st International Conference on Engineering Management, Information Technology and Intelligence - Volume 1: EMITI; ISBN 978-989-758-713-9, SciTePress, pages 639-645. DOI: 10.5220/0012961100004508

in Bibtex Style

@conference{emiti24,
author={Xuanzhou Shi},
title={Performance Comparison and Analysis Between MOON and FedProx in Image Classification},
booktitle={Proceedings of the 1st International Conference on Engineering Management, Information Technology and Intelligence - Volume 1: EMITI},
year={2024},
pages={639-645},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0012961100004508},
isbn={978-989-758-713-9},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 1st International Conference on Engineering Management, Information Technology and Intelligence - Volume 1: EMITI
TI - Performance Comparison and Analysis Between MOON and FedProx in Image Classification
SN - 978-989-758-713-9
AU - Shi X.
PY - 2024
SP - 639
EP - 645
DO - 10.5220/0012961100004508
PB - SciTePress