FedAvg Algorithm Explores Personalization vs. Generalization in HAR

This paper explores the effectiveness of the FedAvg algorithm in the Human Activity Recognition (HAR) domain, focusing on the balance between personalized and generalized accuracy in federated learning. Researchers designed and implemented various testing scenarios to compare centralized, local, and federated learning paradigms, including a simulation of changing client data. The experimental results indicate that FedAvg offers better personalization while maintaining strong generalization compared to traditional centralized learning, though this advantage diminishes under challenging conditions like varying class distributions across clients. AI

IMPACT This research explores trade-offs in federated learning for activity recognition, potentially informing more privacy-preserving and personalized AI systems.

RANK_REASON The cluster contains a research paper submitted to arXiv. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• Alessandro Bogliolo
• alphaXiv
• arXiv
• CatalyzeX Code Finder for Papers
• CORE Recommender
• DagsHub
• FedAvg
• Gotit.pub
• Hugging Face
• Human Activity Recognition
• IArxiv Recommender
• Influence Flower
• ScienceCast

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →