FedRAW method boosts rare-label influence in asynchronous federated learning

Researchers have introduced FedRAW, a novel server-side aggregation method designed to address silent rarity failure in asynchronous federated learning. This failure mode occurs when clients with rare labels contribute too little to the global model, even if overall accuracy appears stable. FedRAW tackles this by implementing client-level update deduplication to prevent frequent clients from dominating the update buffer and by incorporating rare-label-aware weighting to boost the influence of clients carrying less common labels. Experiments on datasets like EMNIST Balanced and CIFAR-10 demonstrate that FedRAW enhances rare-label accuracy while maintaining comparable global accuracy and adding minimal computational overhead. AI

IMPACT Enhances rare-label accuracy in federated learning without compromising overall performance or adding significant computational cost.

RANK_REASON The cluster contains a research paper detailing a new method for federated learning. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• alphaXiv
• arXiv
• CatalyzeX
• CIFAR-10
• DagsHub
• EMNIST Balanced
• FedRAW
• Gotit.pub
• HAM10000
• Hugging Face
• IArxiv
• ISIC-2019
• ScienceCast

• paper
• infra

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