ERIS framework enhances federated learning privacy and scalability

Researchers have introduced ERIS, a new framework for federated learning designed to improve the balance between privacy, utility, and scalability. ERIS utilizes Federated Shard Aggregation (FSA) to partition client updates, distributing aggregation across multiple client-side aggregators. This approach eliminates the central aggregation bottleneck and limits information exposure to any single observer. Experiments show ERIS achieves utility comparable to FedAvg while significantly reducing communication overhead and enhancing robustness against privacy attacks, without requiring complex cryptography. AI

IMPACT This framework could enable more scalable and private training of large models, particularly in scenarios with sensitive data.

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

Read on arXiv cs.LG →

• arXiv
• Connected Papers
• Dario Fenoglio
• Distributed Shifted Compression
• ERIS
• FedAvg
• federated learning
• Federated Shard Aggregation
• Hugging Face
• IArxiv

• paper
• infra

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