Ampere system boosts split federated learning efficiency and accuracy

Researchers have introduced Ampere, a novel system designed to enhance the efficiency and accuracy of split federated learning (SFL). Ampere addresses the limitations of traditional SFL, which often suffers from high communication overhead and reduced accuracy with non-independent and identically distributed (non-IID) data. The new system employs unidirectional inter-block training and a lightweight auxiliary network generation method to significantly decrease on-device computation and device-server communication. Experiments demonstrate that Ampere improves model accuracy by up to 11.70 percentage points, trains up to 18.6x faster, and incurs substantially lower communication and computation costs compared to existing SFL methods, while also showing improved performance on heterogeneous data. AI

IMPACT Improves efficiency and accuracy in federated learning, potentially enabling more complex models on distributed devices.

RANK_REASON Research paper detailing a new system for federated learning. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• Ampere
• CNNS
• federated learning
• Split Federated Learning
• transformers
• Zihan Zhang

• paper
• infra

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