New GAPSL framework enhances federated learning in heterogeneous edge systems

Researchers have developed GAPSL, a new framework for parallel split learning designed to improve federated learning in edge computing systems with heterogeneous data. GAPSL addresses training divergence caused by inconsistent gradient directions across devices by introducing two components: leader gradient identification (LGI) and gradient direction alignment (GDA). LGI selects a robust leader gradient, while GDA uses regularization to align client gradients with this leader, enhancing model convergence and robustness. Experiments show GAPSL outperforms existing benchmarks in accuracy, convergence speed, and system resilience. AI

IMPACT This framework could improve the efficiency and accuracy of AI model training on edge devices with diverse data.

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

Read on arXiv cs.LG →

• GAPSL
• Lin Zheng

• paper
• infra

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