.png){kind=icon}
.png){kind=icon}
Return
FedA2L: Adaptive layer-wise learning rate adjustment in decentralized federated learning
.png){kind=icon}
PRE
AI
V
K
T
DOI：10.1016/j.future.2026.108743.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• Introduces a three-state DFL model for layer-wise conflict analysis. • Proposes adaptive layer-wise LRs using divergence and stability metrics. • Seamlessly integrates into existing DFL protocols without architectural changes. • Achieves 4.94 × faster convergence and 80% lower communication overhead. • Robust to severe non-IID data, large-scale decentralization, and sparse networks.
Journal
F
IF:
6.1
Papers:
6.8K
Citations:
2.3W
Organization
No organization information available
Cited Papers