AUG 10, 2026 · PREPRINT
FedA2L: Adaptive layer-wise learning rate adjustment in decentralized federated learning
arXiv
This is an unrefereed technical preprint proposing a novel algorithmic method for federated learning, demonstrating computational improvements in simulation but lacking peer review, clinical relevance, or human validation.
Reported
Convergence speedup vs. vanilla D…up to 4.94 times
Communication round reduction vs.…up to 59%