SSFL: Discovering Sparse Unified Subnetworks at Initialization for Efficient Federated Learning
SSFL:在初始化时发现稀疏统一子网络以实现高效的联邦学习
机构 * Georgia Institute of Technology(佐治亚理工学院) ; TReNDS Center(TReNDS中心) ; Google DeepMind, Mila - Quebec AI Institute(谷歌DeepMind、魁北克AI研究所) ; Georgia State University(佐治亚州立大学) ; Emory University(埃默里大学) ; Center for Translational Research in Neuroimaging & Data Science (TReNDS)(神经影像与数据科学转化研究中心(TReNDS))
AI总结 SSFL通过在初始化时发现稀疏统一子网络,实现高效联邦学习,减少通信成本并提升准确性与稀疏性权衡。
Comments Published in Transactions on Machine Learning Research (TMLR), 2026
Journal ref Transactions on Machine Learning Research, 2026