Client Drift
Client Drift is a phenomenon in federated learning where local model updates from clients deviate from the global optimization direction due to statistical heterogeneity across decentralized data distributions, resulting in biased or inconsistent convergence.
21
Publications
2K
Citations
86
Authors
35
Institutions
Publications per year
2017–2026
19
Parents
21
SCAFFOLD: Stochastic Controlled Averaging for On-Device Federated Learning.
Sai Praneeth Karimireddy, Satyen Kale, Mehryar Mohri et al. · 2019 · 215 citations
Mime: Mimicking Centralized Stochastic Algorithms in Federated Learning
Sai Praneeth Karimireddy, Martin Jaggi, Satyen Kale et al. · arXiv (Cornell University) · 2020 · 94 citations · Full text
1–5 of 21
86
École Polytechnique Fédérale de Lausanne, Switzerland
3 publications · 1K citations
3
h-index
2
h-index
2
h-index
2
h-index
2
h-index
1–5 of 86
35
University of Chinese Academy of Sciences
Beijing, China
2 h-index
4 publications
188 citations

City University of Hong Kong
Hong Kong, Hong Kong
2 h-index
8 publications
136 citations
École Polytechnique Fédérale de Lausanne
Lausanne, Switzerland
2 h-index
3 publications
923 citations
University of Pittsburgh
Pittsburgh, United States
2 h-index
4 publications
160 citations
Google (United States)
Mountain View, United States
2 h-index
5 publications
458 citations
Simula UiB
Bergen, Norway
1 h-index
4 publications
232 citations

Sun Yat-sen University
Guangzhou, China
1 h-index
1 publications
35 citations
Syracuse University
Syracuse, United States
1 h-index
1 publications
122 citations
1–8 of 35
Repository · 5 publications · 902 citations
5
h-index
2 publications · 157 citations
2
h-index
1 publications · 149 citations
1
h-index
Conference · 1 publications · 58 citations
1
h-index
1 publications · 24 citations
1
h-index
1–5