Revisiting LocalSGD and SCAFFOLD: Improved rates and missing analysis ; PMLR

Λεπτομέρειες βιβλιογραφικής εγγραφής
Τίτλος: Revisiting LocalSGD and SCAFFOLD: Improved rates and missing analysis ; PMLR
Συγγραφείς: Luo, Ruichen, Stich, Sebastian U., Horváth, Samuel, Takáč, Martin
Πηγή: Luo R, Stich SU, Horváth S, Takáč M. Revisiting LocalSGD and SCAFFOLD: Improved rates and missing analysis. In: The 28th International Conference on Artificial Intelligence and Statistics . Vol 258. ML Research Press; 2025:2539-2547.
Στοιχεία εκδότη: ML Research Press
Έτος έκδοσης: 2025
Συλλογή: IST Austria Research Explorer (Institute of Science and Technology)
Περιγραφή: LocalSGD and SCAFFOLD are widely used methods in distributed stochastic optimization, with numerous applications in machine learning, large-scale data processing, and federated learning. However, rigorously establishing their theoretical advantages over simpler methods, such as minibatch SGD (MbSGD), has proven challenging, as existing analyses often rely on strong assumptions, unrealistic premises, or overly restrictive scenarios. In this work, we revisit the convergence properties of LocalSGD and SCAFFOLD under a variety of existing or weaker conditions, including gradient similarity, Hessian similarity, weak convexity, and Lipschitz continuity of the Hessian. Our analysis shows that (i) LocalSGD achieves faster convergence compared to MbSGD for weakly convex functions without requiring stronger gradient similarity assumptions; (ii) LocalSGD benefits significantly from higher-order similarity and smoothness; and (iii) SCAFFOLD demonstrates faster convergence than MbSGD for a broader class of non-quadratic functions. These theoretical insights provide a clearer understanding of the conditions under which LocalSGD and SCAFFOLD outperform MbSGD.
Τύπος εγγράφου: conference object
Γλώσσα: English
ISSN: 2640-3498
Relation: info:eu-repo/semantics/altIdentifier/e-issn/2640-3498; info:eu-repo/semantics/altIdentifier/arxiv/2501.04443
Διαθεσιμότητα: https://research-explorer.ista.ac.at/record/20302
Rights: info:eu-repo/semantics/openAccess
Αριθμός Καταχώρησης: edsbas.840D8141
Βάση Δεδομένων: BASE