Join this FLOW Seminar presentation where Peter Richtarik from KAUST introduces the first optimal parallel SGD algorithm. This 1 hour and 43 minute talk, delivered on April 9th, 2025 as part of the Federated Learning One World Seminar series, explores groundbreaking developments in stochastic gradient descent optimization for parallel computing environments. Discover the theoretical foundations and practical applications of this optimal approach to parallel SGD, presented by a leading researcher in the field. Access additional resources through the seminar website, view archived talks, or learn more about the speaker through his personal website and the FLOW Seminar Twitter account.
Overview
Syllabus
FLOW Seminar #125: Peter Richtarik (KAUST) The First Optimal Parallel SGD
Taught by
Federated Learning One World Seminar
Related Courses
-
The First Asynchronous SGD with Optimal Time Complexity - Seminar #126
-
The First Optimal Distributed SGD in the Presence of Data, Compute and Communication Heterogeneity
-
Federated Learning Seminar - #127
-
SGD in the Large - Average-Case Analysis, Asymptotics, and Stepsize Criticality
-
Hessian-Aware Stochastic Differential Equation Modeling of SGD
-
Towards Understanding the Implicit Regularization Effect of SGD
