activity
20242026
most citedTAMUNA: Doubly Accelerated Distributed Optimization under Partial Participation

3 citations · 3 across the 3 of their papers we have counts for

collaborators

27 papers

cs.LG2026

SpecGradFilter: A Spectral Gradient Filtering Framework for Taming Federated Heterogeneity

Liyang Yuan, Yibo Yang, Dandan Guo +2

Federated Learning (FL) is fundamentally challenged by statistical heterogeneity, where non-identically distributed (non-IID) data induces client drift that severely hampers global…

cs.LG20263 cited

TAMUNA: Doubly Accelerated Distributed Optimization under Partial Participation

Laurent Condat, Ivan Agarský, Grigory Malinovsky +1

In distributed optimization and federated learning, slow and costly communication between parallel devices and the central server constitutes the primary bottleneck. To alleviate t…

cs.LG2026

Demystifying Pipeline Parallelism: First Theory for PipeDream

Ivan Ilin, Peter Richtárik

Training modern machine learning models increasingly requires computation to be distributed across many accelerators. Data parallelism remains the default choice and is often paire…

math.OC2026

Tighter Performance Theory of FedExProx

Wojciech Anyszka, Kaja Gruntkowska, Alexander Tyurin +1

We revisit FedExProx - a recently proposed distributed optimization method designed to enhance convergence properties of parallel proximal algorithms via extrapolation. In the proc…

cs.LG2026

CompressedScaffnew: The First Theoretical Double Acceleration of Communication from Local Training and Compression in Distributed Optimization

Laurent Condat, Ivan Agarský, Peter Richtárik

In distributed optimization, a large number of machines alternate between local computations and communication with a coordinating server. Communication, which can be slow and cost…

cs.LG2026

MAST: Model-Agnostic Sparsified Training

Yury Demidovich, Grigory Malinovsky, Egor Shulgin +1

We introduce a novel optimization problem formulation that departs from the conventional way of minimizing machine learning model loss as a black-box function. Unlike traditional f…