Asynchrony and Acceleration in Gossip Algorithms - Inria - Institut national de recherche en sciences et technologies du numérique Access content directly
Preprints, Working Papers, ... Year :

Asynchrony and Acceleration in Gossip Algorithms

Hadrien Hendrikx
  • Function : Author
  • PersonId : 1053506
Mathieu Even


This paper considers the minimization of a sum of smooth and strongly convex functions dispatched over the nodes of a communication network. Previous works on the subject either focus on synchronous algorithms, which can be heavily slowed down by a few slow nodes (the straggler problem), or consider a historical asynchronous setting (Boyd et al., 2006), which relies on a communication model that cannot be readily implemented in practice, as it does not capture important aspects of asynchronous communications such as non-instantaneous computations and communications. We have two main contributions. 1) We introduce a new communication scheme, based on Loss-Networks, that is programmable in a fully asynchronous and decentralized fashion. We establish empirically and theoretically that it improves over existing synchronous algorithms by depending on local communication delays in the analysis instead of global worst-ones. 2) We provide an acceleration of the standard gossip algorithm in the historical asynchronous model without requiring any additional synchronization.
Fichier principal
Vignette du fichier
Gossip_arxiv_v2.pdf (990.82 Ko) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

hal-02989459 , version 1 (05-11-2020)
hal-02989459 , version 2 (10-02-2021)


  • HAL Id : hal-02989459 , version 2


Hadrien Hendrikx, Laurent Massoulié, Mathieu Even. Asynchrony and Acceleration in Gossip Algorithms. 2021. ⟨hal-02989459v2⟩
79 View
83 Download


Gmail Facebook Twitter LinkedIn More