4.7 Article

On the Decentralized Stochastic Gradient Descent With Markov Chain Sampling

Journal

IEEE TRANSACTIONS ON SIGNAL PROCESSING
Volume 71, Issue -, Pages 2895-2909

Publisher

IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
DOI: 10.1109/TSP.2023.3297053

Keywords

Markov chain sampling; gradient descent; decentralization; distributed machine learning; convergence

Ask authors/readers for more resources

This paper studies the decentralized Markov chain gradient descent (DMGD), a variant of the decentralized stochastic gradient method. It analyzes the convergence rates of DMGD on a connected graph, highlighting the critical dependencies on the graph topology and the mixing time of the Markov chain. The numerical experiments also verify the sample efficiency of DMGD.
The decentralized stochastic gradient method emerges as a promising solution for solving large-scale machine learning problems. This paper studies the decentralized Markov chain gradient descent (DMGD), a variant of the decentralized stochastic gradient method, which draws random samples along the trajectory of a Markov chain. DMGD arises when obtaining independent samples is costly or impossible, excluding the use of the traditional stochastic gradient algorithms. Specifically, we consider the DMGD over a connected graph, where each node only communicates with its neighbors by sending and receiving the intermediate results. We establish both ergodic and nonergodic convergence rates of DMGD, which elucidate the critical dependencies on the topology of the graph that connects all nodes and the mixing time of the Markov chain. We further numerically verify the sample efficiency of DMGD.

Authors

I am an author on this paper
Click your name to claim this paper and add it to your profile.

Reviews

Primary Rating

4.7
Not enough ratings

Secondary Ratings

Novelty
-
Significance
-
Scientific rigor
-
Rate this paper

Recommended

No Data Available
No Data Available