Sorry, you need to enable JavaScript to visit this website.

facebooktwittermailshare

FAST DECENTRALIZED LEARNING VIA HYBRID CONSENSUS ADMM

Abstract: 

The present work introduces the hybrid consensus alternating direction method of multipliers (H-CADMM), a novel framework for optimization over networks which unifies existing distributed optimization approaches, including the centralized and the decentralized consensus ADMM. H-CADMM provides a flexible tool that leverages the underlying graph topology in order to achieve a desirable sweet-spot between node-to-node communication overhead and rate of convergence -- thereby alleviating known limitations of both C-CADMM and D-CADMM. A rigorous analysis of the novel method establishes linear convergence rate, and also guides the choice of parameters to optimize this rate. The novel hybrid update rules of H-CADMM lend themselves to "in-network acceleration" that is shown to effect considerable -- and essentially "free-of-charge" -- performance boost over the fully decentralized ADMM. Comprehensive numerical tests validate the analysis and showcase the potential of the method in tackling efficiently, widely useful learning tasks.

up
0 users have voted:

Paper Details

Authors:
Meng Ma, Athanasios N. Nikolakopoulos, Georgios B. Giannakis
Submitted On:
13 April 2018 - 4:18pm
Short Link:
Type:
Poster
Event:
Presenter's Name:
Meng Ma
Paper Code:
3414
Document Year:
2018
Cite

Document Files

H-CADMM

(340)

Subscribe

[1] Meng Ma, Athanasios N. Nikolakopoulos, Georgios B. Giannakis, "FAST DECENTRALIZED LEARNING VIA HYBRID CONSENSUS ADMM", IEEE SigPort, 2018. [Online]. Available: http://sigport.org/2760. Accessed: Sep. 19, 2020.
@article{2760-18,
url = {http://sigport.org/2760},
author = {Meng Ma; Athanasios N. Nikolakopoulos; Georgios B. Giannakis },
publisher = {IEEE SigPort},
title = {FAST DECENTRALIZED LEARNING VIA HYBRID CONSENSUS ADMM},
year = {2018} }
TY - EJOUR
T1 - FAST DECENTRALIZED LEARNING VIA HYBRID CONSENSUS ADMM
AU - Meng Ma; Athanasios N. Nikolakopoulos; Georgios B. Giannakis
PY - 2018
PB - IEEE SigPort
UR - http://sigport.org/2760
ER -
Meng Ma, Athanasios N. Nikolakopoulos, Georgios B. Giannakis. (2018). FAST DECENTRALIZED LEARNING VIA HYBRID CONSENSUS ADMM. IEEE SigPort. http://sigport.org/2760
Meng Ma, Athanasios N. Nikolakopoulos, Georgios B. Giannakis, 2018. FAST DECENTRALIZED LEARNING VIA HYBRID CONSENSUS ADMM. Available at: http://sigport.org/2760.
Meng Ma, Athanasios N. Nikolakopoulos, Georgios B. Giannakis. (2018). "FAST DECENTRALIZED LEARNING VIA HYBRID CONSENSUS ADMM." Web.
1. Meng Ma, Athanasios N. Nikolakopoulos, Georgios B. Giannakis. FAST DECENTRALIZED LEARNING VIA HYBRID CONSENSUS ADMM [Internet]. IEEE SigPort; 2018. Available from : http://sigport.org/2760