Distributed Optimization for Deep Learning with Gossip Exchange

Michael Blot (1) , David Picard (2, 1) , Nicolas Thome (1) , Matthieu Cord (1)

1 MLIA - Machine Learning and Information Access
2 ETIS - UMR 8051 - Equipes Traitement de l'Information et Systèmes

We address the issue of speeding up the training of convolutional neural networks by studying a distributed method adapted to stochastic gradient descent. Our parallel optimization setup uses several threads, each applying individual gradient descents on a local variable. We propose a new way of sharing information between different threads based on gossip algorithms that show good consensus convergence properties. Our method called GoSGD has the advantage to be fully asynchronous and decentralized.

Mots clés: Optimization; Distributed gradient descent; Neural networks; Learning; Deep; Gossip

Domaines: Traitement du signal et de l'image [eess.SP]; Réseau de neurones [cs.NE]

Michael Blot, David Picard, Nicolas Thome, Matthieu Cord. Distributed Optimization for Deep Learning with Gossip Exchange. Neurocomputing, 2019, 330, pp.287-296. ⟨10.1016/j.neucom.2018.11.002⟩. ⟨hal-01930346⟩