GossipGraD: Scalable Deep Learning using Gossip Communication based Asynchronous Gradient Descent

Explore this paper's citation graph

Summary

GossipGraD can achieve perfect efficiency for these datasets and their associated neural network topologies such as GoogLeNet and ResNet50 and is able to achieve ~100% compute efficiency using 128 NVIDIA Pascal P100 GPUs - while matching the top-1 classification accuracy published in literature.

Type
preprint
Published
2018-03-15
Cited by
108
References
66
Access
Open access

Keywords

Gossip, Asynchronous communication, Computer science, Scalability, Asynchronous learning

References

Cited by

Related papers