Multi-stage Gradient Compression: Overcoming the Communication Bottleneck in Distributed Deep Learning.

Qu Lu, Wantao Liu, Jizhong Han, Jinrong Guo

Browse the full ICONIP paper archive.