Accelerating Distributed GNN Training by Codes-Reference-Cited by-同舟云学术

Accelerating Distributed GNN Training by Codes

Author:

Wang Yanhong¹^ORCID,Guan Tianchan²^ORCID,Niu Dimin²^ORCID,Zou Qiaosha³^ORCID,Zheng Hongzhong²^ORCID,Shi C.-J. Richard⁴^ORCID,Xie Yuan²^ORCID

Affiliation:

1. Institute of Brain-inspired Circuits and Systems, Fudan University, Shanghai, China

2. DAMO Academy, Alibaba Group, Hangzhou, China

3. Graph Computing Center, Zhejiang Lab, Hangzhou, China

4. Department of Electronic and Computer Engineering, University of Washington, Seattle, WA, USA

Funder

Science and Technology Commission of Shanghai Municipality

National Key Research and Development Program of China

Damo Academy through Damo Academy Research Intern Program

Publisher

Institute of Electrical and Electronics Engineers (IEEE)

Subject

Computational Theory and Mathematics,Hardware and Architecture,Signal Processing

Link

Reference55 articles.

1. Deep gradient compression: Reducing the communication bandwidth for distributed training;lin;Proc Int Conf Learn Representations,2018

2. Communication-efficient distributed SGD with sketching;ivkin;Proc Int Conf Neural Inf Process,2019

3. Qsparse-local-SGD: Distributed SGD with quantization, sparsification and local computations;basu;Proc Int Conf Neural Inf Process,2019

4. Gradient sparsification for communication-efficient distributed optimization;wangni;Proc Int Conf Neural Inf Process,2018