Poster
Can Decentralized Algorithms Outperform Centralized Algorithms? A Case Study for Decentralized Parallel Stochastic Gradient Descent
Xiangru Lian · Ce Zhang · Huan Zhang · Cho-Jui Hsieh · Wei Zhang · Ji Liu

Wed Dec 6th 06:30 -- 10:30 PM @ Pacific Ballroom #167 #None

Most distributed machine learning systems nowadays, including TensorFlow and CNTK, are built in a centralized fashion. One bottleneck of centralized algorithms lies on high communication cost on the central node. Motivated by this, we ask, can decentralized algorithms be faster than its centralized counterpart? Although decentralized PSGD (D-PSGD) algorithms have been studied by the control community, existing analysis and theory do not show any advantage over centralized PSGD (C-PSGD) algorithms, simply assuming the application scenario where only the decentralized network is available. In this paper, we study a D-PSGD algorithm and provide the first theoretical analysis that indicates a regime in which decentralized algorithms might outperform centralized algorithms for distributed stochastic gradient descent. This is because D-PSGD has comparable total computational complexities to C-PSGD but requires much less communication cost on the busiest node. We further conduct an empirical study to validate our theoretical analysis across multiple frameworks (CNTK and Torch), different network configurations, and computation platforms up to 112 GPUs. On network configurations with low bandwidth or high latency, D-PSGD can be up to one order of magnitude faster than its well-optimized centralized counterparts.

Author Information

Xiangru Lian (University of Rochester)
Ce Zhang (ETH Zurich)
Huan Zhang
Cho-Jui Hsieh (UCLA, Google)
Wei Zhang (IBM T.J.Watson Research Center)

BE Beijing Univ of Technology 2005 MSc Technical University of Denmark 2008 PhD University of Wisconsin, Madison 2013 All in computer science Published papers in ASPLOS, OOPSLA, OSDI, PLDI, IJCAI, ICDM, NIPS

Ji Liu (University of Rochester)

Related Events (a corresponding poster, oral, or spotlight)

More from the Same Authors