DoubleSqueeze: Parallel Stochastic Gradient Descent with Double-pass Error-Compensated Compression

Hanlin Tang, Chen Yu, Xiangru Lian, Tong Zhang, Ji Liu
Proceedings of the 36th International Conference on Machine Learning, PMLR 97:6155-6165, 2019.

Abstract

A standard approach in large scale machine learning is distributed stochastic gradient training, which requires the computation of aggregated stochastic gradients over multiple nodes on a network. Communication is a major bottleneck in such applications, and in recent years, compressed stochastic gradient methods such as QSGD (quantized SGD) and sparse SGD have been proposed to reduce communication. It was also shown that error compensation can be combined with compression to achieve better convergence in a scheme that each node compresses its local stochastic gradient and broadcast the result to all other nodes over the network in a single pass. However, such a single pass broadcast approach is not realistic in many practical implementations. For example, under the popular parameter-server model for distributed learning, the worker nodes need to send the compressed local gradients to the parameter server, which performs the aggregation. The parameter server has to compress the aggregated stochastic gradient again before sending it back to the worker nodes. In this work, we provide a detailed analysis on this two-pass communication model, with error-compensated compression both on the worker nodes and on the parameter server. We show that the error-compensated stochastic gradient algorithm admits three very nice properties: 1) it is compatible with an arbitrary compression technique; 2) it admits an improved convergence rate than the non error-compensated stochastic gradient method such as QSGD and sparse SGD; 3) it admits linear speedup with respect to the number of workers. The empirical study is also conducted to validate our theoretical results.

Cite this Paper

BibTeX
@InProceedings{pmlr-v97-tang19d, title = {$\texttt{DoubleSqueeze}$: Parallel Stochastic Gradient Descent with Double-pass Error-Compensated Compression}, author = {Tang, Hanlin and Yu, Chen and Lian, Xiangru and Zhang, Tong and Liu, Ji}, booktitle = {Proceedings of the 36th International Conference on Machine Learning}, pages = {6155--6165}, year = {2019}, editor = {Chaudhuri, Kamalika and Salakhutdinov, Ruslan}, volume = {97}, series = {Proceedings of Machine Learning Research}, month = {09--15 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v97/tang19d/tang19d.pdf}, url = {https://proceedings.mlr.press/v97/tang19d.html}, abstract = {A standard approach in large scale machine learning is distributed stochastic gradient training, which requires the computation of aggregated stochastic gradients over multiple nodes on a network. Communication is a major bottleneck in such applications, and in recent years, compressed stochastic gradient methods such as QSGD (quantized SGD) and sparse SGD have been proposed to reduce communication. It was also shown that error compensation can be combined with compression to achieve better convergence in a scheme that each node compresses its local stochastic gradient and broadcast the result to all other nodes over the network in a single pass. However, such a single pass broadcast approach is not realistic in many practical implementations. For example, under the popular parameter-server model for distributed learning, the worker nodes need to send the compressed local gradients to the parameter server, which performs the aggregation. The parameter server has to compress the aggregated stochastic gradient again before sending it back to the worker nodes. In this work, we provide a detailed analysis on this two-pass communication model, with error-compensated compression both on the worker nodes and on the parameter server. We show that the error-compensated stochastic gradient algorithm admits three very nice properties: 1) it is compatible with an arbitrary compression technique; 2) it admits an improved convergence rate than the non error-compensated stochastic gradient method such as QSGD and sparse SGD; 3) it admits linear speedup with respect to the number of workers. The empirical study is also conducted to validate our theoretical results.} }
Endnote
%0 Conference Paper %T DoubleSqueeze: Parallel Stochastic Gradient Descent with Double-pass Error-Compensated Compression %A Hanlin Tang %A Chen Yu %A Xiangru Lian %A Tong Zhang %A Ji Liu %B Proceedings of the 36th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2019 %E Kamalika Chaudhuri %E Ruslan Salakhutdinov %F pmlr-v97-tang19d %I PMLR %P 6155--6165 %U https://proceedings.mlr.press/v97/tang19d.html %V 97 %X A standard approach in large scale machine learning is distributed stochastic gradient training, which requires the computation of aggregated stochastic gradients over multiple nodes on a network. Communication is a major bottleneck in such applications, and in recent years, compressed stochastic gradient methods such as QSGD (quantized SGD) and sparse SGD have been proposed to reduce communication. It was also shown that error compensation can be combined with compression to achieve better convergence in a scheme that each node compresses its local stochastic gradient and broadcast the result to all other nodes over the network in a single pass. However, such a single pass broadcast approach is not realistic in many practical implementations. For example, under the popular parameter-server model for distributed learning, the worker nodes need to send the compressed local gradients to the parameter server, which performs the aggregation. The parameter server has to compress the aggregated stochastic gradient again before sending it back to the worker nodes. In this work, we provide a detailed analysis on this two-pass communication model, with error-compensated compression both on the worker nodes and on the parameter server. We show that the error-compensated stochastic gradient algorithm admits three very nice properties: 1) it is compatible with an arbitrary compression technique; 2) it admits an improved convergence rate than the non error-compensated stochastic gradient method such as QSGD and sparse SGD; 3) it admits linear speedup with respect to the number of workers. The empirical study is also conducted to validate our theoretical results.
APA
Tang, H., Yu, C., Lian, X., Zhang, T. & Liu, J.. (2019). DoubleSqueeze: Parallel Stochastic Gradient Descent with Double-pass Error-Compensated Compression. Proceedings of the 36th International Conference on Machine Learning, in Proceedings of Machine Learning Research 97:6155-6165 Available from https://proceedings.mlr.press/v97/tang19d.html.

Related Material