Nothing Special   »   [go: up one dir, main page]

skip to main content
10.1145/2897518.2897582acmconferencesArticle/Chapter ViewAbstractPublication PagesstocConference Proceedingsconference-collections
research-article

Communication lower bounds for statistical estimation problems via a distributed data processing inequality

Published: 19 June 2016 Publication History

Abstract

We study the tradeoff between the statistical error and communication cost of distributed statistical estimation problems in high dimensions. In the distributed sparse Gaussian mean estimation problem, each of the m machines receives n data points from a d-dimensional Gaussian distribution with unknown mean θ which is promised to be k-sparse. The machines communicate by message passing and aim to estimate the mean θ. We provide a tight (up to logarithmic factors) tradeoff between the estimation error and the number of bits communicated between the machines. This directly leads to a lower bound for the distributed sparse linear regression problem: to achieve the statistical minimax error, the total communication is at least Ω(min{n,d}m), where n is the number of observations that each machine receives and d is the ambient dimension. These lower results improve upon Shamir (NIPS'14) and Steinhardt-Duchi (COLT'15) by allowing multi-round iterative communication model. We also give the first optimal simultaneous protocol in the dense case for mean estimation. As our main technique, we prove a distributed data processing inequality, as a generalization of usual data processing inequalities, which might be of independent interest and useful for other problems.

References

[1]
{BJKS04} Ziv Bar-Yossef, T. S. Jayram, Ravi Kumar, and D. Sivakumar. An information statistics approach to data stream and communication complexity. J. Comput. Syst. Sci., 68(4):702–732, 2004.
[2]
{BWZ15} Christos Boutsidis, David P. Woodruff, and Peilin Zhong. Optimal principal component analysis in distributed and streaming models. CoRR, abs/1504.06729, 2015.
[3]
{BYJKS04} Ziv Bar-Yossef, T. S. Jayram, Ravi Kumar, and D. Sivakumar. An information statistics approach to data stream and communication complexity. J. Comput. Syst. Sci., 68(4), 2004.
[4]
{DAW12} John C Duchi, Alekh Agarwal, and Martin J Wainwright. Dual averaging for distributed optimization: convergence analysis and network scaling. Automatic Control, IEEE Transactions on, 57(3):592–606, 2012.
[5]
{DJWZ14} John C. Duchi, Michael I. Jordan, Martin J. Wainwright, and Yuchen Zhang. Information-theoretic lower bounds for distributed statistical estimation with communication constraints. CoRR, abs/1405.0782, 2014.
[6]
{GMN14} Ankit Garg, Tengyu Ma, and Huy L. Nguyen. On communication cost of distributed statistical estimation and dimensionality. In NIPS, pages 2726–2734, 2014.
[7]
{Jay09} T.S. Jayram. Hellinger strikes back: A note on the multi-party information complexity of and. In Irit Dinur, Klaus Jansen, Joseph Naor, and José Rolim, editors, Approximation, Randomization, and Combinatorial Optimization. Algorithms and Techniques, volume 5687 of Lecture Notes in Computer Science, pages 562–573. Springer Berlin Heidelberg, 2009.
[8]
{KVW14} Ravi Kannan, Santosh Vempala, and David P. Woodruff. Principal component analysis and higher correlations for distributed data. In Proceedings of The 27th Conference on Learning Theory, COLT 2014, Barcelona, Spain, June 13-15, 2014, pages 1040–1057, 2014.
[9]
{LBKW14} Yingyu Liang, Maria-Florina Balcan, Vandana Kanchanapally, and David P. Woodruff. Improved distributed principal component analysis. In NIPS, pages 3113–3121, 2014.
[10]
{LSLT15} Jason D Lee, Yuekai Sun, Qiang Liu, and Jonathan E Taylor. Communication-efficient sparse regression: a one-shot approach. arXiv preprint arXiv:1503.04337, 2015.
[11]
{Rag14} Maxim Raginsky. Strong data processing inequalities and $Φ$-sobolev inequalities for discrete channels. CoRR, abs/1411.3575, 2014.
[12]
{SD15} Jacob Steinhardt and John C. Duchi. Minimax rates for memory-bounded sparse linear regression. In Proceedings of The 28th Conference on Learning Theory, COLT 2015, Paris, France, July 3-6, 2015, pages 1564–1587, 2015.
[13]
{Sha14} Ohad Shamir. Fundamental limits of online and distributed algorithms for statistical learning and estimation. In NIPS, pages 163–171, 2014.
[14]
{SSZ14} Ohad Shamir, Nathan Srebro, and Tong Zhang. Communication-efficient distributed optimization using an approximate newton-type method. In Proceedings of the 31th International Conference on Machine Learning, ICML 2014, pages 1000–1008, 2014.
[15]
{WZ12} David P. Woodruff and Qin Zhang. Tight bounds for distributed functional monitoring. STOC, 2012.
[16]
{WZ16} David P. Woodruff and Peilin Zhong. Distributed low rank approximation of implicit functions of a matrix. CoRR, abs/1601.07721, 2016.
[17]
{ZDJW13} Yuchen Zhang, John C. Duchi, Michael I. Jordan, and Martin J. Wainwright. Information-theoretic lower bounds for distributed statistical estimation with communication constraints. In NIPS, pages 2328–2336, 2013.
[18]
{ZDW13} Yuchen Zhang, John C. Duchi, and Martin J. Wainwright. Communication-efficient algorithms for statistical optimization. Journal of Machine Learning Research, 14(1):3321–3363, 2013.
[19]
{ZX15} Yuchen Zhang and Lin Xiao. Communication-efficient distributed optimization of self-concordant empirical loss. CoRR, abs/1501.00263, 2015.

Cited By

View all
  • (2024)Communication-Constrained Hypothesis Testing: Optimality, Robustness, and Reverse Data Processing InequalitiesIEEE Transactions on Information Theory10.1109/TIT.2023.333402470:1(389-414)Online publication date: Jan-2024
  • (2024)Optimal Rates for Nonparametric Density Estimation Under Communication ConstraintsIEEE Transactions on Information Theory10.1109/TIT.2023.332590270:3(1939-1961)Online publication date: Mar-2024
  • (2024)Statistical Inference With Limited Memory: A SurveyIEEE Journal on Selected Areas in Information Theory10.1109/JSAIT.2024.34812965(623-644)Online publication date: 2024
  • Show More Cited By

Index Terms

  1. Communication lower bounds for statistical estimation problems via a distributed data processing inequality

        Recommendations

        Comments

        Please enable JavaScript to view thecomments powered by Disqus.

        Information & Contributors

        Information

        Published In

        cover image ACM Conferences
        STOC '16: Proceedings of the forty-eighth annual ACM symposium on Theory of Computing
        June 2016
        1141 pages
        ISBN:9781450341325
        DOI:10.1145/2897518
        Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for components of this work owned by others than ACM must be honored. Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. Request permissions from [email protected]

        Sponsors

        Publisher

        Association for Computing Machinery

        New York, NY, United States

        Publication History

        Published: 19 June 2016

        Permissions

        Request permissions for this article.

        Check for updates

        Author Tags

        1. Communication complexity
        2. Information complexity
        3. statistical estimation

        Qualifiers

        • Research-article

        Conference

        STOC '16
        Sponsor:
        STOC '16: Symposium on Theory of Computing
        June 19 - 21, 2016
        MA, Cambridge, USA

        Acceptance Rates

        Overall Acceptance Rate 1,469 of 4,586 submissions, 32%

        Contributors

        Other Metrics

        Bibliometrics & Citations

        Bibliometrics

        Article Metrics

        • Downloads (Last 12 months)54
        • Downloads (Last 6 weeks)7
        Reflects downloads up to 24 Nov 2024

        Other Metrics

        Citations

        Cited By

        View all
        • (2024)Communication-Constrained Hypothesis Testing: Optimality, Robustness, and Reverse Data Processing InequalitiesIEEE Transactions on Information Theory10.1109/TIT.2023.333402470:1(389-414)Online publication date: Jan-2024
        • (2024)Optimal Rates for Nonparametric Density Estimation Under Communication ConstraintsIEEE Transactions on Information Theory10.1109/TIT.2023.332590270:3(1939-1961)Online publication date: Mar-2024
        • (2024)Statistical Inference With Limited Memory: A SurveyIEEE Journal on Selected Areas in Information Theory10.1109/JSAIT.2024.34812965(623-644)Online publication date: 2024
        • (2024)SAM: An Efficient Approach With Selective Aggregation of Models in Federated LearningIEEE Internet of Things Journal10.1109/JIOT.2024.337382211:11(20769-20783)Online publication date: 1-Jun-2024
        • (2024)Lq Lower Bounds on Distributed Estimation via Fisher Information2024 IEEE International Symposium on Information Theory (ISIT)10.1109/ISIT57864.2024.10619530(91-96)Online publication date: 7-Jul-2024
        • (2024)Efficient Unbiased Sparsification2024 IEEE International Symposium on Information Theory (ISIT)10.1109/ISIT57864.2024.10619383(1854-1859)Online publication date: 7-Jul-2024
        • (2024)BAFFLE: A Baseline of Backpropagation-Free Federated LearningComputer Vision – ECCV 202410.1007/978-3-031-73226-3_6(89-109)Online publication date: 1-Nov-2024
        • (2024)Robust multitask learning in high dimensions under memory constraintStatistical Analysis and Data Mining: The ASA Data Science Journal10.1002/sam.1170017:3Online publication date: 17-Jun-2024
        • (2023)On robust streaming for learning with expertsProceedings of the 37th International Conference on Neural Information Processing Systems10.5555/3666122.3669602(79518-79539)Online publication date: 10-Dec-2023
        • (2023)Privacy amplification via compressionProceedings of the 37th International Conference on Neural Information Processing Systems10.5555/3666122.3669152(69202-69227)Online publication date: 10-Dec-2023
        • Show More Cited By

        View Options

        Login options

        View options

        PDF

        View or Download as a PDF file.

        PDF

        eReader

        View online with eReader.

        eReader

        Media

        Figures

        Other

        Tables

        Share

        Share

        Share this Publication link

        Share on social media