The reusable holdout: Preserving validity in adaptive data analysis

Testing hypotheses privately Large data sets offer a vast scope for testing already-formulated ideas and exploring new ones. Unfortunately, researchers who attempt to do both on the same data set run the risk of making false discoveries, even when testing and exploration are carried out on distinct subsets of data. Based on ideas drawn from differential privacy, Dwork et al. now provide a theoretical solution. Ideas are tested against aggregate information, whereas individual data set components remain confidential. Preserving that privacy also preserves statistical inference validity. Science, this issue p. 636 A statistical approach allows large data sets to be reanalyzed to test new hypotheses. Misapplication of statistical data analysis is a common cause of spurious discoveries in scientific research. Existing approaches to ensuring the validity of inferences drawn from data assume a fixed procedure to be performed, selected before the data are examined. In common practice, however, data analysis is an intrinsically adaptive process, with new analyses generated on the basis of data exploration, as well as the results of previous analyses on the same data. We demonstrate a new approach for addressing the challenges of adaptivity based on insights from privacy-preserving data analysis. As an application, we show how to safely reuse a holdout data set many times to validate the results of adaptively chosen analyses.

[1]  D. Freedman.  A Note on Screening Regression Equations , 1983 .

[2]  W. E. Riggs.  An Experimental Evaluation , 1983 .

[3]  Y. Benjamini,et al.  Controlling the false discovery rate: a practical and powerful approach to multiple testing , 1995 .

[4]  M. Kearns.  Efficient noise-tolerant learning from statistical queries , 1998, JACM.

[5]  André Elisseeff,et al.  Stability and Generalization , 2002, J. Mach. Learn. Res..

[6]  VACANT-PROPERTY Policy,et al.  THE BROOKINGS INSTITUTION , 2002 .

[7]  Juha Reunanen,et al.  Overfitting in Making Comparisons Between Variable Selection Methods , 2003, J. Mach. Learn. Res..

[8]  T. Poggio,et al.  General conditions for predictivity in learning theory , 2004, Nature.

[9]  D. Ruppert.  The Elements of Statistical Learning: Data Mining, Inference, and Prediction , 2004 .

[10]  J. Ioannidis.  Why Most Published Research Findings Are False , 2005, PLoS medicine.

[11]  Moni Naor,et al.  Our Data, Ourselves: Privacy Via Distributed Noise Generation , 2006, EUROCRYPT.

[12]  Kunle Olukotun,et al.  Map-Reduce for Machine Learning on Multicore , 2006, NIPS.

[13]  Cynthia Dwork,et al.  Calibrating Noise to Sensitivity in Private Data Analysis , 2006, TCC.

[14]  Mohammad Taghi Hajiaghayi,et al.  Regret minimization and the price of total anarchy , 2008, STOC.

[15]  C. Apté,et al.  Proceedings of the SIAM International Conference on Data Mining, SDM 2008, April 24-26, 2008, Atlanta, Georgia, USA , 2010, SDM.

[16]  Dean P. Foster,et al.  α‐investing: a procedure for sequential control of expected false discoveries , 2008 .

[17]  R. Rosenfeld.  Nature , 2009, Otolaryngology--head and neck surgery : official journal of American Academy of Otolaryngology-Head and Neck Surgery.

[18]  Adam Tauman Kalai,et al.  On the equilibria of alternating move games , 2010, SODA '10.

[19]  Ohad Shamir,et al.  Learnability, Stability and Uniform Convergence , 2010, J. Mach. Learn. Res..

[20]  Tim Roughgarden,et al.  Interactive privacy via the median mechanism , 2009, STOC '10.

[21]  Gavin C. Cawley,et al.  On Over-fitting in Model Selection and Subsequent Selection Bias in Performance Evaluation , 2010, J. Mach. Learn. Res..

[22]  Leif D. Nelson,et al.  False-Positive Psychology , 2011, Psychological science.

[23]  Aaron Roth,et al.  Privately releasing conjunctions and the statistical query barrier , 2010, STOC '11.

[24]  Saharon Rosset,et al.  The Quality Preserving Database: A Computational Framework for Encouraging Collaboration, Enhancing Power and Controlling False Discovery , 2011, IEEE/ACM Transactions on Computational Biology and Bioinformatics.

[25]  Aaron Roth,et al.  Fast Private Data Release Algorithms for Sparse Queries , 2013, APPROX-RANDOM.

[26]  Aaron Roth,et al.  A learning theory approach to non-interactive database privacy , 2008, STOC.

[27]  Nicole Immorlica,et al.  Constrained signaling for welfare and revenue maximization , 2013, SECO.

[28]  Zhiyi Huang,et al.  New techniques for computation over private data , 2013 .

[29]  Paul W. Goldberg,et al.  Bounds for the Query Complexity of Approximate Equilibria , 2016, ACM Trans. Economics and Comput..

[30]  Justin Hsu,et al.  Differential privacy for the analyst via private equilibrium computation , 2012, STOC '13.

[31]  Aaron Roth,et al.  The Algorithmic Foundations of Differential Privacy , 2014, Found. Trends Theor. Comput. Sci..

[32]  Aaron Roth,et al.  Exploiting Metric Structure for Efficient Private Query Release , 2014, SODA.

[33]  Marco Gaboardi,et al.  Dual Query: Practical Private Query Release for High Dimensional Data , 2014, ICML.

[34]  Gilles Barthe,et al.  Higher-Order Approximate Relational Refinement Types for Mechanism Design and Differential Privacy , 2014, POPL.

[35]  Andreas Haeberlen,et al.  Differential Privacy: An Economic Method for Choosing Epsilon , 2014, 2014 IEEE 27th Computer Security Foundations Symposium.

[36]  A. Gelman,et al.  The statistical crisis in science , 2014 .

[37]  Aaron Roth,et al.  Inducing Approximately Optimal Flow Using Truthful Mediators , 2015, EC.

[38]  Aaron Roth,et al.  Accuracy for Sale: Aggregating Data with a Variance Constraint , 2015, ITCS.

[39]  Adel Javanmard,et al.  On Online Control of False Discovery Rate , 2015, ArXiv.

[40]  Aaron Roth,et al.  Watch and learn: optimizing from revealed preferences feedback , 2015, SECO.

[41]  Sampath Kannan,et al.  Approximately Stable, School Optimal, and Student-Truthful Many-to-One Matchings (via Differential Privacy) , 2014, SODA.

[42]  Salil P. Vadhan,et al.  Theory of Cryptography , 2016, Lecture Notes in Computer Science.

[43]  Aaron Roth,et al.  Adaptive Learning with Robust Generalization Guarantees , 2016, COLT.

[44]  Aaron Roth,et al.  Fairness in Learning: Classic and Contextual Bandits , 2016, NIPS.

[45]  Aaron Roth,et al.  Learning from Rational Behavior: Predicting Solutions to Unknown Linear Programs , 2016, NIPS.

[46]  Aaron Roth,et al.  The Strange Case of Privacy in Equilibrium Models , 2015, EC.