Private PAC learning implies finite Littlestone dimension

We show that every approximately differentially private learning algorithm (possibly improper) for a class H with Littlestone dimension d requires Ω(log*(d)) examples. As a corollary it follows that the class of thresholds over ℕ can not be learned in a private manner; this resolves open questions due to [Bun et al. 2015] and [Feldman and Xiao, 2015]. We leave as an open question whether every class with a finite Littlestone dimension can be learned by an approximately differentially private algorithm.

[1]  Anand D. Sarwate,et al.  Differentially Private Empirical Risk Minimization , 2009, J. Mach. Learn. Res..

[2]  W. Hodges.  CLASSIFICATION THEORY AND THE NUMBER OF NON‐ISOMORPHIC MODELS , 1980 .

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

[4]  Michael C. Laskowski,et al.  Vapnik-Chervonenkis classes of definable sets , 1992 .

[5]  Raef Bassily,et al.  Model-Agnostic Private Learning via Stability , 2018, ArXiv.

[6]  Maria-Florina Balcan,et al.  Statistical Active Learning Algorithms for Noise Tolerance and Differential Privacy , 2013, Algorithmica.

[7]  David Haussler,et al.  Learnability and the Vapnik-Chervonenkis dimension , 1989, JACM.

[8]  Vitaly Feldman,et al.  Privacy-preserving Prediction , 2018, COLT.

[9]  Shai Ben-David,et al.  Agnostic Online Learning , 2009, COLT.

[10]  Wilfrid Hodges,et al.  A Shorter Model Theory , 1997 .

[11]  Chen C. Chang,et al.  Model Theory: Third Edition (Dover Books On Mathematics) By C.C. Chang;H. Jerome Keisler;Mathematics , 1966 .

[12]  Cynthia Dwork,et al.  Calibrating Noise to Sensitivity in Private Data Analysis , 2016, J. Priv. Confidentiality.

[13]  Kamalika Chaudhuri,et al.  The Large Margin Mechanism for Differentially Private Maximization , 2014, NIPS.

[14]  Stephen E. Fienberg,et al.  Learning with Differential Privacy: Stability, Learnability and the Sufficiency and Necessity of ERM Principle , 2015, J. Mach. Learn. Res..

[15]  Raef Bassily,et al.  Algorithmic stability for adaptive data analysis , 2015, STOC.

[16]  Vitaly Feldman,et al.  Sample Complexity Bounds on Differentially Private Learning via Communication Complexity , 2014, SIAM J. Comput..

[17]  Amos Beimel,et al.  Bounds on the sample complexity for private learning and private data release , 2010, Machine Learning.

[18]  Amos Beimel,et al.  Characterizing the sample complexity of private learners , 2013, ITCS '13.

[19]  Seth Neel,et al.  Accuracy First: Selecting a Differential Privacy Level for Accuracy Constrained ERM , 2017, NIPS.

[20]  Kobbi Nissim,et al.  Simultaneous Private Learning of Multiple Concepts , 2015, ITCS.

[21]  Dhruv Mubayi,et al.  A survey of quantitative bounds for hypergraph Ramsey problems , 2017 .

[22]  Vladimir Vapnik,et al.  Chervonenkis: On the uniform convergence of relative frequencies of events to their probabilities , 1971 .

[23]  Cynthia Dwork,et al.  Practical privacy: the SuLQ framework , 2005, PODS.

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

[25]  N. Littlestone.  Learning Quickly When Irrelevant Attributes Abound: A New Linear-Threshold Algorithm , 1987, 28th Annual Symposium on Foundations of Computer Science (sfcs 1987).

[26]  Roi Livni,et al.  Honest Compressions and Their Application to Compression Schemes , 2013, COLT.

[27]  Shai Ben-David,et al.  Understanding Machine Learning: From Theory to Algorithms , 2014 .

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

[29]  Amos Beimel,et al.  Learning Privately with Labeled and Unlabeled Examples , 2014, Algorithmica.

[30]  Amos Beimel,et al.  Private Learning and Sanitization: Pure vs. Approximate Differential Privacy , 2013, APPROX-RANDOM.

[31]  Marek Karpinski,et al.  Polynomial Bounds for VC Dimension of Sigmoidal and General Pfaffian Neural Networks , 1997, J. Comput. Syst. Sci..

[32]  Hunter Chase,et al.  Model Theory and Machine Learning , 2019, Bull. Symb. Log..

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

[34]  P. Erdös,et al.  Combinatorial Theorems on Classifications of Subsets of a Given Set , 1952 .

[35]  Ling Huang,et al.  Learning in a Large Function Space: Privacy-Preserving Mechanisms for SVM Learning , 2009, J. Priv. Confidentiality.

[36]  Sofya Raskhodnikova,et al.  What Can We Learn Privately? , 2008, 2008 49th Annual IEEE Symposium on Foundations of Computer Science.

[37]  Kobbi Nissim,et al.  Differentially Private Release and Learning of Threshold Functions , 2015, 2015 IEEE 56th Annual Symposium on Foundations of Computer Science.

[38]  Mark Bun,et al.  New Separations in the Complexity of Differential Privacy , 2016 .

[39]  Salil P. Vadhan,et al.  The Complexity of Differential Privacy , 2017, Tutorials on the Foundations of Cryptography.

[40]  Cynthia Dwork,et al.  Differential privacy and robust statistics , 2009, STOC '09.