Abstract
Hoeffding racing algorithms are used to achieve computational speedups in settings where the goal is to select a “best” option among a set of alternatives, but the amount of data is so massive that scoring all candidates using every data point is too costly. The key is to construct confidence intervals for scores of candidates that are used to eliminate options sequentially as more samples are processed. We propose a tighter version of Hoeffding racing based on empirical Bernstein inequalities, where a jackknife estimate is used in place of the unknown variance. We provide rigorous proofs of the accuracy of our confidence intervals in the case of U-statistics and entropy estimators, and demonstrate the efficacy of our racing algorithms with synthetic experiments.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
References
Antos, A., Kontoyiannis, I.: Convergence properties of functional estimates for discrete distributions. Random Structures and Algorithms 19(3-4), 163–193 (2001)
Arcones, M.: A Bernstein-type inequality for U-statistics and U-processes. Statistics and Probability Letters 22(3), 239–247 (1995)
Audibert, J.-Y., Munos, R., Szepasvari, C.: Exploration-exploitation tradeoff using variance estimates in multi-armed bandits. Theoretical Computer Science 410(19), 1876–1902 (2009)
Audibert, J.-Y., Munos, R., Szepesvári, C.: Tuning bandit algorithms in stochastic environments. In: Hutter, M., Servedio, R.A., Takimoto, E. (eds.) ALT 2007. LNCS (LNAI), vol. 4754, pp. 150–165. Springer, Heidelberg (2007)
Bernstein, S.N.: The Theory of Probabilities. Gastehizdat Publishing House, Moscow (1946)
Berry, D.A., Fristedt, B.: Bandit Problems. Chapman and Hall (1985)
Boucheron, S., Lugosi, G., Massart, P.: Concentration inequalities using the entropy method. Annals of Probability 31(3), 1583–1614 (2003)
DasGupta, A.: Asymptotic Theory of Statistics and Probability. Springer (2008)
Domingos, P., Hulten, G.: Mining high-speed data streams. In: KDD, pp. 71–80 (2000)
Dubhashi, D.P., Panconesi, A.: Concentration of Measure for the Analysis of Randomized Algorithms. Cambridge University Press (2009)
Efron, B., Stein, C.: The jackknife estimator of variance. Annals of Statistics 9, 586–596 (1981)
Even-Dar, E., Mannor, S., Mansour, Y.: Action elimination and stopping conditions for the multi-armed bandit and reinforcement learning problems. JMLR 7, 1079–1105 (2006)
Hoeffding, W.: Probability inequalities for sums of bounded random variables. JASA 58(301), 13–30 (1963)
Ikonomovska, E., Gama, J., Zenko., B., Dzeroski, S.: Speeding up Hoeffding-based regression trees with options. In: Proceedings of ICML, pp. 537–544 (2011)
Jin, R., Agrawal, G.: Efficient decision tree construction on streaming data. In: Proceedings of the 9th ACM SIGKDD, pp. 571–576. ACM, New York (2003)
Lee, A.J.: U-statistics: Theory and Practice. CRC Press (1990)
Maron, O., Moore, A.W.: Hoeffding races: Accelerating model selection search for classification and function approximation. Advances in NIPS, 59–66 (1993)
McDiarmid, C.: On the method of bounded differences. Surveys in Combinatorics 141, 148–188 (1989)
Mnih, V., Szepesvári, C., Audibert, J.-Y.: Empirical Bernstein stopping. In: Proceedings of ICML, vol. 307, pp. 672–679. ACM (2008)
Paninski, L.: Estimation of entropy and mutual information. Neural Computation 15(6), 1191–1253 (2003)
Peel, T., Anthoine, S., Ralaivola, L.: Empirical Bernstein inequalities for U-statistics. Advances in NIPS 23, 1903–1911 (2010)
Pfahringer, B., Holmes, G., Kirkby, R.: New options for Hoeffding trees. In: Australian Conference on Artificial Intelligence, pp. 90–99 (2007)
Serfling, R.: Approximation theorems of mathematical statistics, Series in Probability and Mathematical Statistics, New York, NY (1980)
Steele, J.M.: An Efron-Stein inequality for nonsymmetric statistics. Annals of Statistics 14(2), 753–758 (1986)
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2013 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Loh, PL., Nowozin, S. (2013). Faster Hoeffding Racing: Bernstein Races via Jackknife Estimates. In: Jain, S., Munos, R., Stephan, F., Zeugmann, T. (eds) Algorithmic Learning Theory. ALT 2013. Lecture Notes in Computer Science(), vol 8139. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-40935-6_15
Download citation
DOI: https://doi.org/10.1007/978-3-642-40935-6_15
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-642-40934-9
Online ISBN: 978-3-642-40935-6
eBook Packages: Computer ScienceComputer Science (R0)