skip to main content
article
Free access

Follow the leader if you can, hedge if you must

Published: 01 January 2014 Publication History

Abstract

Follow-the-Leader (FTL) is an intuitive sequential prediction strategy that guarantees constant regret in the stochastic setting, but has poor performance for worst-case data. Other hedging strategies have better worst-case guarantees but may perform much worse than FTL if the data are not maximally adversarial. We introduce the FlipFlop algorithm, which is the first method that provably combines the best of both worlds. As a stepping stone for our analysis, we develop AdaHedge, which is a new way of dynamically tuning the learning rate in Hedge without using the doubling trick. AdaHedge refines a method by Cesa-Bianchi, Mansour, and Stoltz (2007), yielding improved worst-case guarantees. By interleaving AdaHedge and FTL, FlipFlop achieves regret within a constant factor of the FTL regret, without sacrificing AdaHedge's worst-case guarantees. AdaHedge and FlipFlop do not need to know the range of the losses in advance; moreover, unlike earlier methods, both have the intuitive property that the issued weights are invariant under rescaling and translation of the losses. The losses are also allowed to be negative, in which case they may be interpreted as gains.

References

[1]
Jean-Yves Audibert. PAC-Bayesian statistical learning theory. PhD thesis, Université Paris VI, 2004.
[2]
Peter Auer, Nicolò Cesa-Bianchi, and Claudio Gentile. Adaptive and self-confident on-line learning algorithms. Journal of Computer and System Sciences, 64:48-75, 2002.
[3]
Olivier Catoni. PAC-Bayesian Supervised Classification. Lecture Notes-Monograph Series. IMS, 2007.
[4]
Nicolò Cesa-Bianchi and Gábor Lugosi. Prediction, learning, and games. Cambridge University Press, 2006.
[5]
Nicolò Cesa-Bianchi, Yoav Freund, David Haussler, David P. Helmbold, Robert E. Schapire, and Manfred K. Warmuth. How to use expert advice. Journal of the ACM, 44(3):427-485, 1997.
[6]
Nicolò Cesa-Bianchi, Yishay Mansour, and Gilles Stoltz. Improved second-order bounds for prediction with expert advice. Machine Learning, 66(2/3):321-352, 2007.
[7]
Kamalika Chaudhuri, Yoav Freund, and Daniel Hsu. A parameter-free hedging algorithm. In Advances in Neural Information Processing Systems 22 (NIPS 2009), pages 297-305, 2009.
[8]
Alexey V. Chernov and Vladimir Vovk. Prediction with advice of unknown number of experts. In Peter Grünwald and Peter Spirtes, editors, UAI, pages 117-125. AUAI Press, 2010.
[9]
Marie Devaine, Pierre Gaillard, Yannig Goude, and Gilles Stoltz. Forecasting electricity consumption by aggregating specialized experts; a review of the sequential aggregation of specialized experts, with an application to Slovakian and French country-wide one-day-ahead (half-)hourly predictions. Machine Learning, 90(2):231-260, February 2013.
[10]
Yoav Freund and Robert E. Schapire. A decision-theoretic generalization of on-line learning and an application to boosting. Journal of Computer and System Sciences, 55:119-139, 1997.
[11]
Yoav Freund and Robert E. Schapire. Adaptive game playing using multiplicative weights. Games and Economic Behavior, 29:79-103, 1999.
[12]
Sébastien Gerchinovitz. Prédiction de suites individuelles et cadre statistique classique: étude de quelques liens autour de la régression parcimonieuse et des techniques d'agrégation. PhD thesis, Université Paris-Sud, 2011.
[13]
Peter Grünwald. Safe learning: bridging the gap between Bayes, MDL and statistical learning theory via empirical convexity. In Proceedings of the 24th International Conference on Learning Theory (COLT 2011), pages 551-573, 2011.
[14]
Peter Grünwald. The safe Bayesian: learning the learning rate via the mixability gap. In Proceedings of the 23rd International Conference on Algorithmic Learning Theory (ALT 2012), 2012.
[15]
Peter Grünwald and John Langford. Suboptimal behavior of Bayes and MDL in classification under misspecification. Machine Learning, 66(2-3):119-149, 2007. DOI 10.1007/s10994-007-0716-7.
[16]
László Györfi and György Ottucsák. Sequential prediction of unbounded stationary time series. IEEE Transactions on Information Theory, 53(5):1866-1872, 2007.
[17]
Elad Hazan and Satyen Kale. Extracting certainty from uncertainty: Regret bounded by variation in costs. In Proceedings of the 21st Annual Conference on Learning Theory (COLT), pages 57-67, 2008.
[18]
Marcus Hutter and Jan Poland. Adaptive online prediction by following the perturbed leader. Journal of Machine Learning Research, 6:639-660, 2005.
[19]
Adam Kalai and Santosh Vempala. Efficient algorithms for online decision. In Proceedings of the 16st Annual Conference on Learning Theory (COLT), pages 506-521, 2003.
[20]
Yuri Kalnishkan and Michael V. Vyugin. The weak aggregating algorithm and weak mixability. In Proceedings of the 18th Annual Conference on Learning Theory (COLT), pages 188-203, 2005.
[21]
Wouter M. Koolen, Manfred K. Warmuth, and Jyrki Kivinen. Hedging structured concepts. In A.T. Kalai and M. Mohri, editors, Proceedings of the 23rd Annual Conference on Learning Theory (COLT 2010), pages 93-105, 2010.
[22]
Nick Littlestone and Manfred K. Warmuth. The weighted majority algorithm. Information and Computation, 108(2):212-261, 1994.
[23]
Koji Tsuda, Gunnar Rätsch, and Manfred K. Warmuth. Matrix exponentiated gradient updates for on-line learning and Bregman projection. Journal of Machine Learning Research, 6:995-1018, 2005.
[24]
Tim van Erven, Peter Grünwald, Wouter M. Koolen, and Steven de Rooij. Adaptive hedge. In Advances in Neural Information Processing Systems 24 (NIPS 2011), pages 1656-1664, 2011.
[25]
Vladimir Vovk. A game of prediction with expert advice. Journal of Computer and System Sciences, 56(2):153-173, 1998.
[26]
Vladimir Vovk. Competitive on-line statistics. International Statistical Review, 69(2):213- 248, 2001.
[27]
Vladimir Vovk, Akimichi Takemura, and Glenn Shafer. Defensive forecasting. In Proceedings of AISTATS 2005, 2005. Archive version available at http://www.vovk.net/df.
[28]
Tong Zhang. Information theoretical upper and lower bounds for statistical estimation. IEEE Transactions on Information Theory, 52(4):1307-1321, 2006.

Cited By

View all

Recommendations

Comments

Information & Contributors

Information

Published In

cover image The Journal of Machine Learning Research
The Journal of Machine Learning Research  Volume 15, Issue 1
January 2014
4085 pages
ISSN:1532-4435
EISSN:1533-7928
Issue’s Table of Contents

Publisher

JMLR.org

Publication History

Published: 01 January 2014
Revised: 01 January 2014
Published in JMLR Volume 15, Issue 1

Author Tags

  1. hedge
  2. learning rate
  3. mixability
  4. online learning
  5. prediction with expert advice

Qualifiers

  • Article

Contributors

Other Metrics

Bibliometrics & Citations

Bibliometrics

Article Metrics

  • Downloads (Last 12 months)125
  • Downloads (Last 6 weeks)16
Reflects downloads up to 22 Dec 2024

Other Metrics

Citations

Cited By

View all

View Options

View options

PDF

View or Download as a PDF file.

PDF

eReader

View online with eReader.

eReader

Login options

Full Access

Media

Figures

Other

Tables

Share

Share

Share this Publication link

Share on social media