Leading Strategies in Competitive On-Line Prediction

  • Vladimir Vovk
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 4264)


We start from a simple asymptotic result for the problem of on-line regression with the quadratic loss function: the class of continuous limited-memory prediction strategies admits a “leading prediction strategy”, which not only asymptotically performs at least as well as any continuous limited-memory strategy but also satisfies the property that the excess loss of any continuous limited-memory strategy is determined by how closely it imitates the leading strategy. More specifically, for any class of prediction strategies constituting a reproducing kernel Hilbert space we construct a leading strategy, in the sense that the loss of any prediction strategy whose norm is not too large is determined by how closely it imitates the leading strategy. This result is extended to the loss functions given by Bregman divergences and by strictly proper scoring rules.


Loss Function Reproduce Kernel Hilbert Space Prediction Strategy Predictable Process Quadratic Loss Function 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. 1.
    Adams, R.A., Fournier, J.J.F.: Sobolev Spaces, 2nd edn. Pure and Applied Mathematics, vol. 140. Academic Press, Amsterdam (2003)MATHGoogle Scholar
  2. 2.
    Auer, P., Cesa-Bianchi, N., Gentile, C.: Adaptive and self-confident on-line learning algorithms. Journal of Computer and System Sciences 64, 48–75 (2002)MATHCrossRefMathSciNetGoogle Scholar
  3. 3.
    Azoury, K.S., Warmuth, M.K.: Relative loss bounds for on-line density estimation with the exponential family of distributions. Machine Learning 43, 211–246 (2001)MATHCrossRefGoogle Scholar
  4. 4.
    Blackwell, D., Dubins, L.: Merging of opinions with increasing information. Annals of Mathematical Statistics 33, 882–886 (1962)MATHCrossRefMathSciNetGoogle Scholar
  5. 5.
    Bregman, L.M.: The relaxation method of finding the common point of convex sets and its application to the solution of problems in convex programming. USSR Computational Mathematics and Physics 7, 200–217 (1967)CrossRefGoogle Scholar
  6. 6.
    Cesa-Bianchi, N., Long, P.M., Warmuth, M.K.: Worst-case quadratic loss bounds for on-line prediction of linear functions by gradient descent. IEEE Transactions on Neural Networks 7, 604–619 (1996)CrossRefGoogle Scholar
  7. 7.
    Cesa-Bianchi, N., Lugosi, G.: Prediction, Learning, and Games. Cambridge University Press, Cambridge (2006)MATHCrossRefGoogle Scholar
  8. 8.
    Cox, D.R., Hinkley, D.V.: Theoretical Statistics. Chapman and Hall, London (1974)MATHGoogle Scholar
  9. 9.
    Dawid, A.P.: Statistical theory: the prequential approach. Journal of the Royal Statistical Society A 147, 278–292 (1984)MATHCrossRefMathSciNetGoogle Scholar
  10. 10.
    Dawid, A.P.: Calibration-based empirical probability (with discussion). Annals of Statistics 13, 1251–1285 (1985)MATHCrossRefMathSciNetGoogle Scholar
  11. 11.
    Dawid, A.P.: Proper measures of discrepancy, uncertainty and dependence, with applications to predictive experimental design. Technical Report 139, Department of Statistical Science, University College London, November 1994. This technical report was revised (and its title was slightly changed) in August 1998Google Scholar
  12. 12.
    Dawid, A.P.: Probability, causality and the empirical world: a Bayes–de Finetti–Popper–Borel synthesis. Statistical Science 19, 44–57 (2004)MATHCrossRefMathSciNetGoogle Scholar
  13. 13.
    Ellul, J.: The Technological Bluff. Eerdmans, Grand Rapids, MI (1990), Translated by Bromiley, G.W.: The French original: Le bluff technologique, Hachette, Paris, 1988Google Scholar
  14. 14.
    Helmbold, D.P., Kivinen, J., Warmuth, M.K.: Relative loss bounds for single neurons. IEEE Transactions on Neural Networks 10, 1291–1304 (1999)CrossRefGoogle Scholar
  15. 15.
    Herbster, M., Warmuth, M.K.: Tracking the best linear predictor. Journal of Machine Learning Research 1, 281–309 (2001)MATHCrossRefMathSciNetGoogle Scholar
  16. 16.
    Kabanov, Y.M., Liptser, R.S., Shiryaev, A.N.: To the question of absolute continuity and singularity of probability measures. Matematicheskii Sbornik 104, 227–247 (1977) (in Russian)Google Scholar
  17. 17.
    Kivinen, J., Warmuth, M.K.: Relative loss bounds for multidimensional regression problems. Machine Learning 45, 301–329 (2001)MATHCrossRefGoogle Scholar
  18. 18.
    Levin, L.A.: On the notion of a random sequence. Soviet Mathematics Doklady 14, 1413–1416 (1973)MATHGoogle Scholar
  19. 19.
    Martin-Löf, P.: The definition of random sequences. Information and Control 9, 602–619 (1966)CrossRefMathSciNetGoogle Scholar
  20. 20.
    Schnorr, C.P.: Zufälligkeit und Wahrscheinlichkeit. Springer, Berlin (1971)MATHGoogle Scholar
  21. 21.
    Shafer, G., Vovk, V.: Probability and Finance: It’s Only a Game!. Wiley, New York (2001)CrossRefGoogle Scholar
  22. 22.
    Solomonoff, R.J.: Complexity-based induction systems: comparisons and convergence theorems. IEEE Transactions on Information Theory IT-24, 422–432 (1978)MATHCrossRefMathSciNetGoogle Scholar
  23. 23.
    Ville, J.: Etude critique de la notion de collectif. In: Gauthier-Villars, Paris (1939)Google Scholar
  24. 24.
    Vovk, V.: On a randomness criterion. Soviet Mathematics Doklady 35, 656–660 (1987)MATHGoogle Scholar
  25. 25.
    Vovk, V.: Probability theory for the Brier game. Theoretical Computer Science, 1997 261, 57–79 (2001); Conference version in: Li, M. (ed.) ALT 1997. LNCS, vol. 1316, pp. 57–79. Springer, Heidelberg (1997)Google Scholar
  26. 26.
    Vovk, V.: Defensive prediction with expert advice. In: Jain, S., Simon, H.U., Tomita, E. (eds.) ALT 2005. LNCS (LNAI), vol. 3734, pp. 444–458. Springer, Heidelberg (2005); Full version: Technical Report arXiv:cs.LG/0506041 “Competitive on-line learning with a convex loss function” (version 3), e-Print archive (September 2005)CrossRefGoogle Scholar
  27. 27.
    Vovk, V.: Non-asymptotic calibration and resolution. In: Jain, S., Simon, H.U., Tomita, E. (eds.) ALT 2005. LNCS (LNAI), vol. 3734, pp. 429–443. Springer, Heidelberg (2005); A version of this paper can be downloaded from the e-Print archive (arXiv:cs.LG/0506004)CrossRefGoogle Scholar
  28. 28.
    Vovk, V.: Competing with Markov prediction strategies. Technical report, e-Print archive (July 2006)Google Scholar
  29. 29.
    Vovk, V.: Competing with stationary prediction strategies. Technical Report arXiv:cs.LG/0607067, e-Print archive (July 2006)Google Scholar
  30. 30.
    Vovk, V.: Competing with wild prediction rules. In: Lugosi, G., Simon, H.U. (eds.) COLT 2006. LNCS (LNAI), vol. 4005, pp. 559–573. Springer, Heidelberg (2006); Full version: Technical Report arXiv:cs.LG/0512059 (version 2), e-Print archive (January 2006) CrossRefGoogle Scholar
  31. 31.
    Vovk, V.: Leading strategies in competitive on-line prediction. Technical Report arXiv:cs.LG/0607134, e-Print archive (July 2006)Google Scholar
  32. 32.
    Vovk, V.: On-line regression competitive with reproducing kernel Hilbert spaces. Technical Report arXiv:cs.LG/00511058 (version 2), e-Print archive (January 2006); Cai, J.-Y., Cooper, S.B., Li, A. (eds.) TAMC 2006. LNCS, vol. 3959, pp. 452–463. Springer, Heidelberg (extended abstract, 2006)Google Scholar
  33. 33.
    Vovk, V.: Predictions as statements and decisions. In: Lugosi, G., Simon, H.U. (eds.) COLT 2006. LNCS (LNAI), vol. 4005, p. 4. Springer, Heidelberg (2006); Full version: Technical Report arXiv:cs.LG/0606093, e-Print archive (June 2006) CrossRefGoogle Scholar
  34. 34.
    Vovk, V., Takemura, A., Shafer, G.: Defensive forecasting. In: Cowell, R.G., Ghahramani, Z. (eds.) Proceedings of the Tenth International Workshop on Artificial Intelligence and Statistics. Society for Artificial Intelligence and Statistics, pp. 365–372 (2005), Available electronically at:

Copyright information

© Springer-Verlag Berlin Heidelberg 2006

Authors and Affiliations

  • Vladimir Vovk
    • 1
  1. 1.Computer Learning Research Centre, Department of Computer ScienceUniversity of LondonEgham, SurreyUK

Personalised recommendations