Supermartingales in prediction with expert advice

被引:8
作者
Chernov, Alexey [1 ]
Kalnishkan, Yuri [1 ]
Zhdanov, Fedor [1 ]
Vovk, Vladimir [1 ]
机构
[1] Univ London, Comp Learning Res Ctr, Dept Comp Sci, Egham TW20 0EX, Surrey, England
基金
英国工程与自然科学研究理事会;
关键词
Prediction with expert advice; Defensive forecasting algorithm; Aggregating algorithm; PROPER SCORING RULES; INTERNAL REGRET; GAME;
D O I
10.1016/j.tcs.2010.04.003
中图分类号
TP301 [理论、方法];
学科分类号
081202 ;
摘要
The paper applies the method of defensive forecasting, based on the use of game-theoretic supermartingales, to prediction with expert advice. In the traditional setting of a countable number of experts and a finite number of outcomes, the Defensive Forecasting Algorithm is very close to the well-known Aggregating Algorithm. Not only the performance guarantees but also the predictions are the same for these two methods of fundamentally different nature. The paper also discusses a new setting where the experts can give advice conditional on the learner's future decision. Both the algorithms can be adapted to the new setting and give the same performance guarantees as in the traditional setting. Finally, an application of defensive forecasting to a setting with several loss functions is outlined. (C) 2010 Elsevier B.V. All rights reserved.
引用
收藏
页码:2647 / 2669
页数:23
相关论文
共 32 条
  • [1] Agarwal R. P., 2001, CAMBRIDGE TRACTS MAT, V141
  • [2] [Anonymous], 2001, Probability and Finance: It's Only a Game!
  • [3] Blackwell D, 1954, Theory of Games and Statistical Decisions
  • [4] Blum A, 2007, J MACH LEARN RES, V8, P1307
  • [5] Cesa-Bianchi N., 2006, PREDICTION LEARNING
  • [6] CHERNOV A, 2009, ARXIV09024127V1CSLG
  • [7] Chernov A, 2008, LECT NOTES ARTIF INT, V5254, P199, DOI 10.1007/978-3-540-87987-9_19
  • [8] Chernov A, 2009, LECT NOTES ARTIF INT, V5809, P8, DOI 10.1007/978-3-642-04414-4_6
  • [9] The geometry of proper scoring rules
    Dawid, A. P.
    [J]. ANNALS OF THE INSTITUTE OF STATISTICAL MATHEMATICS, 2007, 59 (01) : 77 - 93
  • [10] FOSTER D, 1999, GAME ECON BEHAV, V29, P104