Defensive prediction with expert advice

被引:0
作者
Vovk, V [1 ]
机构
[1] Royal Holloway Univ London, Dept Comp Sci, Comp Learning Res Ctr, Egham TW20 0EX, Surrey, England
来源
ALGORITHMIC LEARNING THEORY | 2005年 / 3734卷
关键词
D O I
暂无
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
The theory of prediction with expert advice usually deals with countable or finite-dimensional pools of experts. In this paper we give similar results for pools of decision rules belonging to an infinite-dimensional functional spare which we call the Fermi-Sobolev space. For example, it is shown that for a wide class of loss functions (including the standard square, absolute, and log loss functions) the average loss of the master algorithm, over the first N steps, does not exceed the average loss of the best decision rule with a bounded Fermi-Sobolev norm plus O(N-1/2). Our proof techniques are very different from the standard ones and axe based on recent results about defensive forecasting. Given the probabilities produced by a defensive forecasting algorithm, which are known to be well calibrated and to have high resolution in the long run, we use the Expected Loss Minimization principle to find a suitable decision.
引用
收藏
页码:444 / 458
页数:15
相关论文
共 13 条
  • [1] [Anonymous], 1957, GEN TOPOLOGY
  • [2] How to use expert advice
    CesaBianchi, N
    Freund, Y
    Haussler, D
    Helmbold, DP
    Schapire, RE
    Warmuth, MK
    [J]. JOURNAL OF THE ACM, 1997, 44 (03) : 427 - 485
  • [3] ENGELKING R, 1989, SIGMA SERIES PURE MA, V6
  • [4] Freund Y., 1996, Proceedings of the Ninth Annual Conference on Computational Learning Theory, P89, DOI 10.1145/238061.238072
  • [5] KALNISHKAN Y, 2005, P 18 ANN C LEARN THE
  • [6] ONLINE LEARNING OF SMOOTH FUNCTIONS OF A SINGLE VARIABLE
    KIMBER, D
    LONG, PM
    [J]. THEORETICAL COMPUTER SCIENCE, 1995, 148 (01) : 141 - 156
  • [7] Exponentiated gradient versus gradient descent for linear predictors
    Kivinen, J
    Warmuth, MK
    [J]. INFORMATION AND COMPUTATION, 1997, 132 (01) : 1 - 63
  • [8] Improved bounds about on-line learning of smooth-functions of a single variable
    Long, PM
    [J]. THEORETICAL COMPUTER SCIENCE, 2000, 241 (1-2) : 25 - 35
  • [9] Rockafellar, 2015, CONVEX ANAL
  • [10] Vovk V, 2001, INT STAT REV, V69, P213