论文信息 - The one-inclusion graph algorithm is near-optimal for the prediction model of learning

The one-inclusion graph algorithm is near-optimal for the prediction model of learning

Haussler, Littlestone and Warmuth (1994) described a general-purpose algorithm for learning according to the prediction model, and proved an upper bound on the probability that their algorithm makes a mistake in terms of the number of examples seen and the Vapnik-Chervonenkis (VC) dimension of the concept class being learned. We show that their bound is within a factor of 1+o(1) of the best possible such bound for any algorithm.

Yi Li | Aravind Srinivasan | Philip M. Long

[1] Devdatt P. Dubhashi,et al. Balls and bins: A study in negative dependence , 1996, Random Struct. Algorithms.

[2] David Haussler,et al. Predicting {0,1}-functions on randomly drawn points , 1988, COLT '88.

[3] Y. Chien,et al. Pattern classification and scene analysis , 1974 .

[4] C. L. Mallows. An inequality involving multinomial probabilities , 1968 .

[5] Rajeev Motwani,et al. Randomized Algorithms , 1995, SIGA.

[6] Shun-ichi Amari,et al. A Theory of Pattern Recognition , 1968 .

[7] Leslie G. Valiant,et al. A general lower bound on the number of examples needed for learning , 1988, COLT '88.

[8] Peter E. Hart,et al. Pattern classification and scene analysis , 1974, A Wiley-Interscience publication.

[9] László Györfi,et al. A Probabilistic Theory of Pattern Recognition , 1996, Stochastic Modelling and Applied Probability.