论文信息 - A Note on the PAC Bayesian Theorem

A Note on the PAC Bayesian Theorem

We prove general exponential moment inequalities for averages of [0,1]-valued iid random variables and use them to tighten the PAC Bayesian Theorem. The logarithmic dependence on the sample count in the enumerator of the PAC Bayesian bound is halved.

Andreas Maurer | Andreas Maurer

[1] Thomas M. Cover,et al. Elements of Information Theory , 2005 .

[2] David A. McAllester. Simplified PAC-Bayesian Margin Bounds , 2003, COLT.

[3] H. Chernoff. A Measure of Asymptotic Efficiency for Tests of a Hypothesis Based on the sum of Observations , 1952 .

[4] Jean-Yves Audibert,et al. PAC-Bayesian Generic Chaining , 2003, NIPS.

[5] W. Hoeffding. Probability Inequalities for sums of Bounded Random Variables , 1963 .

[6] John Shawe-Taylor,et al. PAC Bayes and Margins , 2003 .

[7] C. McDiarmid. Concentration , 1862, The Dental register.

[8] Matthias W. Seeger,et al. PAC-Bayesian Generalisation Error Bounds for Gaussian Process Classification , 2003, J. Mach. Learn. Res..

[9] Nello Cristianini,et al. Neural Information Processing Systems (NIPS) , 2003 .

[10] David A. McAllester. Some PAC-Bayesian Theorems , 1998, COLT' 98.

[11] O. Catoni. A PAC-Bayesian approach to adaptive classification , 2004 .

[12] David A. McAllester. PAC-Bayesian Stochastic Model Selection , 2003, Machine Learning.