ML Anthology

Distribution-Dependent PAC-Bayes Priors

Guy Lever, François Laviolette, John Shawe-Taylor
ALT 2010 pp. 119-133
doi:10.1007/978-3-642-16108-7_13 /alt/2010/lever2010alt-distributiondependent/

Abstract

We develop the idea that the PAC-Bayes prior can be informed by the data-generating distribution. We prove sharp bounds for an existing framework, and develop insights into function class complexity in this model and suggest means of controlling it with new algorithms. In particular we consider controlling capacity with respect to the unknown geometry of the data-generating distribution. We finally extend this localization to more practical learning methods.

PDF ALT Semantic Scholar

Cite

Text

Lever et al. "Distribution-Dependent PAC-Bayes Priors." International Conference on Algorithmic Learning Theory, 2010. doi:10.1007/978-3-642-16108-7_13

Markdown

[Lever et al. "Distribution-Dependent PAC-Bayes Priors." International Conference on Algorithmic Learning Theory, 2010.](https://mlanthology.org/alt/2010/lever2010alt-distributiondependent/) doi:10.1007/978-3-642-16108-7_13

BibTeX

@inproceedings{lever2010alt-distributiondependent,
  title     = {{Distribution-Dependent PAC-Bayes Priors}},
  author    = {Lever, Guy and Laviolette, François and Shawe-Taylor, John},
  booktitle = {International Conference on Algorithmic Learning Theory},
  year      = {2010},
  pages     = {119-133},
  doi       = {10.1007/978-3-642-16108-7_13},
  url       = {https://mlanthology.org/alt/2010/lever2010alt-distributiondependent/}
}