A Variational Approach to Learning Curves

NeurIPS 2001 pp. 463-469

/neurips/2001/malzahn2001neurips-variational/

Abstract

We combine the replica approach from statistical physics with a varia- tional approach to analyze learning curves analytically. We apply the method to Gaussian process regression. As a main result we derive ap- proximative relations between empirical error measures, the generaliza- tion error and the posterior variance.

PDF NeurIPS Semantic Scholar

Cite

Text

Malzahn and Opper. "A Variational Approach to Learning Curves." Neural Information Processing Systems, 2001.

Markdown

[Malzahn and Opper. "A Variational Approach to Learning Curves." Neural Information Processing Systems, 2001.](https://mlanthology.org/neurips/2001/malzahn2001neurips-variational/)

BibTeX

@inproceedings{malzahn2001neurips-variational,
  title     = {{A Variational Approach to Learning Curves}},
  author    = {Malzahn, Dörthe and Opper, Manfred},
  booktitle = {Neural Information Processing Systems},
  year      = {2001},
  pages     = {463-469},
  url       = {https://mlanthology.org/neurips/2001/malzahn2001neurips-variational/}
}