Well-Calibrated Predictions from Online Compression Models

Abstract

It has been shown recently that Transductive Confidence Machine (TCM) is automatically well-calibrated when used in the on-line mode and provided that the data sequence is generated by an exchangeable distribution. In this paper we strengthen this result by relaxing the assumption of exchangeability of the data-generating distribution to the much weaker assumption that the data agrees with a given “on-line compression model”.

Cite

Text

Vovk. "Well-Calibrated Predictions from Online Compression Models." International Conference on Algorithmic Learning Theory, 2003. doi:10.1007/978-3-540-39624-6_22

Markdown

[Vovk. "Well-Calibrated Predictions from Online Compression Models." International Conference on Algorithmic Learning Theory, 2003.](https://mlanthology.org/alt/2003/vovk2003alt-wellcalibrated/) doi:10.1007/978-3-540-39624-6_22

BibTeX

@inproceedings{vovk2003alt-wellcalibrated,
  title     = {{Well-Calibrated Predictions from Online Compression Models}},
  author    = {Vovk, Vladimir},
  booktitle = {International Conference on Algorithmic Learning Theory},
  year      = {2003},
  pages     = {268-282},
  doi       = {10.1007/978-3-540-39624-6_22},
  url       = {https://mlanthology.org/alt/2003/vovk2003alt-wellcalibrated/}
}