Criterion of Calibration for Transductive Confidence Machine with Limited Feedback
Abstract
This paper is concerned with the problem of on-line prediction in the situation where some data is unlabelled and can never be used for prediction, and even when data is labelled, the labels may arrive with a delay. We construct a modification of randomised Transductive Confidence Machine for this case and prove a necessary and sufficient condition for its predictions being calibrated, in the sense that in the long run they are wrong with a prespecified probability under the assumption that data is generated independently by same distribution. The condition for calibration turns out to be very weak: feedback should be given on more than a logarithmic fraction of steps.
Cite
Text
Nouretdinov and Vovk. "Criterion of Calibration for Transductive Confidence Machine with Limited Feedback." International Conference on Algorithmic Learning Theory, 2003. doi:10.1007/978-3-540-39624-6_21Markdown
[Nouretdinov and Vovk. "Criterion of Calibration for Transductive Confidence Machine with Limited Feedback." International Conference on Algorithmic Learning Theory, 2003.](https://mlanthology.org/alt/2003/nouretdinov2003alt-criterion/) doi:10.1007/978-3-540-39624-6_21BibTeX
@inproceedings{nouretdinov2003alt-criterion,
title = {{Criterion of Calibration for Transductive Confidence Machine with Limited Feedback}},
author = {Nouretdinov, Ilia and Vovk, Vladimir},
booktitle = {International Conference on Algorithmic Learning Theory},
year = {2003},
pages = {259-267},
doi = {10.1007/978-3-540-39624-6_21},
url = {https://mlanthology.org/alt/2003/nouretdinov2003alt-criterion/}
}