Conformal calibrators

Vladimir Vovk, Ivan Petej, Paolo Toccaceli, Alexander Gammerman, Ernst Ahlberg, Lars Carlsson
Proceedings of the Ninth Symposium on Conformal and Probabilistic Prediction and Applications, PMLR 128:84-99, 2020.

Abstract

Most existing examples of full conformal predictive systems, split conformal predictive systems, and cross-conformal predictive systems impose severe restrictions on the adaptation of predictive distributions to the test object at hand. In this paper we develop split conformal predictive systems that are fully adaptive. Our method consists in calibrating existing predictive systems; the input predictive system is not supposed to satisfy any properties of validity, whereas the output predictive system is guaranteed to be calibrated in probability.

Cite this Paper


BibTeX
@InProceedings{pmlr-v128-vovk20a, title = {Conformal calibrators}, author = {Vovk, Vladimir and Petej, Ivan and Toccaceli, Paolo and Gammerman, Alexander and Ahlberg, Ernst and Carlsson, Lars}, booktitle = {Proceedings of the Ninth Symposium on Conformal and Probabilistic Prediction and Applications}, pages = {84--99}, year = {2020}, editor = {Gammerman, Alexander and Vovk, Vladimir and Luo, Zhiyuan and Smirnov, Evgueni and Cherubin, Giovanni}, volume = {128}, series = {Proceedings of Machine Learning Research}, month = {09--11 Sep}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v128/vovk20a/vovk20a.pdf}, url = {https://proceedings.mlr.press/v128/vovk20a.html}, abstract = {Most existing examples of full conformal predictive systems, split conformal predictive systems, and cross-conformal predictive systems impose severe restrictions on the adaptation of predictive distributions to the test object at hand. In this paper we develop split conformal predictive systems that are fully adaptive. Our method consists in calibrating existing predictive systems; the input predictive system is not supposed to satisfy any properties of validity, whereas the output predictive system is guaranteed to be calibrated in probability.} }
Endnote
%0 Conference Paper %T Conformal calibrators %A Vladimir Vovk %A Ivan Petej %A Paolo Toccaceli %A Alexander Gammerman %A Ernst Ahlberg %A Lars Carlsson %B Proceedings of the Ninth Symposium on Conformal and Probabilistic Prediction and Applications %C Proceedings of Machine Learning Research %D 2020 %E Alexander Gammerman %E Vladimir Vovk %E Zhiyuan Luo %E Evgueni Smirnov %E Giovanni Cherubin %F pmlr-v128-vovk20a %I PMLR %P 84--99 %U https://proceedings.mlr.press/v128/vovk20a.html %V 128 %X Most existing examples of full conformal predictive systems, split conformal predictive systems, and cross-conformal predictive systems impose severe restrictions on the adaptation of predictive distributions to the test object at hand. In this paper we develop split conformal predictive systems that are fully adaptive. Our method consists in calibrating existing predictive systems; the input predictive system is not supposed to satisfy any properties of validity, whereas the output predictive system is guaranteed to be calibrated in probability.
APA
Vovk, V., Petej, I., Toccaceli, P., Gammerman, A., Ahlberg, E. & Carlsson, L.. (2020). Conformal calibrators. Proceedings of the Ninth Symposium on Conformal and Probabilistic Prediction and Applications, in Proceedings of Machine Learning Research 128:84-99 Available from https://proceedings.mlr.press/v128/vovk20a.html.

Related Material