Exact and Robust Conformal Inference Methods for Predictive Machine Learning with Dependent Data

Victor Chernozhukov, Kaspar Wüthrich, Zhu Yinchu
; Proceedings of the 31st Conference On Learning Theory, PMLR 75:732-749, 2018.

Abstract

We extend conformal inference to general settings that allow for time series data. Our proposal is developed as a randomization method and accounts for potential serial dependence by including block structures in the permutation scheme. As a result, the proposed method retains the exact, model-free validity when the data are i.i.d. or more generally exchangeable, similar to usual conformal inference methods. When exchangeability fails, as is the case for common time series data, the proposed approach is approximately valid under weak assumptions on the conformity score.

Cite this Paper


BibTeX
@InProceedings{pmlr-v75-chernozhukov18a, title = {Exact and Robust Conformal Inference Methods for Predictive Machine Learning with Dependent Data}, author = {Chernozhukov, Victor and W\"{u}thrich, Kaspar and Yinchu, Zhu}, booktitle = {Proceedings of the 31st Conference On Learning Theory}, pages = {732--749}, year = {2018}, editor = {Sébastien Bubeck and Vianney Perchet and Philippe Rigollet}, volume = {75}, series = {Proceedings of Machine Learning Research}, address = {}, month = {06--09 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v75/chernozhukov18a/chernozhukov18a.pdf}, url = {http://proceedings.mlr.press/v75/chernozhukov18a.html}, abstract = {We extend conformal inference to general settings that allow for time series data. Our proposal is developed as a randomization method and accounts for potential serial dependence by including block structures in the permutation scheme. As a result, the proposed method retains the exact, model-free validity when the data are i.i.d. or more generally exchangeable, similar to usual conformal inference methods. When exchangeability fails, as is the case for common time series data, the proposed approach is approximately valid under weak assumptions on the conformity score.} }
Endnote
%0 Conference Paper %T Exact and Robust Conformal Inference Methods for Predictive Machine Learning with Dependent Data %A Victor Chernozhukov %A Kaspar Wüthrich %A Zhu Yinchu %B Proceedings of the 31st Conference On Learning Theory %C Proceedings of Machine Learning Research %D 2018 %E Sébastien Bubeck %E Vianney Perchet %E Philippe Rigollet %F pmlr-v75-chernozhukov18a %I PMLR %J Proceedings of Machine Learning Research %P 732--749 %U http://proceedings.mlr.press %V 75 %W PMLR %X We extend conformal inference to general settings that allow for time series data. Our proposal is developed as a randomization method and accounts for potential serial dependence by including block structures in the permutation scheme. As a result, the proposed method retains the exact, model-free validity when the data are i.i.d. or more generally exchangeable, similar to usual conformal inference methods. When exchangeability fails, as is the case for common time series data, the proposed approach is approximately valid under weak assumptions on the conformity score.
APA
Chernozhukov, V., Wüthrich, K. & Yinchu, Z.. (2018). Exact and Robust Conformal Inference Methods for Predictive Machine Learning with Dependent Data. Proceedings of the 31st Conference On Learning Theory, in PMLR 75:732-749

Related Material