SpeedCP: Fast Kernel-based Conditional Conformal Prediction

Yating Liu, Yeo Jin Jung, Zixuan Wu, Sowon Jeong, Claire Donnat
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:76980-77027, 2026.

Abstract

Conformal prediction provides distribution-free prediction sets with finite-sample conditional guarantees. RKHS-based frameworks—while promising for complex covariate shifts—suffer from prohibitive computational costs. To guarantee conditional validity under such shifts while ensuring feasibility, we build upon the framework of Gibbs et al. (2025) by introducing a stable and efficient algorithm that computes the full solution path of the regularized RKHS conformal optimization problem, at essentially the same cost as a single kernel quantile fit. Our approach provides simultaneous hyperparameter tuning which provides smoothness control and data-adaptive calibration. To extend the method to high-dimensional settings, we further integrate our approach with low-rank latent embeddings that capture conditional validity in a data-driven latent space. Empirically, our method provides reliable conditional coverage across a variety of modern black-box predictors, improving the interval length of Gibbs et al. (2025) by 30%, while achieving a 40-fold speedup.

Cite this Paper


BibTeX
@InProceedings{pmlr-v306-liu26bq, title = {{S}peed{CP}: Fast Kernel-based Conditional Conformal Prediction}, author = {Liu, Yating and Jung, Yeo Jin and Wu, Zixuan and Jeong, Sowon and Donnat, Claire}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {76980--77027}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/liu26bq/liu26bq.pdf}, url = {https://proceedings.mlr.press/v306/liu26bq.html}, abstract = {Conformal prediction provides distribution-free prediction sets with finite-sample conditional guarantees. RKHS-based frameworks—while promising for complex covariate shifts—suffer from prohibitive computational costs. To guarantee conditional validity under such shifts while ensuring feasibility, we build upon the framework of Gibbs et al. (2025) by introducing a stable and efficient algorithm that computes the full solution path of the regularized RKHS conformal optimization problem, at essentially the same cost as a single kernel quantile fit. Our approach provides simultaneous hyperparameter tuning which provides smoothness control and data-adaptive calibration. To extend the method to high-dimensional settings, we further integrate our approach with low-rank latent embeddings that capture conditional validity in a data-driven latent space. Empirically, our method provides reliable conditional coverage across a variety of modern black-box predictors, improving the interval length of Gibbs et al. (2025) by 30%, while achieving a 40-fold speedup.} }
Endnote
%0 Conference Paper %T SpeedCP: Fast Kernel-based Conditional Conformal Prediction %A Yating Liu %A Yeo Jin Jung %A Zixuan Wu %A Sowon Jeong %A Claire Donnat %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-liu26bq %I PMLR %P 76980--77027 %U https://proceedings.mlr.press/v306/liu26bq.html %V 306 %X Conformal prediction provides distribution-free prediction sets with finite-sample conditional guarantees. RKHS-based frameworks—while promising for complex covariate shifts—suffer from prohibitive computational costs. To guarantee conditional validity under such shifts while ensuring feasibility, we build upon the framework of Gibbs et al. (2025) by introducing a stable and efficient algorithm that computes the full solution path of the regularized RKHS conformal optimization problem, at essentially the same cost as a single kernel quantile fit. Our approach provides simultaneous hyperparameter tuning which provides smoothness control and data-adaptive calibration. To extend the method to high-dimensional settings, we further integrate our approach with low-rank latent embeddings that capture conditional validity in a data-driven latent space. Empirically, our method provides reliable conditional coverage across a variety of modern black-box predictors, improving the interval length of Gibbs et al. (2025) by 30%, while achieving a 40-fold speedup.
APA
Liu, Y., Jung, Y.J., Wu, Z., Jeong, S. & Donnat, C.. (2026). SpeedCP: Fast Kernel-based Conditional Conformal Prediction. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:76980-77027 Available from https://proceedings.mlr.press/v306/liu26bq.html.

Related Material