CP$^2$: Leveraging Geometry for Conformal Prediction via Canonicalization

Putri A Van der Linden, Alexander Timans, Erik J Bekkers
Proceedings of the Forty-first Conference on Uncertainty in Artificial Intelligence, PMLR 286:2642-2658, 2025.

Abstract

We study the problem of *conformal prediction* (CP) under geometric data shifts, where data samples are susceptible to transformations such as rotations or flips. While CP endows prediction models with *post-hoc* uncertainty quantification and formal coverage guarantees, their practicality breaks under distribution shifts that deteriorate model performance. To address this issue, we propose integrating geometric information-such as geometric pose-into the conformal procedure to reinstate its guarantees and ensure robustness under geometric shifts. In particular, we explore recent advancements on pose *canonicalization* as a suitable information extractor for this purpose. Evaluating the combined approach across discrete and continuous shifts and against equivariant and augmentation-based baselines, we find that integrating geometric information with CP yields a principled way to address geometric shifts while maintaining broad applicability to black-box predictors.

Cite this Paper

BibTeX
@InProceedings{pmlr-v286-linden25a, title = {CP$^2$: Leveraging Geometry for Conformal Prediction via Canonicalization}, author = {Van der Linden, Putri A and Timans, Alexander and Bekkers, Erik J}, booktitle = {Proceedings of the Forty-first Conference on Uncertainty in Artificial Intelligence}, pages = {2642--2658}, year = {2025}, editor = {Chiappa, Silvia and Magliacane, Sara}, volume = {286}, series = {Proceedings of Machine Learning Research}, month = {21--25 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v286/main/assets/linden25a/linden25a.pdf}, url = {https://proceedings.mlr.press/v286/linden25a.html}, abstract = {We study the problem of *conformal prediction* (CP) under geometric data shifts, where data samples are susceptible to transformations such as rotations or flips. While CP endows prediction models with *post-hoc* uncertainty quantification and formal coverage guarantees, their practicality breaks under distribution shifts that deteriorate model performance. To address this issue, we propose integrating geometric information-such as geometric pose-into the conformal procedure to reinstate its guarantees and ensure robustness under geometric shifts. In particular, we explore recent advancements on pose *canonicalization* as a suitable information extractor for this purpose. Evaluating the combined approach across discrete and continuous shifts and against equivariant and augmentation-based baselines, we find that integrating geometric information with CP yields a principled way to address geometric shifts while maintaining broad applicability to black-box predictors.} }
Endnote
%0 Conference Paper %T CP$^2$: Leveraging Geometry for Conformal Prediction via Canonicalization %A Putri A Van der Linden %A Alexander Timans %A Erik J Bekkers %B Proceedings of the Forty-first Conference on Uncertainty in Artificial Intelligence %C Proceedings of Machine Learning Research %D 2025 %E Silvia Chiappa %E Sara Magliacane %F pmlr-v286-linden25a %I PMLR %P 2642--2658 %U https://proceedings.mlr.press/v286/linden25a.html %V 286 %X We study the problem of *conformal prediction* (CP) under geometric data shifts, where data samples are susceptible to transformations such as rotations or flips. While CP endows prediction models with *post-hoc* uncertainty quantification and formal coverage guarantees, their practicality breaks under distribution shifts that deteriorate model performance. To address this issue, we propose integrating geometric information-such as geometric pose-into the conformal procedure to reinstate its guarantees and ensure robustness under geometric shifts. In particular, we explore recent advancements on pose *canonicalization* as a suitable information extractor for this purpose. Evaluating the combined approach across discrete and continuous shifts and against equivariant and augmentation-based baselines, we find that integrating geometric information with CP yields a principled way to address geometric shifts while maintaining broad applicability to black-box predictors.
APA
Van der Linden, P.A., Timans, A. & Bekkers, E.J.. (2025). CP$^2$: Leveraging Geometry for Conformal Prediction via Canonicalization. Proceedings of the Forty-first Conference on Uncertainty in Artificial Intelligence, in Proceedings of Machine Learning Research 286:2642-2658 Available from https://proceedings.mlr.press/v286/linden25a.html.

Related Material