TULiP: Test-time Uncertainty Estimation via Linearization and Weight Perturbation

Yuhui Zhang, Dongshen Wu, Yuichiro Wada, Takafumi Kanamori

Keywords: Out-of-distribution detection, Uncertainty Quantification, Lazy Training, Neural Tangent Kernel

TL;DR: We propose TULiP, a theoretically-driven uncertainty estimator for OOD detection applicable for pre-trained models.

Abstract: A reliable uncertainty estimation method is the foundation of many modern out-of-distribution (OOD) detectors, which are critical for safe deployments of deep learning models in the open world. In this work, we propose TULiP, a novel, theoretically-driven, post-hoc uncertainty estimator for OOD detection. Our method considers a hypothetical perturbation applied to the network prior to convergence. Based on linearized training dynamics, we bound the effect of such perturbation, resulting in an uncertainty score computable by perturbing model parameters. Ultimately, our approach computes uncertainty from a set of sampled predictions, thus not limited to classification problems. We visualize our bound on synthetic regression and classification datasets. Furthermore, we demonstrate the effectiveness of TULiP using large-scale OOD detection benchmarks for image classification. Our method exhibits state-of-the-art performance, particularly for near-distribution samples.

Primary Area: probabilistic methods (Bayesian methods, variational inference, sampling, UQ, etc.)

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 11751