Skip to Main content Skip to Navigation
Conference papers

Beyond First-Order Uncertainty Estimation with Evidential Models for Open-World Recognition

Abstract : In this paper, we tackle the challenge of jointly quantifying in-distribution and out-of-distribution (OOD) uncertainties. We introduce KLoS, a KL-divergence measure defined on the classprobability simplex. By leveraging the secondorder uncertainty representation provided by evidential models, KLoS captures more than existing first-order uncertainty measures such as predictive entropy. We design an auxiliary neural network, KLoSNet, to learn a refined measure directly aligned with the evidential training objective. Experiments show that KLoSNet acts as a class-wise density estimator and outperforms current uncertainty measures in the realistic context where no OOD data is available during training. We also report comparisons in the presence of OOD training samples, which shed a new light on the impact of the vicinity of this data with OOD test data.
Complete list of metadata
Contributor : Charles Corbière Connect in order to contact the contributor
Submitted on : Friday, September 17, 2021 - 1:03:42 PM
Last modification on : Friday, August 5, 2022 - 2:54:01 PM
Long-term archiving on: : Saturday, December 18, 2021 - 6:43:44 PM


Files produced by the author(s)


  • HAL Id : hal-03347628, version 1


Charles Corbière, Marc Lafon, Nicolas Thome, Matthieu Cord, Patrick Pérez. Beyond First-Order Uncertainty Estimation with Evidential Models for Open-World Recognition. ICML 2021 Workshop on Uncertainty and Robustness in Deep Learning, Sep 2021, Virtual, Austria. ⟨hal-03347628⟩



Record views


Files downloads