Hybrid Energy Based Model in the Feature Space for Out-of-Distribution Detection

Marc Lafon, Elias Ramzi, Clément Rambour, Nicolas Thome
Proceedings of the 40th International Conference on Machine Learning, PMLR 202:18250-18268, 2023.

Abstract

Out-of-distribution (OOD) detection is a critical requirement for the deployment of deep neural networks. This paper introduces the HEAT model, a new post-hoc OOD detection method estimating the density of in-distribution (ID) samples using hybrid energy-based models (EBM) in the feature space of a pre-trained backbone. HEAT complements prior density estimators of the ID density, e.g. parametric models like the Gaussian Mixture Model (GMM), to provide an accurate yet robust density estimation. A second contribution is to leverage the EBM framework to provide a unified density estimation and to compose several energy terms. Extensive experiments demonstrate the significance of the two contributions. HEAT sets new state-of-the-art OOD detection results on the CIFAR-10 / CIFAR-100 benchmark as well as on the large-scale Imagenet benchmark. The code is available at: https://github.com/MarcLafon/heatood.

Cite this Paper

BibTeX
@InProceedings{pmlr-v202-lafon23a, title = {Hybrid Energy Based Model in the Feature Space for Out-of-Distribution Detection}, author = {Lafon, Marc and Ramzi, Elias and Rambour, Cl\'{e}ment and Thome, Nicolas}, booktitle = {Proceedings of the 40th International Conference on Machine Learning}, pages = {18250--18268}, year = {2023}, editor = {Krause, Andreas and Brunskill, Emma and Cho, Kyunghyun and Engelhardt, Barbara and Sabato, Sivan and Scarlett, Jonathan}, volume = {202}, series = {Proceedings of Machine Learning Research}, month = {23--29 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v202/lafon23a/lafon23a.pdf}, url = {https://proceedings.mlr.press/v202/lafon23a.html}, abstract = {Out-of-distribution (OOD) detection is a critical requirement for the deployment of deep neural networks. This paper introduces the HEAT model, a new post-hoc OOD detection method estimating the density of in-distribution (ID) samples using hybrid energy-based models (EBM) in the feature space of a pre-trained backbone. HEAT complements prior density estimators of the ID density, e.g. parametric models like the Gaussian Mixture Model (GMM), to provide an accurate yet robust density estimation. A second contribution is to leverage the EBM framework to provide a unified density estimation and to compose several energy terms. Extensive experiments demonstrate the significance of the two contributions. HEAT sets new state-of-the-art OOD detection results on the CIFAR-10 / CIFAR-100 benchmark as well as on the large-scale Imagenet benchmark. The code is available at: https://github.com/MarcLafon/heatood.} }
Endnote
%0 Conference Paper %T Hybrid Energy Based Model in the Feature Space for Out-of-Distribution Detection %A Marc Lafon %A Elias Ramzi %A Clément Rambour %A Nicolas Thome %B Proceedings of the 40th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2023 %E Andreas Krause %E Emma Brunskill %E Kyunghyun Cho %E Barbara Engelhardt %E Sivan Sabato %E Jonathan Scarlett %F pmlr-v202-lafon23a %I PMLR %P 18250--18268 %U https://proceedings.mlr.press/v202/lafon23a.html %V 202 %X Out-of-distribution (OOD) detection is a critical requirement for the deployment of deep neural networks. This paper introduces the HEAT model, a new post-hoc OOD detection method estimating the density of in-distribution (ID) samples using hybrid energy-based models (EBM) in the feature space of a pre-trained backbone. HEAT complements prior density estimators of the ID density, e.g. parametric models like the Gaussian Mixture Model (GMM), to provide an accurate yet robust density estimation. A second contribution is to leverage the EBM framework to provide a unified density estimation and to compose several energy terms. Extensive experiments demonstrate the significance of the two contributions. HEAT sets new state-of-the-art OOD detection results on the CIFAR-10 / CIFAR-100 benchmark as well as on the large-scale Imagenet benchmark. The code is available at: https://github.com/MarcLafon/heatood.
APA
Lafon, M., Ramzi, E., Rambour, C. & Thome, N.. (2023). Hybrid Energy Based Model in the Feature Space for Out-of-Distribution Detection. Proceedings of the 40th International Conference on Machine Learning, in Proceedings of Machine Learning Research 202:18250-18268 Available from https://proceedings.mlr.press/v202/lafon23a.html.

Related Material