Beyond Mahalanobis Distance for Textual OOD Detection
Pierre Colombo, Eduardo Dadalto Câmara Gomes, Guillaume Staerman, Nathan Noiry, Pablo Piantanida
Abstract
As the number of AI systems keeps growing, it is fundamental to implement and develop efficient control mechanisms to ensure the safe and proper functioning of machine learning (ML) systems. Reliable out-of-distribution (OOD) detection aims to detect test samples that are statistically far from the training distribution, as they might cause failures of in-production systems. In this paper, we propose a new detector called TRUSTED. Different from previous works, TRUSTED key components (i) include a novel OOD score relying on the concept of statistical data depth, (ii) rely on the idea’s full potential that all hidden layers of the network carry information regarding OOD. Our extensive experiments, comparing over 51k model configurations including different checkpoints, seed and various datasets, demonstrate that TRUSTED achieve state-of-the-art performances by producing an improvement of over 3 AUROC points.
BibTeX
@inproceedings{
colombo2022beyond,
title={Beyond Mahalanobis Distance for Textual {OOD} Detection},
author={Pierre Colombo and Eduardo Dadalto C{\^a}mara Gomes and Guillaume Staerman and Nathan Noiry and Pablo Piantanida},
booktitle={Advances in Neural Information Processing Systems},
editor={Alice H. Oh and Alekh Agarwal and Danielle Belgrave and Kyunghyun Cho},
year={2022},
url={https://openreview.net/forum?id=ReB7CCByD6U}
}