On the Learnability of Out-of-distribution Detection

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: Fang, Zhen, Li, Yixuan, Liu, Feng, Han, Bo, Lu, Jie
Format: Preprint
Publié: 2024
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866916197349457920
author Fang, Zhen
Li, Yixuan
Liu, Feng
Han, Bo
Lu, Jie
author_facet Fang, Zhen
Li, Yixuan
Liu, Feng
Han, Bo
Lu, Jie
contents Supervised learning aims to train a classifier under the assumption that training and test data are from the same distribution. To ease the above assumption, researchers have studied a more realistic setting: out-of-distribution (OOD) detection, where test data may come from classes that are unknown during training (i.e., OOD data). Due to the unavailability and diversity of OOD data, good generalization ability is crucial for effective OOD detection algorithms, and corresponding learning theory is still an open problem. To study the generalization of OOD detection, this paper investigates the probably approximately correct (PAC) learning theory of OOD detection that fits the commonly used evaluation metrics in the literature. First, we find a necessary condition for the learnability of OOD detection. Then, using this condition, we prove several impossibility theorems for the learnability of OOD detection under some scenarios. Although the impossibility theorems are frustrating, we find that some conditions of these impossibility theorems may not hold in some practical scenarios. Based on this observation, we next give several necessary and sufficient conditions to characterize the learnability of OOD detection in some practical scenarios. Lastly, we offer theoretical support for representative OOD detection works based on our OOD theory.
format Preprint
id arxiv_https___arxiv_org_abs_2404_04865
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle On the Learnability of Out-of-distribution Detection
Fang, Zhen
Li, Yixuan
Liu, Feng
Han, Bo
Lu, Jie
Machine Learning
Computer Vision and Pattern Recognition
Supervised learning aims to train a classifier under the assumption that training and test data are from the same distribution. To ease the above assumption, researchers have studied a more realistic setting: out-of-distribution (OOD) detection, where test data may come from classes that are unknown during training (i.e., OOD data). Due to the unavailability and diversity of OOD data, good generalization ability is crucial for effective OOD detection algorithms, and corresponding learning theory is still an open problem. To study the generalization of OOD detection, this paper investigates the probably approximately correct (PAC) learning theory of OOD detection that fits the commonly used evaluation metrics in the literature. First, we find a necessary condition for the learnability of OOD detection. Then, using this condition, we prove several impossibility theorems for the learnability of OOD detection under some scenarios. Although the impossibility theorems are frustrating, we find that some conditions of these impossibility theorems may not hold in some practical scenarios. Based on this observation, we next give several necessary and sufficient conditions to characterize the learnability of OOD detection in some practical scenarios. Lastly, we offer theoretical support for representative OOD detection works based on our OOD theory.
title On the Learnability of Out-of-distribution Detection
topic Machine Learning
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2404.04865