Abstract:The presence of noisy examples in the training set inevitably hampers the performance of out-of-distribution (OOD) detection. In this paper, we investigate a previously overlooked problem called OOD detection under asymmetric open-set noise, which is frequently encountered and significantly reduces the identifiability of OOD examples. We analyze the generating process of asymmetric open-set noise and observe the influential role of the confounding variable, entangling many open-set noisy examples with partial in-distribution (ID) examples referred to as hard-ID examples due to spurious-related characteristics. To address the issue of the confounding variable, we propose a novel method called Adversarial Confounder REmoving (ACRE) that utilizes progressive optimization with adversarial learning to curate three collections of potential examples (easy-ID, hard-ID, and open-set noisy) while simultaneously developing invariant representations and reducing spurious-related representations. Specifically, by obtaining easy-ID examples with minimal confounding effect, we learn invariant representations from ID examples that aid in identifying hard-ID and open-set noisy examples based on their similarity to the easy-ID set. By triplet adversarial learning, we achieve the joint minimization and maximization of distribution discrepancies across the three collections, enabling the dual elimination of the confounding variable. We also leverage potential open-set noisy examples to optimize a K+1-class classifier, further removing the confounding variable and inducing a tailored K+1-Guided scoring function. Theoretical analysis establishes the feasibility of ACRE, and extensive experiments demonstrate its effectiveness and generalization. Code is available at https://github.com/Anonymous-re-ssl/ACRE0 .

Diffusion based Semantic Outlier Generation via Nuisance Awareness for Out-of-Distribution Detection

Out-of-Distribution Detection with Virtual Outlier Smoothing

Non-Linear Outlier Synthesis for Out-of-Distribution Detection

Non-Parametric Outlier Synthesis

Visual Out-of-Distribution Detection in Open-Set Noisy Environments

OAL: Enhancing OOD Detection Using Latent Diffusion

Out-of-distribution Detection with Implicit Outlier Transformation

Diffusion-based Layer-wise Semantic Reconstruction for Unsupervised Out-of-Distribution Detection

RONF: Reliable Outlier Synthesis under Noisy Feature Space for Out-of-Distribution Detection

Feature Space Singularity for Out-of-Distribution Detection.

Semantic enhanced for out-of-distribution detection

Feature Space Singularity for Out-of-Distribution Detection

SUOD: A Scalable Unsupervised Outlier Detection Framework.

Going Beyond Conventional OOD Detection

Robustness to Spurious Correlations Improves Semantic Out-of-Distribution Detection

Out-of-distribution Detection with Diffusion-based Neighborhood

SEE-OoD: Supervised Exploration For Enhanced Out-of-Distribution Detection

DOS: Diverse Outlier Sampling for Out-of-Distribution Detection

Three Factors to Improve Out-of-Distribution Detection

Out-of-Distribution Detection with Hilbert-Schmidt Independence Optimization

Out-of-distribution Detection Learning with Unreliable Out-of-distribution Sources