MADOD: Generalizing OOD Detection to Unseen Domains via G-Invariance Meta-Learning

TL;DR

MADOD is a meta-learning framework that improves out-of-distribution detection across unseen domains by leveraging G-invariance and energy-based regularization, effectively handling both covariate and semantic shifts without test-time adaptation.

Contribution

It introduces a novel meta-learning approach with task construction and energy regularization to enhance domain-invariant features and OOD detection in unseen domains.

Findings

Achieves up to 20.81% improvement in AUPR for OOD detection.

Maintains competitive in-distribution classification accuracy.

Operates effectively without test-time adaptation.

Abstract

Real-world machine learning applications often face simultaneous covariate and semantic shifts, challenging traditional domain generalization and out-of-distribution (OOD) detection methods. We introduce Meta-learned Across Domain Out-of-distribution Detection (MADOD), a novel framework designed to address both shifts concurrently. MADOD leverages meta-learning and G-invariance to enhance model generalizability and OOD detection in unseen domains. Our key innovation lies in task construction: we randomly designate in-distribution classes as pseudo-OODs within each meta-learning task, simulating OOD scenarios using existing data. This approach, combined with energy-based regularization, enables the learning of robust, domain-invariant features while calibrating decision boundaries for effective OOD detection. Operating in a test domain-agnostic setting, MADOD eliminates the need for adaptation during inference, making it suitable for scenarios where test data is unavailable. Extensive experiments on real-world and synthetic datasets demonstrate MADOD's superior performance in semantic OOD detection across unseen domains, achieving an AUPR improvement of 8.48% to 20.81%, while maintaining competitive in-distribution classification accuracy, representing a significant advancement in handling both covariate and semantic shifts.