Beyond Binary Out-of-Distribution Detection: Characterizing Distributional Shifts with Multi-Statistic Diffusion Trajectories

2025-10-20

AI Summary

Existing out-of-distribution (OOD) detection methods predominantly rely on single-scalar scores, limiting their ability to distinguish among distinct types of distribution shifts and hindering interpretability and downstream utility. To address this, we propose DISC—a novel, diffusion-based OOD detection and classification framework that leverages denoising trajectories. DISC models statistical discrepancies across multiple feature levels throughout the iterative denoising process, yielding high-dimensional trajectory representations that jointly enable OOD detection and fine-grained shift-type identification (e.g., rotation, additive noise, domain shift). By departing from the conventional scalar-scoring paradigm, DISC achieves state-of-the-art detection performance on both image and tabular benchmarks while significantly improving OOD type classification accuracy. Comprehensive experiments demonstrate DISC’s effectiveness, robustness, and generalizability across diverse datasets and shift scenarios.

Technology Category

• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Computer Vision: Diffusion Models for Vision
• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Web Mining and Content Analysis: Content-based information diffusion

Abstract

Detecting out-of-distribution (OOD) data is critical for machine learning, be it for safety reasons or to enable open-ended learning. However, beyond mere detection, choosing an appropriate course of action typically hinges on the type of OOD data encountered. Unfortunately, the latter is generally not distinguished in practice, as modern OOD detection methods collapse distributional shifts into single scalar outlier scores. This work argues that scalar-based methods are thus insufficient for OOD data to be properly contextualized and prospectively exploited, a limitation we overcome with the introduction of DISC: Diffusion-based Statistical Characterization. DISC leverages the iterative denoising process of diffusion models to extract a rich, multi-dimensional feature vector that captures statistical discrepancies across multiple noise levels. Extensive experiments on image and tabular benchmarks show that DISC matches or surpasses state-of-the-art detectors for OOD detection and, crucially, also classifies OOD type, a capability largely absent from prior work. As such, our work enables a shift from simple binary OOD detection to a more granular detection.

Problem

• Characterizing diverse distributional shifts beyond binary detection
• Overcoming scalar score limitations with multi-dimensional statistical features
• Enabling granular OOD type classification for contextualized decision-making

Innovation

• Leverages diffusion models for multi-dimensional feature extraction
• Characterizes distributional shifts across multiple noise levels
• Classifies OOD types beyond binary detection capability