Distribution Prototype Diffusion Learning for Open-set Supervised Anomaly Detection

2025-02-28

AI Summary

In open-set supervised anomaly detection (OSAD), existing pseudo-anomaly generation methods neglect the prior distribution of normal samples and yield ambiguous decision boundaries. To address this, we propose a diffusion-based modeling framework grounded in a learnable Gaussian prototype space. First, we construct learnable Gaussian prototypes that explicitly characterize the normal data distribution. Second, we integrate a Schrödinger bridge mechanism to enable directional diffusion—drawing normal samples toward prototypes while actively repelling anomalies. Third, we impose a feature dispersion constraint on a hyperspherical latent space to explicitly enforce compact yet well-separated normal-class representations. This work is the first to unify distributional prototype learning, Schrödinger bridge diffusion, and hyperspherical representation learning within a single OSAD framework. Evaluated on nine standard benchmarks, our method achieves state-of-the-art performance, significantly improving intra-class compactness and inter-class separability.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Machine Learning: Learning with Manifolds

Application Category

• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

In Open-set Supervised Anomaly Detection (OSAD), the existing methods typically generate pseudo anomalies to compensate for the scarcity of observed anomaly samples, while overlooking critical priors of normal samples, leading to less effective discriminative boundaries. To address this issue, we propose a Distribution Prototype Diffusion Learning (DPDL) method aimed at enclosing normal samples within a compact and discriminative distribution space. Specifically, we construct multiple learnable Gaussian prototypes to create a latent representation space for abundant and diverse normal samples and learn a Schr""odinger bridge to facilitate a diffusive transition toward these prototypes for normal samples while steering anomaly samples away. Moreover, to enhance inter-sample separation, we design a dispersion feature learning way in hyperspherical space, which benefits the identification of out-of-distribution anomalies. Experimental results demonstrate the effectiveness and superiority of our proposed DPDL, achieving state-of-the-art performance on 9 public datasets.

Problem

• Addresses scarcity of anomaly samples in Open-set Supervised Anomaly Detection.
• Proposes Distribution Prototype Diffusion Learning for compact normal sample representation.
• Enhances anomaly detection by separating normal and anomaly samples effectively.

Innovation

• Gaussian prototypes for normal sample representation
• Schrödinger bridge for diffusive transition guidance
• Hyperspherical dispersion for anomaly separation