Towards Real Unsupervised Anomaly Detection Via Confident Meta-Learning

2025-08-04

AI Summary

Existing unsupervised anomaly detection methods are, in practice, semi-supervised: they rely on manually curated anomaly-free training sets, introducing human bias and limiting generalization. This work proposes Confident Meta-learning (CoMet), the first framework enabling truly unsupervised end-to-end learning directly on raw data containing unknown anomalies. CoMet integrates Soft Confident Learning with Meta-Learning: it suppresses noisy samples via confidence-weighted loss and enhances robustness through meta-validation loss covariance regularization. The framework is model-agnostic, fully differentiable, and compatible with mainstream backbone architectures. Extensive experiments on MVTec-AD, VIADUCT, and KSDD2 benchmarks demonstrate significant improvements over state-of-the-art methods—achieving new performance records across multiple metrics. Crucially, CoMet exhibits strong robustness to varying proportions of anomalies in training data, confirming its practical viability in real-world scenarios where clean normal data cannot be guaranteed.

Technology Category

• Machine Learning: Adversarial Learning & Robustness
• Computer Vision: Adversarial Attacks & Robustness
• Data Mining & Knowledge Management: Anomaly/Outlier Detection

Application Category

• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Security and Privacy: Data transparency and provenance

Abstract

So-called unsupervised anomaly detection is better described as semi-supervised, as it assumes all training data are nominal. This assumption simplifies training but requires manual data curation, introducing bias and limiting adaptability. We propose Confident Meta-learning (CoMet), a novel training strategy that enables deep anomaly detection models to learn from uncurated datasets where nominal and anomalous samples coexist, eliminating the need for explicit filtering. Our approach integrates Soft Confident Learning, which assigns lower weights to low-confidence samples, and Meta-Learning, which stabilizes training by regularizing updates based on training validation loss covariance. This prevents overfitting and enhances robustness to noisy data. CoMet is model-agnostic and can be applied to any anomaly detection method trainable via gradient descent. Experiments on MVTec-AD, VIADUCT, and KSDD2 with two state-of-the-art models demonstrate the effectiveness of our approach, consistently improving over the baseline methods, remaining insensitive to anomalies in the training set, and setting a new state-of-the-art across all datasets.

Problem

• Enables anomaly detection with uncurated mixed nominal and anomalous data
• Uses confident meta-learning to reduce bias from manual data filtering
• Improves robustness and performance across diverse anomaly detection datasets

Innovation

• Confident Meta-learning for uncurated datasets
• Soft Confident Learning weights low-confidence samples
• Meta-Learning stabilizes training via validation loss