CADIC: Continual Anomaly Detection Based on Incremental Coreset

2025-11-10

AI Summary

Continual anomaly detection (CAD) faces dual challenges: learning normal patterns for new tasks under dynamic data distributions while mitigating catastrophic forgetting. Existing embedding-based approaches rely on task-specific sub-memory banks, leading to fragmented knowledge and poor scalability. This paper proposes a unified memory bank mechanism, employing a shared, fixed-size incremental coreset to enable cross-task knowledge accumulation and efficient memory updates; anomaly scores are computed via embedding learning combined with nearest-neighbor matching. On MVTec AD and Visa datasets, the method achieves image-level AUROC scores of 0.972 and 0.891, respectively, and attains 100% anomaly detection rate on the E-Paper dataset—substantially outperforming state-of-the-art methods. The core contribution lies in decoupling task boundaries, thereby enhancing model flexibility and scalability without requiring explicit task identifiers or architectural expansion.

Technology Category

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

Application Category

• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

The primary objective of Continual Anomaly Detection (CAD) is to learn the normal patterns of new tasks under dynamic data distribution assumptions while mitigating catastrophic forgetting. Existing embedding-based CAD approaches continuously update a memory bank with new embeddings to adapt to sequential tasks. However, these methods require constructing class-specific sub-memory banks for each task, which restricts their flexibility and scalability. To address this limitation, we propose a novel CAD framework where all tasks share a unified memory bank. During training, the method incrementally updates embeddings within a fixed-size coreset, enabling continuous knowledge acquisition from sequential tasks without task-specific memory fragmentation. In the inference phase, anomaly scores are computed via a nearest-neighbor matching mechanism, achieving state-of-the-art detection accuracy. We validate the method through comprehensive experiments on MVTec AD and Visa datasets. Results show that our approach outperforms existing baselines, achieving average image-level AUROC scores of 0.972 (MVTec AD) and 0.891 (Visa). Notably, on a real-world electronic paper dataset, it demonstrates 100% accuracy in anomaly sample detection, confirming its robustness in practical scenarios. The implementation will be open-sourced on GitHub.

Problem

• Addresses catastrophic forgetting in continual anomaly detection tasks
• Eliminates task-specific memory banks to improve flexibility and scalability
• Enables continuous knowledge acquisition through incremental coreset updates

Innovation

• Unified memory bank shared across all tasks
• Incremental coreset updates for continuous learning
• Nearest-neighbor matching for anomaly score computation