An Encode-then-Decompose Approach to Unsupervised Time Series Anomaly Detection on Contaminated Training Data--Extended Version

2025-10-21

AI Summary

Time-series anomaly detection suffers from model degradation under unsupervised training with anomalous contamination. To address this, we propose a novel “encoding-then-decomposition” paradigm: within the autoencoder’s latent space, we disentangle stable representations—robust to anomalies—from auxiliary representations—capturing local dynamics—and introduce a mutual-information-based anomaly scoring mechanism, replacing contamination-sensitive reconstruction error. Our method jointly integrates representation decomposition, mutual information estimation, and contrastive learning to support multivariate modeling. Evaluated on eight standard benchmarks across varying contamination ratios, our approach achieves state-of-the-art or competitive performance, demonstrating substantial improvements in robustness and generalization.

Technology Category

• Machine Learning: Deep Generative Models & Autoencoders
• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Computer Vision: Representation Learning for Vision

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Time series anomaly detection is important in modern large-scale systems and is applied in a variety of domains to analyze and monitor the operation of diverse systems. Unsupervised approaches have received widespread interest, as they do not require anomaly labels during training, thus avoiding potentially high costs and having wider applications. Among these, autoencoders have received extensive attention. They use reconstruction errors from compressed representations to define anomaly scores. However, representations learned by autoencoders are sensitive to anomalies in training time series, causing reduced accuracy. We propose a novel encode-then-decompose paradigm, where we decompose the encoded representation into stable and auxiliary representations, thereby enhancing the robustness when training with contaminated time series. In addition, we propose a novel mutual information based metric to replace the reconstruction errors for identifying anomalies. Our proposal demonstrates competitive or state-of-the-art performance on eight commonly used multi- and univariate time series benchmarks and exhibits robustness to time series with different contamination ratios.

Problem

• Detecting anomalies in contaminated time series training data
• Improving robustness of autoencoder representations against anomalies
• Replacing reconstruction errors with mutual information metrics

Innovation

• Encode-then-decompose paradigm for robust representation
• Decomposing representations into stable and auxiliary components
• Mutual information metric replacing reconstruction error detection