COMET: Codebook-based Online-adaptive Multi-scale Embedding for Time-series Anomaly Detection

2026-02-02

AI Summary

Existing methods for time series anomaly detection struggle to simultaneously model multi-scale temporal dependencies and multivariate correlations, and are often sensitive to distribution shifts during inference. To address these limitations, this work proposes a framework based on multi-scale patch encoding and vector-quantized codebooks, featuring a core-set dual-scoring mechanism for fine-grained anomaly discrimination. Furthermore, the approach introduces an online codebook adaptation strategy grounded in contrastive learning, coupled with dynamic pseudo-label updates to enhance model generalization under distributional shifts. Evaluated across five benchmark datasets, the method achieves state-of-the-art performance in 36 out of 45 metrics, significantly outperforming existing approaches.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Computer Vision: Multi-modal Vision

Application Category

• Graph Algorithms and Modeling for the Web: Algorithms and analysis for heterogeneous, signed, attributed, multi-relational, temporal, higher-order, and annotated Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

Time series anomaly detection is a critical task across various industrial domains. However, capturing temporal dependencies and multivariate correlations within patch-level representation learning remains underexplored, and reliance on single-scale patterns limits the detection of anomalies across different temporal ranges. Furthermore, focusing on normal data representations makes models vulnerable to distribution shifts at inference time. To address these limitations, we propose Codebook-based Online-adaptive Multi-scale Embedding for Time-series anomaly detection (COMET), which consists of three key components: (1) Multi-scale Patch Encoding captures temporal dependencies and inter-variable correlations across multiple patch scales. (2) Vector-Quantized Coreset learns representative normal patterns via codebook and detects anomalies with a dual-score combining quantization error and memory distance. (3) Online Codebook Adaptation generates pseudo-labels based on codebook entries and dynamically adapts the model at inference through contrastive learning. Experiments on five benchmark datasets demonstrate that COMET achieves the best performance in 36 out of 45 evaluation metrics, validating its effectiveness across diverse environments.

Problem

• time-series anomaly detection
• temporal dependencies
• multivariate correlations
• multi-scale patterns
• distribution shift

Innovation

• multi-scale patch encoding
• vector-quantized coreset
• online codebook adaptation
• time-series anomaly detection
• contrastive learning