DCdetector: Dual Attention Contrastive Representation Learning for Time Series Anomaly Detection

DCdetector: контрастивное обучение представлений временных рядов с двойным вниманием для обнаружения аномалий
Qingsong Wen, Tian Zhou, Chaoli Zhang, Liang Sun, Yiyuan Yang, Liang Sun
2023-08-04

contrastive representation learningdual attentionmulti-scale learningpermutation invariant representationtime series anomaly detection
Time series anomaly detection is critical for a wide range of applications. It aims to identify deviant samples from the normal sample distribution in time series. The most fundamental challenge for this task is to learn a representation map that enables effective discrimination of anomalies. Reconstruction-based methods still dominate, but the representation learning with anomalies might hurt the performance with its large abnormal loss. On the other hand, contrastive learning aims to find a representation that can clearly distinguish any instance from the others, which can bring a more natural and promising representation for time series anomaly detection. In this paper, we propose DCdetector, a multi-scale dual attention contrastive representation learning model. DCdetector utilizes a novel dual attention asymmetric design to create the permutated environment and pure contrastive loss to guide the learning process, thus learning a permutation invariant representation with superior discrimination abilities. Extensive experiments show that DCdetector achieves state-of-the-art results on multiple time series anomaly detection benchmark datasets. Code is publicly available at https://github.com/DAMO-DI-ML/KDD2023-DCdetector.
1
A pure contrastive loss guides representation learning without relying on anomaly reconstruction losses that may distort representations.
2
DCdetector introduces a multi-scale dual-attention contrastive representation learning model for time series anomaly detection.
3
Extensive experiments report state-of-the-art performance across multiple time series anomaly detection benchmark datasets.
4
Its dual-attention asymmetric design creates permuted environments and enables learning of permutation-invariant representations.
5
The learned representations provide stronger anomaly discrimination than reconstruction-oriented approaches.

time series data and their normal/anomalous samples

discriminative, permutation-invariant representation learning for identifying anomalies in time series

Publication Details
Publication Date
2023-08-04
Journal
Publisher
ISSN
Cited by
453
Access Type
Author Information
Authors
Qingsong Wen
Tian Zhou
Chaoli Zhang
Liang Sun
Yiyuan Yang
Liang Sun
Explore further
Open the scid.ai AI chat with a ready-made request: it will find papers on a similar topic and help build a literature review.
Find similar papers in the chat →
Make a presentation
100%