LARA: A Light and Anti-overfitting Retraining Approach for Unsupervised Time Series Anomaly Detection
Feiyi Chen, Zhen Qin, Mengchu Zhou, Yingying Zhang, Shuiguang Deng, Lunting Fan, Guansong Pang, Qingsong Wen
Abstract
Most of current anomaly detection models assume that the normal pattern remains the same all the time. However, the normal patterns of web services can change dramatically and frequently over time. The model trained on old-distribution data becomes outdated and ineffective after such changes. Retraining the whole model whenever the pattern is changed is computationally expensive. Further, at the beginning of normal pattern changes, there is not enough observation data from the new distribution. Retraining a large neural network model with limited data is vulnerable to overfitting. Thus, we propose a Light Anti-overfitting Retraining Approach (LARA) based on deep variational auto-encoders for time series anomaly detection. In LARA we make the following three major contributions: 1) the retraining process is designed as a convex problem such that overfitting is prevented and the retraining process can converge fast; 2) a novel ruminate block is introduced, which can leverage the historical data without the need to store them; 3) we mathematically and experimentally prove that when fine-tuning the latent vector and reconstructed data, the linear formations can achieve the least adjusting errors between the ground truths and the fine-tuned ones. Moreover, we have performed many experiments to verify that retraining LARA with even a limited amount of data from new distribution can achieve competitive performance in comparison with the state-of-the-art anomaly detection models trained with sufficient data. Besides, we verify its light computational overhead.
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
4 top-tier papers have cited it since. Know what they found before your reviewers do.
Lune can draw on
the full paper · the top-tier papers that cite it · related work
Free to start. No credit card required.
Lineage
- Learning Multi-Pattern Normalities in the Frequency Domain for Efficient Time Series Anomaly Detection cites DCdetector: Dual Attention Contrastive Representation Learning for Time Series Anomaly Detection
- Learning Multi-Pattern Normalities in the Frequency Domain for Efficient Time Series Anomaly Detection cites TranAD: Deep Transformer Networks for Anomaly Detection in Multivariate Time Series Data
- Learning Multi-Pattern Normalities in the Frequency Domain for Efficient Time Series Anomaly Detection cites Timeseries Anomaly Detection using Temporal Hierarchical One-Class Network
- Learning Multi-Pattern Normalities in the Frequency Domain for Efficient Time Series Anomaly Detection cites Anomaly Transformer: Time Series Anomaly Detection with Association Discrepancy
- Cluster-Wide Task Slowdown Detection in Cloud System cites DCdetector: Dual Attention Contrastive Representation Learning for Time Series Anomaly Detection
- Cluster-Wide Task Slowdown Detection in Cloud System cites TranAD: Deep Transformer Networks for Anomaly Detection in Multivariate Time Series Data
- Cluster-Wide Task Slowdown Detection in Cloud System cites Timeseries Anomaly Detection using Temporal Hierarchical One-Class Network
- Cluster-Wide Task Slowdown Detection in Cloud System cites Anomaly Transformer: Time Series Anomaly Detection with Association Discrepancy
- DCdetector: Dual Attention Contrastive Representation Learning for Time Series Anomaly Detection cites Timeseries Anomaly Detection using Temporal Hierarchical One-Class Network
- DCdetector: Dual Attention Contrastive Representation Learning for Time Series Anomaly Detection cites Anomaly Transformer: Time Series Anomaly Detection with Association Discrepancy
- Noise-Resilient Point-wise Anomaly Detection in Time Series Using Weak Segment Labels cites DCdetector: Dual Attention Contrastive Representation Learning for Time Series Anomaly Detection
- Noise-Resilient Point-wise Anomaly Detection in Time Series Using Weak Segment Labels cites Timeseries Anomaly Detection using Temporal Hierarchical One-Class Network
- Noise-Resilient Point-wise Anomaly Detection in Time Series Using Weak Segment Labels cites Anomaly Transformer: Time Series Anomaly Detection with Association Discrepancy
- Anomaly Transformer: Time Series Anomaly Detection with Association Discrepancy cites Timeseries Anomaly Detection using Temporal Hierarchical One-Class Network
Cited by top-tier papers4
- Learning Multi-Pattern Normalities in the Frequency Domain for Efficient Time Series Anomaly DetectionFeiyi Chen, Yingying Zhang, Zhen Qin, Lunting Fan et al.ICDE 2024 · 10 citations
- Cluster-Wide Task Slowdown Detection in Cloud SystemFeiyi Chen, Yingying Zhang, Lunting Fan, Yuxuan Liang et al.KDD 2024 · 2 citations
- United We Stand: Towards End-to-End Log-based Fault Diagnosis via Interactive Multi-Task LearningMinghua He, Chiming Duan, Pei Xiao, Tong Jia et al.ASE 2025 · 1 citation
- Noise-Resilient Point-wise Anomaly Detection in Time Series Using Weak Segment LabelsYaxuan Wang, Hao Cheng, Jing Xiong, Qingsong Wen et al.KDD 2025
Builds on11
- Anomaly Transformer: Time Series Anomaly Detection with Association DiscrepancyJiehui Xu, Haixu Wu, Jianmin Wang, Mingsheng LongICLR 2022 · 960 citations
- TranAD: Deep Transformer Networks for Anomaly Detection in Multivariate Time Series DataShreshth Tuli, Giuliano Casale, Nicholas R. JenningsVLDB 2022 · 930 citations
- Timeseries Anomaly Detection using Temporal Hierarchical One-Class NetworkLifeng Shen, Zhuocong Li, James T. KwokNeurIPS 2020 · 454 citations
- DCdetector: Dual Attention Contrastive Representation Learning for Time Series Anomaly DetectionYiyuan Yang, Chaoli Zhang, Tian Zhou, Qingsong Wen et al.KDD 2023 · 244 citations
- Learning Unsupervised Metaformer for Anomaly DetectionJhih-Ciang Wu, Ding-Jie Chen, Chiou-Shann Fuh, Tyng-Luh LiuICCV 2021 · 101 citations
Related papers
- Revisiting VAE for Unsupervised Time Series Anomaly Detection: A Frequency PerspectiveZexin Wang, Changhua Pei, Minghua Ma, Xin Wang et al.WWW 2024 · 90 citations
- Anomaly Detection in Time Series with Robust Variational Quasi-Recurrent AutoencodersTung Kieu, Bin Yang, Chenjuan Guo, Razvan-Gabriel Cirstea et al.ICDE 2022 · 60 citations
- A Semi-Supervised VAE Based Active Anomaly Detection Framework in Multivariate Time Series for Online SystemsTao Huang, Pengfei Chen, Ruipeng LiWWW 2022 · 74 citations
- Unsupervised Anomaly Detection on Microservice Traces through Graph VAEZhe Xie, Haowen Xu, Wenxiao Chen, Wanxue Li et al.WWW 2023 · 44 citations
- KAN-AD: Time Series Anomaly Detection with Kolmogorov-Arnold NetworksQuan Zhou, Changhua Pei, Fei Sun, Jing Han et al.ICML 2025
