Deep Generative model with Hierarchical Latent Factors for Time Series Anomaly Detection

Cristian I. Challu, Peihong Jiang, Ying Nian Wu, Laurent Callot
Proceedings of The 25th International Conference on Artificial Intelligence and Statistics, PMLR 151:1643-1654, 2022.

Abstract

Multivariate time series anomaly detection has become an active area of research in recent years, with Deep Learning models outperforming previous approaches on benchmark datasets. Among reconstruction-based models, most previous work has focused on Variational Autoencoders and Generative Adversarial Networks. This work presents DGHL, a new family of generative models for time series anomaly detection, trained by maximizing the observed likelihood by posterior sampling and alternating back-propagation. A top-down Convolution Network maps a novel hierarchical latent space to time series windows, exploiting temporal dynamics to encode information efficiently. Despite relying on posterior sampling, it is computationally more efficient than current approaches, with up to 10x shorter training times than RNN based models. Our method outperformed current state-of-the-art models on four popular benchmark datasets. Finally, DGHL is robust to variable features between entities and accurate even with large proportions of missing values, settings with increasing relevance with the advent of IoT. We demonstrate the superior robustness of DGHL with novel occlusion experiments in this literature. Our code is available at https://github.com/cchallu/dghl.

Cite this Paper


BibTeX
@InProceedings{pmlr-v151-challu22a, title = { Deep Generative model with Hierarchical Latent Factors for Time Series Anomaly Detection }, author = {Challu, Cristian I. and Jiang, Peihong and Nian Wu, Ying and Callot, Laurent}, booktitle = {Proceedings of The 25th International Conference on Artificial Intelligence and Statistics}, pages = {1643--1654}, year = {2022}, editor = {Camps-Valls, Gustau and Ruiz, Francisco J. R. and Valera, Isabel}, volume = {151}, series = {Proceedings of Machine Learning Research}, month = {28--30 Mar}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v151/challu22a/challu22a.pdf}, url = {https://proceedings.mlr.press/v151/challu22a.html}, abstract = { Multivariate time series anomaly detection has become an active area of research in recent years, with Deep Learning models outperforming previous approaches on benchmark datasets. Among reconstruction-based models, most previous work has focused on Variational Autoencoders and Generative Adversarial Networks. This work presents DGHL, a new family of generative models for time series anomaly detection, trained by maximizing the observed likelihood by posterior sampling and alternating back-propagation. A top-down Convolution Network maps a novel hierarchical latent space to time series windows, exploiting temporal dynamics to encode information efficiently. Despite relying on posterior sampling, it is computationally more efficient than current approaches, with up to 10x shorter training times than RNN based models. Our method outperformed current state-of-the-art models on four popular benchmark datasets. Finally, DGHL is robust to variable features between entities and accurate even with large proportions of missing values, settings with increasing relevance with the advent of IoT. We demonstrate the superior robustness of DGHL with novel occlusion experiments in this literature. Our code is available at https://github.com/cchallu/dghl. } }
Endnote
%0 Conference Paper %T Deep Generative model with Hierarchical Latent Factors for Time Series Anomaly Detection %A Cristian I. Challu %A Peihong Jiang %A Ying Nian Wu %A Laurent Callot %B Proceedings of The 25th International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2022 %E Gustau Camps-Valls %E Francisco J. R. Ruiz %E Isabel Valera %F pmlr-v151-challu22a %I PMLR %P 1643--1654 %U https://proceedings.mlr.press/v151/challu22a.html %V 151 %X Multivariate time series anomaly detection has become an active area of research in recent years, with Deep Learning models outperforming previous approaches on benchmark datasets. Among reconstruction-based models, most previous work has focused on Variational Autoencoders and Generative Adversarial Networks. This work presents DGHL, a new family of generative models for time series anomaly detection, trained by maximizing the observed likelihood by posterior sampling and alternating back-propagation. A top-down Convolution Network maps a novel hierarchical latent space to time series windows, exploiting temporal dynamics to encode information efficiently. Despite relying on posterior sampling, it is computationally more efficient than current approaches, with up to 10x shorter training times than RNN based models. Our method outperformed current state-of-the-art models on four popular benchmark datasets. Finally, DGHL is robust to variable features between entities and accurate even with large proportions of missing values, settings with increasing relevance with the advent of IoT. We demonstrate the superior robustness of DGHL with novel occlusion experiments in this literature. Our code is available at https://github.com/cchallu/dghl.
APA
Challu, C.I., Jiang, P., Nian Wu, Y. & Callot, L.. (2022). Deep Generative model with Hierarchical Latent Factors for Time Series Anomaly Detection . Proceedings of The 25th International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 151:1643-1654 Available from https://proceedings.mlr.press/v151/challu22a.html.

Related Material