Meta-learning for Robust Anomaly Detection

Atsutoshi Kumagai, Tomoharu Iwata, Hiroshi Takahashi, Yasuhiro Fujiwara
Proceedings of The 26th International Conference on Artificial Intelligence and Statistics, PMLR 206:675-691, 2023.

Abstract

We propose a meta-learning method to improve the anomaly detection performance on unseen target tasks that have only unlabeled data. Existing meta-learning methods for anomaly detection have shown remarkable performance but require labeled data in target tasks. Although they can treat unlabeled data as normal assuming anomalies in the unlabeled data are negligible, this assumption is often violated in practice. As a result, the methods have low performance. Our method meta-learns with related tasks that have labeled and unlabeled data such that the expected test anomaly detection performance is directly improved when the anomaly detector is adapted to given unlabeled data. Our method is based on autoencoders (AEs), which are widely used neural network-based anomaly detectors. We model anomalous attributes for each unlabeled instance in the reconstruction loss of the AE, which are used to prevent the anomalies from being reconstructed; they can remove the effect of the anomalies. We formulate adaptation to the unlabeled data as a learning problem of the last layer of the AE and the anomalous attributes. This formulation enables the optimum solution to be obtained with a closed-form alternate update formula, which is preferable to efficiently maximize the expected test anomaly detection performance. The effectiveness of our method is experimentally shown with four real-world datasets.

Cite this Paper

BibTeX
@InProceedings{pmlr-v206-kumagai23a, title = {Meta-learning for Robust Anomaly Detection}, author = {Kumagai, Atsutoshi and Iwata, Tomoharu and Takahashi, Hiroshi and Fujiwara, Yasuhiro}, booktitle = {Proceedings of The 26th International Conference on Artificial Intelligence and Statistics}, pages = {675--691}, year = {2023}, editor = {Ruiz, Francisco and Dy, Jennifer and van de Meent, Jan-Willem}, volume = {206}, series = {Proceedings of Machine Learning Research}, month = {25--27 Apr}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v206/kumagai23a/kumagai23a.pdf}, url = {https://proceedings.mlr.press/v206/kumagai23a.html}, abstract = {We propose a meta-learning method to improve the anomaly detection performance on unseen target tasks that have only unlabeled data. Existing meta-learning methods for anomaly detection have shown remarkable performance but require labeled data in target tasks. Although they can treat unlabeled data as normal assuming anomalies in the unlabeled data are negligible, this assumption is often violated in practice. As a result, the methods have low performance. Our method meta-learns with related tasks that have labeled and unlabeled data such that the expected test anomaly detection performance is directly improved when the anomaly detector is adapted to given unlabeled data. Our method is based on autoencoders (AEs), which are widely used neural network-based anomaly detectors. We model anomalous attributes for each unlabeled instance in the reconstruction loss of the AE, which are used to prevent the anomalies from being reconstructed; they can remove the effect of the anomalies. We formulate adaptation to the unlabeled data as a learning problem of the last layer of the AE and the anomalous attributes. This formulation enables the optimum solution to be obtained with a closed-form alternate update formula, which is preferable to efficiently maximize the expected test anomaly detection performance. The effectiveness of our method is experimentally shown with four real-world datasets.} }
Endnote
%0 Conference Paper %T Meta-learning for Robust Anomaly Detection %A Atsutoshi Kumagai %A Tomoharu Iwata %A Hiroshi Takahashi %A Yasuhiro Fujiwara %B Proceedings of The 26th International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2023 %E Francisco Ruiz %E Jennifer Dy %E Jan-Willem van de Meent %F pmlr-v206-kumagai23a %I PMLR %P 675--691 %U https://proceedings.mlr.press/v206/kumagai23a.html %V 206 %X We propose a meta-learning method to improve the anomaly detection performance on unseen target tasks that have only unlabeled data. Existing meta-learning methods for anomaly detection have shown remarkable performance but require labeled data in target tasks. Although they can treat unlabeled data as normal assuming anomalies in the unlabeled data are negligible, this assumption is often violated in practice. As a result, the methods have low performance. Our method meta-learns with related tasks that have labeled and unlabeled data such that the expected test anomaly detection performance is directly improved when the anomaly detector is adapted to given unlabeled data. Our method is based on autoencoders (AEs), which are widely used neural network-based anomaly detectors. We model anomalous attributes for each unlabeled instance in the reconstruction loss of the AE, which are used to prevent the anomalies from being reconstructed; they can remove the effect of the anomalies. We formulate adaptation to the unlabeled data as a learning problem of the last layer of the AE and the anomalous attributes. This formulation enables the optimum solution to be obtained with a closed-form alternate update formula, which is preferable to efficiently maximize the expected test anomaly detection performance. The effectiveness of our method is experimentally shown with four real-world datasets.
APA
Kumagai, A., Iwata, T., Takahashi, H. & Fujiwara, Y.. (2023). Meta-learning for Robust Anomaly Detection. Proceedings of The 26th International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 206:675-691 Available from https://proceedings.mlr.press/v206/kumagai23a.html.

Related Material