TSA on AutoPilot: Self-tuning Self-supervised Time Series Anomaly Detection

Boje Deforce, Meng-Chieh Lee, Bart Baesens, Estefanía Serral Asensio, Jaemin Yoo, Leman Akoglu

Keywords: Self-supervised Learning, Time-series Anomaly Detection, Hyperparameter Optimization

TL;DR: We present a self-supervised time-series anomaly detection method, which can self-tune augmentation hyperparameters.

Abstract: Time series anomaly detection (TSAD) finds many applications such as monitoring environmental sensors, industry KPIs, patient biomarkers, etc. A two-fold challenge for TSAD is a versatile and unsupervised model that can detect various \textit{different types} of time series anomalies (spikes, discontinuities, trend shifts, etc.) \textit{without any labeled data}. Self-supervised models in particular tackle unsupervised TSAD by transforming the input via various augmentations to create pseudo anomalies for training. However, their performance is sensitive to the choice of augmentation, which is hard to choose in practice, while there exists no effort in the literature on data augmentation tuning for TSAD without labels. Our work aims to fill this gap. We introduce TSAP for \textit{TSA ``on autoPilot”}, which can (self-)tune augmentation hyperparameters end-to-end. It stands on two key components: a differentiable augmentation architecture and an unsupervised validation loss to effectively assess the alignment between augmentation type and anomaly type. Case studies show TSAP’s ability to effectively select the (discrete) augmentation type and associated (continuous) hyperparameters.

Submission Number: 35