ExCal detects anomalies in multivariate time series without optimization
ExCal: Optimization-Free Multivariate Time Series Anomaly Detection via Excess Calibration
Machine LearningArtificial Intelligence
Summary
Detecting unusual patterns in complex data that changes over time is hard, especially without labeled examples. The authors developed ExCal, a method that spots anomalies in multi-dimensional time series data by using a combination of simple statistical tools and comparisons to normal patterns learned beforehand, without needing complex training. ExCal processes the entire sequence together rather than step-by-step, which helps it perform well on many tested datasets. The method ranked second among 24 others and runs quickly, but its usefulness for real-time detection is still unknown.
What this means in practice
- •For data scientists: Analyze multivariate time series to detect anomalies without needing labeled data or gradient-based training.
- •For operations teams: Detect unusual events in batch-processed sensor or system logs where online, real-time detection is not required.
Authors
Kadir-Kaan Özer, René Ebeling, Markus Enzweiler
Abstract
We introduce ExCal, a normal-only multivariate time series anomaly detector that combines raw state retrieval, a signal/derivative subspace, and robust statistics across temporal scales. Each expert map receives a high normality floor from the training prefix. During scoring, positive excess is retained, attenuated for broadly firing maps, and combined with a small rank term. The method fits reference memories and projections without gradient optimization or anomaly labels. Prevalence and ranks depend on the complete scored sequence, making inference batch and transductive rather than online. On 180 series from 17 TSB-AD-M families, ExCal obtains AUC-PR $0.421\pm0.37$ and VUS-PR $0.452\pm0.37$ over three seeds, ranking second among 24 methods. Holm-adjusted paired tests detect no significant difference from AxonAD or StreamVAE, which are the two leading non-pretrained methods on TSB-AD-M, on either primary metric, while finding significant positive differences against the other 21 baselines. A separate single-run probe over three series measures median model-wrapper time of $0.0662$ s. Ablations associate the largest primary-metric changes with raw retrieval and excess-magnitude preservation. The evaluated score is decomposable by expert, but its diagnostic utility and online suitability remain untested.