Variational Inference for On-line Anomaly Detection in High-Dimensional Time Series

TL;DR

This paper introduces a variational inference approach using Stochastic Recurrent Networks to detect anomalies in high-dimensional time series data, particularly applied to robot sensor sequences, enabling robust online and offline anomaly detection.

Contribution

It applies variational inference with STORN to high-dimensional time series, demonstrating effective online and offline anomaly detection in robotics data.

Findings

Robust anomaly detection achieved both online and offline.

Effective modeling of complex temporal and spatial structures.

Demonstrated on robot sensor data with promising results.

Abstract

Approximate variational inference has shown to be a powerful tool for modeling unknown complex probability distributions. Recent advances in the field allow us to learn probabilistic models of sequences that actively exploit spatial and temporal structure. We apply a Stochastic Recurrent Network (STORN) to learn robot time series data. Our evaluation demonstrates that we can robustly detect anomalies both off- and on-line.