Research map / Time series / Anomaly detection
Detecting Regions of Maximal Divergence for Spatio-Temporal Anomaly Detection
Björn Barz, Erik Rodner, Yanira Guanche García, Joachim Denzler
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use Gaussian model instead of KDE in MDI | MDI · Synthetic anomaly set | Mean AP: improves on the best baseline by 286% | Improved |
| add time-delay embedding (kappa=6, tau=2) to Gaussian MDI (ablation) | MDI Gaussian · Synthetic anomaly set | Mean AP: without embedding the Gaussian model would be worse than KDE and on par with the baselines | Improved |
| add interval proposals (theta=1.5) (ablation) | MDI Gaussian unbiased KL · Synthetic anomaly set | processing time (s): 216 -> 5.2 | Improved |
| add interval proposals (ablation) | MDI variants · Synthetic anomaly set | Mean AP: increased the detection performance by up to 125% | Improved |
| use unbiased KL instead of KL (ablation) | MDI Gaussian · Synthetic anomaly set | Mean AP: did not perform better overall | No clear effect |
| use unbiased KL instead of KL (ablation) | MDI Gaussian · meanshift5, meanshift5_hard, meanshift_hard | AP: AP twice as high | Improved |
| use cross entropy instead of KL divergence (ablation) | MDI Gaussian · Synthetic anomaly set | Mean AP | Improved |
| use Jensen-Shannon divergence instead of KL (ablation) | MDI · Synthetic anomaly set | Mean AP | Mixed |
| add interval proposals (ablation) | MDI · North Sea storms | processing time (s): 27 seconds -> half a second | Improved |
| add interval proposals (ablation) | MDI · North Atlantic SLP | processing time: 4 hours -> 22 seconds | Improved |
| add interval proposals (ablation) | MDI · ViSOR traffic video | processing time: 13 hours -> 5 minutes | Improved |
Closest papers by meaning
- Detect, Explain, Interpret: An End-to-End Benchmark for Time Series Anomaly Detection, Explainability and Interpretability. (NeurIPS 2026 · similarity 0.85)
- Structured Temporal Causality for Interpretable Multivariate Time Series Anomaly Detection (NeurIPS 2025 · similarity 0.85)
- TAIL-MIL: Time-Aware and Instance-Learnable Multiple Instance Learning for Multivariate Time Series Anomaly Detection (AAAI 2025 · similarity 0.84)
- SARAD: Spatial Association-Aware Anomaly Detection and Diagnosis for Multivariate Time Series (NeurIPS 2024 · similarity 0.84)
- ScatterAD: Temporal-Topological Scattering Mechanism for Time Series Anomaly Detection (NeurIPS 2025 · similarity 0.84)
- Retrieval Over Training: Similarity search-based Model Selection for Time Series Anomaly Detection (NeurIPS 2026 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
