Research map / Time series / Anomaly detection
EVEREST: A Transformer for Probabilistic Rare-Event Anomaly Detection with Evidential and Tail-Aware Uncertainty
Antanas Žilinskas, Robert Noel Shorten, Jakub Marecek
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 |
|---|---|---|---|
| add attention bottleneck instead of mean pooling (ablation) | EVEREST · M5-72h | TSS: 0.319 -> 0.746; +0.427 TSS | Improved |
| add EVT head (ablation) | EVEREST · M5-72h | TSS: 0.461 -> 0.746; +0.285 TSS | Improved |
| add evidential NIG head (ablation) | EVEREST · M5-72h | TSS: 0.682 -> 0.746; +0.064 TSS | Improved |
| add precursor auxiliary head (ablation) | EVEREST · M5-72h | TSS: -0.650 TSS | Improved |
| use focal loss instead of cross-entropy (ablation) | EVEREST · M5-72h | TSS: -0.537 TSS | Improved |
| use mixed precision (AMP) instead of FP32 (ablation) | EVEREST · M5-72h | TSS: 0.000 -> 0.746 | Improved |
| add composite schedule with gamma annealing and joint training (ablation) | EVEREST · M5-72h | TSS: +0.045 TSS | Improved |
| add evidential head (calibration effect) (ablation) | EVEREST · M5-72h | ECE: 0.0111 -> 0.0110 | No clear effect |
| add EVT head (calibration effect) (ablation) | EVEREST · M5-72h | ECE: 0.0336 -> 0.0110 | Improved |
| vary EVT exceedance quantile u in {0.85, 0.90, 0.95} (ablation) | EVEREST · M5-72h | TSS: TSS 0.903 to 0.932; ECE 0.0119 to 0.0164 | No clear effect |
| vary evidential and EVT loss weights over 5x5 grid (0.5 to 1.5 multipliers) (ablation) | EVEREST · M5-72h | TSS and ECE | No clear effect |
| apply EVEREST to SKAB valve anomalies without task-specific tuning | EVEREST · SKAB | F1 (%): 98.2 ± 1.7 vs 91–96 | Improved |
Closest papers by meaning
- The Elephant in the Room: Towards A Reliable Time-Series Anomaly Detection Benchmark (NeurIPS 2024 · similarity 0.85)
- Detect, Explain, Interpret: An End-to-End Benchmark for Time Series Anomaly Detection, Explainability and Interpretability. (NeurIPS 2026 · similarity 0.84)
- COMET: Codebook-based Online-adaptive Multi-scale Embedding for Time-series Anomaly Detection (NeurIPS 2026 · similarity 0.84)
- Rare event modeling with self-regularized normalizing flows: what can we learn from a single failure? (ICLR 2025 · similarity 0.84)
- Structured Temporal Causality for Interpretable Multivariate Time Series Anomaly Detection (NeurIPS 2025 · similarity 0.84)
- TAIL-MIL: Time-Aware and Instance-Learnable Multiple Instance Learning for Multivariate Time Series Anomaly Detection (AAAI 2025 · 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.
