Research map / Time series / Anomaly detection
Drift doesn't Matter: Dynamic Decomposition with Diffusion Reconstruction for Unstable Multivariate Time Series Anomaly Detection
Chengsen Wang, Zirui Zhuang, Qi Qi, Jingyu Wang, Xingyu Wang, Haifeng Sun, Jianxin Liao
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 temporal transformer (ablation) | D3R · PSM/SMD/SWaT average | affiliation F1: 0.7493 -> 0.8034 | Improved |
| add spatial transformer (ablation) | D3R · PSM/SMD/SWaT average | affiliation F1: 0.7673 -> 0.8034 | Improved |
| add time attention in data-time mix-attention (ablation) | D3R · PSM/SMD/SWaT average | affiliation F1: 0.7347 -> 0.8034 | Improved |
| add data attention in data-time mix-attention (ablation) | D3R · PSM/SMD/SWaT average | affiliation F1: 0.7403 -> 0.8034 | Improved |
| add offset subtraction (ablation) | D3R · PSM/SMD/SWaT average | affiliation F1: 0.7690 -> 0.8034 | Improved |
| add disturbance strategy (random vertical drift in training) (ablation) | D3R · PSM/SMD/SWaT average | affiliation F1: 0.7249 -> 0.8034 | Improved |
| use noise diffusion reconstruction instead of VAE (ablation) | D3R · PSM/SMD/SWaT average | affiliation F1: 0.7677 -> 0.8034 | Improved |
| add trend retaining in noise diffusion (ablation) | D3R · SMD | affiliation F1: 0.8006 -> 0.8682 | Improved |
| add trend retaining in noise diffusion (ablation) | D3R · PSM | affiliation F1: 0.7592 -> 0.7609 | No clear effect |
| add spatial transformer (ablation) | D3R · PSM | affiliation F1: 0.7571 -> 0.7609 | No clear effect |
| replace external noise-diffusion bottleneck with internal VAE latent size, scaling hyperparameter (ablation) | D3R · SMD | F1 span across hyperparameter scale: 8.96% (0.7496->0.8392) vs 3.30% (0.8381->0.8711) | Improved |
| use D3R instead of best baseline | SMD · SMD | affiliation F1: 0.7571 -> 0.8682 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Anomaly Transformer — Anomaly Transformer: Time Series Anomaly Detection with Association Discrepancy (ICLR 2022) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Structured Temporal Causality for Interpretable Multivariate Time Series Anomaly Detection (NeurIPS 2025 · similarity 0.89)
- The Elephant in the Room: Towards A Reliable Time-Series Anomaly Detection Benchmark (NeurIPS 2024 · similarity 0.88)
- CATCH: Channel-Aware Multivariate Time Series Anomaly Detection via Frequency Patching (ICLR 2025 · similarity 0.87)
- GCAD: Anomaly Detection in Multivariate Time Series from the Perspective of Granger Causality (AAAI 2025 · similarity 0.87)
- COMET: Codebook-based Online-adaptive Multi-scale Embedding for Time-series Anomaly Detection (NeurIPS 2026 · similarity 0.86)
- Retrieval Over Training: Similarity search-based Model Selection for Time Series Anomaly Detection (NeurIPS 2026 · similarity 0.86)
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.
