Multivariate Time-series Anomaly Detection on SMD 2-4
90.8AUROCCANDI
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| CANDIalpha=0.5%, Backbone=MLP-based autoencoder2026.04 | 90.8 | 60.8 | 35.2 | |
| CANDIalpha=1.0%, Backbone=MLP-based autoencoder2026.04 | 90.8 | 60.8 | 37.2 | |
| CANDIalpha=5.0%, Backbone=MLP-based autoencoder2026.04 | 89.9 | 60 | 51.2 | |
| M2N2alpha=0.5%, Backbone=MLP-based autoencoder2026.04 | 89.5 | 60.5 | 35.5 | |
| M2N2alpha=1.0%, Backbone=MLP-based autoencoder2026.04 | 89.5 | 60.5 | 37.7 | |
| M2N2alpha=5.0%, Backbone=MLP-based autoencoder2026.04 | 82.8 | 46.1 | 31.1 | |
| Pretrainedalpha=0.5%, Backbone=MLP-based autoencoder2026.04 | 82.1 | 45.7 | 35.7 | |
| Pretrainedalpha=1.0%, Backbone=MLP-based autoencoder2026.04 | 82.1 | 45.7 | 37.8 | |
| Pretrainedalpha=5.0%, Backbone=MLP-based autoencoder2026.04 | 82.1 | 45.7 | 31.6 |