Accueil Actualités Articles de recherche récents Tutoriels Ensembles de données Wiki SOTA Modèles LLM Classement GPU Événements

Français

Unsupervised Anomaly Detection On Anoshift

Métriques

ROC-AUC FAR

ROC-AUC IID

ROC-AUC NEAR

ROC-AUC-ID (In-Distribution setup)

Résultats

Résultats de performance de divers modèles sur ce benchmark

Nom du modèle	ROC-AUC FAR	ROC-AUC IID	ROC-AUC NEAR	ROC-AUC-ID (In-Distribution setup)	Paper Title	Repository
OC-SVM	49.57	76.86	71.43	68.73	AnoShift: A Distribution Shift Benchmark for Unsupervised Anomaly Detection
deepSVDD	34.53	92.67	87.00	88.24	AnoShift: A Distribution Shift Benchmark for Unsupervised Anomaly Detection
IsoForest	27.16	86.09	75.26	81.27	AnoShift: A Distribution Shift Benchmark for Unsupervised Anomaly Detection
ACR-DSVDD (zero-shot, anomaly ratio=1%)	62	-	-	-	Zero-Shot Anomaly Detection via Batch Normalization
COPOD	50.42	85.62	54.24	80.89	AnoShift: A Distribution Shift Benchmark for Unsupervised Anomaly Detection
ACR-NTL (zero-shot, test anomaly ratio=1%)	62.5	-	-	-	Zero-Shot Anomaly Detection via Batch Normalization
SO-GAAL	49.35	50.48	54.55	49.90	AnoShift: A Distribution Shift Benchmark for Unsupervised Anomaly Detection
ACR-NTL (zero-shot, test anomaly ratio=20%)	62	-	-	-	Zero-Shot Anomaly Detection via Batch Normalization
ECOD Li et al. (2022)	49.19	84.76	44.87	79.41	AnoShift: A Distribution Shift Benchmark for Unsupervised Anomaly Detection
LOF	34.96	91.5	79.29	87.61	AnoShift: A Distribution Shift Benchmark for Unsupervised Anomaly Detection
Internal Contrastive Learning	22.45	84.86	52.26	66.99	AnoShift: A Distribution Shift Benchmark for Unsupervised Anomaly Detection
AE for anomalies	19.96	81	44.06	64.08	AnoShift: A Distribution Shift Benchmark for Unsupervised Anomaly Detection
ACR-DSVDD (zero-shot, anomaly ratio=20%)	59.1	-	-	-	Zero-Shot Anomaly Detection via Batch Normalization
BERT	28.15	84.54	86.05	79.62	AnoShift: A Distribution Shift Benchmark for Unsupervised Anomaly Detection
LUNAR	28.19	85.75	49.03	78.53	AnoShift: A Distribution Shift Benchmark for Unsupervised Anomaly Detection

0 of 15 row(s) selected.