AD-GCL improves graph anomaly detection on low-degree tail nodes by pruning head-node edges to forge tail-like views and completing tail neighborhoods using the model's own anomaly scores.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Revisiting Graph Contrastive Learning on Anomaly Detection: A Structural Imbalance Perspective
AD-GCL improves graph anomaly detection on low-degree tail nodes by pruning head-node edges to forge tail-like views and completing tail neighborhoods using the model's own anomaly scores.