Active Mix Up for Graph Anomaly Detection
Published:
Graph anomaly detection (GAD) aims to learn a function that can detect anomalous entities in a graph. In this tiny paper, we explore node-level anomaly detection. Although graph data is prevalent, ground truth labels are hard to acquire. Hence, active learning can be used to obtain soft labels for efficient supervised learning. Furthermore, we can make better use of the scarce labeled data by applying a data augmentation strategy such as mixup. In this paper, we propose AMUGraph, a method for Active anomaly detection that uses MixUp for data augmentation using soft label on Graphs.