Classification Based Anomaly Detection for General Data

Page link: https://arxiv.org/abs/2005.02359

Considers the semi-supervised scenario to detect anomalous data. The training set consists of normal examples (which contains no anomalies). After training the anomaly detector, anomalies in the test data are detected. The test data contains both normal and anomalous examples. This supervision is easy to obtain in many practical settings and is less difficult than the fully-unsupervised case.

The idea is to use a classification method as compared to the reconstruction and statistically based methods.

Introduces a novel technique, GOAD: The method first transforms the data into M subspaces and learns a feature space such that inter-class separation is larger than intra-class separation. For the learned features, the distance from the cluster center is correlated with the likelihood of anomaly.