FCDD - Explainable Anomaly Detection

Anonymous

Keywords: anomaly-detection, explainability, one-class classification

Abstract: Anomaly detection (AD) is the task of identifying anomalies in a corpus of data. There are many real-life applications where we find anomalies including applications in healthcare and manufacturing. We briefly discuss some of the existing approaches for solving this problem before focusing on one specific approach. FCDD, an approach introduced in a previous ICLR paper , provides an explainable way to solve anomaly detection, while providing SOTA performance. In this blog, we go over the paper and discuss the advantages and disadvantages of the methods involved in it. We also review the various additional experiments from the authors of the paper which were included in the appendix to the main paper.

ICLR Paper: https://openreview.net/forum?id=A5VV3UyIQz