History-based Anomaly Detector: an Adversarial Approach to Anomaly Detection
This addresses anomaly detection, a challenging problem in many areas, with an incremental improvement over existing GAN-based methods.
The authors tackled the problem of anomaly detection by proposing HistoryAD, a self-supervised adversarial method that compares samples to training history from a GAN, achieving top-tier results on several datasets.
Anomaly detection is a difficult problem in many areas and has recently been subject to a lot of attention. Classifying unseen data as anomalous is a challenging matter. Latest proposed methods rely on Generative Adversarial Networks (GANs) to estimate the normal data distribution, and produce an anomaly score prediction for any given data. In this article, we propose a simple yet new adversarial method to tackle this problem, denoted as History-based anomaly detector (HistoryAD). It consists of a self-supervised model, trained to recognize 'normal' samples by comparing them to samples based on the training history of a previously trained GAN. Quantitative and qualitative results are presented evaluating its performance. We also present a comparison to several state-of-the-art methods for anomaly detection showing that our proposal achieves top-tier results on several datasets.