Our work on anomaly detection was accepted to IJCNN 2022!

Our paper “D-AnoGAN: Anomaly Detection in Disconnected Data Manifolds with Generative Adversarial Networks” was accepted for presentation at IJCNN 2022! We explore the problem of unsupervised anomaly detection in disconnected data manifolds. We show that a multi-generator network can be combined with a bandit to learn to cluster data into different manifolds, leading to improved performance on several datasets.