Targeted Collapse Regularized Autoencoder for Anomaly Detection

A simple regularization technique that penalizes the norm of latent representations in autoencoders can significantly improve their performance for anomaly detection.