Improving Deep Unsupervised Anomaly Detection by Exploiting VAE Latent Space Distribution

Anomaly detection methods exploiting autoencoders (AE) have shown good performances. Unfortunately, deep non-linear architectures are able to perform high dimensionality reduction while keeping reconstruction error low, thus worsening outlier detecting performances of AEs. To alleviate the above problem, recently some authors have proposed to exploit Variational autoencoders (VAE), which arise as a variant of standard AEs designed for generative purposes. The key idea of VAEs is take into account a regularization term constraining the organization of the latent space. However, VAEs share with standard AEs the problem that they generalize so well that they can also well reconstruct anomalies. In this work we argue that the approach of selecting the worst reconstructed examples as anomalies is too simplistic if a VAE architecture is employed. We show that outliers tend to lie in the sparsest regions of the combined latent/error space and propose a novel unsupervised anomaly detection algorithm, called VAEOut, that identifies outliers by performing density estimation in this augmented feature space. The proposed approach shows sensible improvements in terms of detection performances over the standard approach based on the reconstruction error.