VAE with Hyperspherical Coordinates: Improving Anomaly Detection from Hypervolume-Compressed Latent Space

Alejandro Ascárate, Leo Lebrat, Rodrigo Santa Cruz, Clinton Fookes, Olivier Salvado

Keywords: variational autoencoder, anomaly detection, unsupervised, semi-supervised, high dimensional statistics, hyperspherical coordinates

TL;DR: Compressing the latent vectors towards a given direction on the hypersphere improves the anomaly detection ability of the VAE

Abstract: Variational autoencoders (VAE) encode data into lower-dimensional latent vectors before decoding those vectors back to data. Once trained, one can hope to detect out-of-distribution (abnormal) latent vectors, but several issues arise when the latent space is high dimensional. This includes an exponential growth of the hypervolume with the dimension, which severely affects the generative capacity of the VAE. In this paper, we draw insights from high dimensional statistics: in these regimes, the latent vectors of a standard VAE are distributed on the `equators' of a hypersphere, challenging the detection of anomalies. We propose to formulate the latent variables of a VAE using hyperspherical coordinates, which allows compressing the latent vectors towards a given direction on the hypersphere, thereby allowing for a more expressive approximate posterior. We show that this improves both the fully unsupervised and semi-supervised anomaly detection ability of the VAE, achieving the best performance on the datasets we considered, outperforming existing methods. For the unsupervised and semi-supervised modalities, respectively, these are: i) detecting unusual landscape from the Mars Rover camera and unusual Galaxies from ground based imagery (complex, real world datasets); ii) standard benchmarks like Cifar10 and subsets of ImageNet as the in-distribution (ID) class.

Primary Area: unsupervised, self-supervised, semi-supervised, and supervised representation learning

Submission Number: 10167