Variational Auto Encoder fused with Gaussian Process for Unsupervised Anomaly Detection

Abstract

The identification of abnormal data in high-dimensional and high-complexity situations is a challenging subject. In order to improve the accuracy of abnormal data detection, in this article, we first use Variational Auto Encoder (VAE) to extract the features of high-dimensional data to achieve the effect of data dimensionality reduction. Then the Gaussian process model is applied to establish the latent space's feature distribution field and guide the data's displacement in the latent space. The joint probability distribution matrix of the latent space data is adopted to establish the feedback channel between latent space data distribution and VAE data mapping. Finally, the characteristic distribution field of the Gaussian process model is fed back to the VAE to realize the separation of normal data and abnormal data at the data mapping level.