Anomaly Detection based on Zero-Shot Outlier Synthesis and Hierarchical Feature Distillation

TL;DR

This paper introduces a hierarchical latent space approach for zero-shot anomaly detection by synthesizing outliers from distilled inlier features, enabling robust classifiers without real outlier data.

Contribution

It proposes a novel hierarchical feature distillation method combined with variational autoencoders to generate synthetic outliers for improved anomaly detection.

Findings

Effective synthetic outlier generation improves detection accuracy.

Robust feature representations enhance classifier performance.

Method outperforms existing anomaly detection techniques on benchmarks.

Abstract

Anomaly detection suffers from unbalanced data since anomalies are quite rare. Synthetically generated anomalies are a solution to such ill or not fully defined data. However, synthesis requires an expressive representation to guarantee the quality of the generated data. In this paper, we propose a two-level hierarchical latent space representation that distills inliers' feature-descriptors (through autoencoders) into more robust representations based on a variational family of distributions (through a variational autoencoder) for zero-shot anomaly generation. From the learned latent distributions, we select those that lie on the outskirts of the training data as synthetic-outlier generators. And, we synthesize from them, i.e., generate negative samples without seen them before, to train binary classifiers. We found that the use of the proposed hierarchical structure for feature distillation and fusion creates robust and general representations that allow us to synthesize pseudo outlier samples. And in turn, train robust binary classifiers for true outlier detection (without the need for actual outliers during training). We demonstrate the performance of our proposal on several benchmarks for anomaly detection.