Neural Transformation Learning for Deep Anomaly Detection Beyond Images

TL;DR

This paper introduces a learnable transformation-based method for deep anomaly detection that effectively handles various data types beyond images, including time series and tabular data, by embedding transformed data into a semantic space.

Contribution

It proposes an end-to-end approach for anomaly detection using neural networks to learn transformations that distinguish normal from anomalous data across multiple data domains.

Findings

Outperforms existing methods on time series anomaly detection.

Achieves higher accuracy on domain-specific tabular datasets.

Effective in both one-vs.-rest and n-vs.-rest settings.

Abstract

Data transformations (e.g. rotations, reflections, and cropping) play an important role in self-supervised learning. Typically, images are transformed into different views, and neural networks trained on tasks involving these views produce useful feature representations for downstream tasks, including anomaly detection. However, for anomaly detection beyond image data, it is often unclear which transformations to use. Here we present a simple end-to-end procedure for anomaly detection with learnable transformations. The key idea is to embed the transformed data into a semantic space such that the transformed data still resemble their untransformed form, while different transformations are easily distinguishable. Extensive experiments on time series demonstrate that our proposed method outperforms existing approaches in the one-vs.-rest setting and is competitive in the more challenging n-vs.-rest anomaly detection task. On tabular datasets from the medical and cyber-security domains, our method learns domain-specific transformations and detects anomalies more accurately than previous work.