Challenges for Unsupervised Anomaly Detection in Particle Physics

TL;DR

This paper investigates the challenges of using variational autoencoders for anomaly detection in particle physics, highlighting hyperparameter sensitivity and exploring optimal transport methods as alternatives.

Contribution

It uncovers the dependence of autoencoder performance on hyperparameters and introduces optimal transport distances as a competitive anomaly detection method.

Findings

Hyperparameters significantly influence autoencoder effectiveness.

Optimal transport distances can match autoencoder performance.

Background representation choices impact signal detection.

Abstract

Anomaly detection relies on designing a score to determine whether a particular event is uncharacteristic of a given background distribution. One way to define a score is to use autoencoders, which rely on the ability to reconstruct certain types of data (background) but not others (signals). In this paper, we study some challenges associated with variational autoencoders, such as the dependence on hyperparameters and the metric used, in the context of anomalous signal (top and $W$) jets in a QCD background. We find that the hyperparameter choices strongly affect the network performance and that the optimal parameters for one signal are non-optimal for another. In exploring the networks, we uncover a connection between the latent space of a variational autoencoder trained using mean-squared-error and the optimal transport distances within the dataset. We then show that optimal transport distances to representative events in the background dataset can be used directly for anomaly detection, with performance comparable to the autoencoders. Whether using autoencoders or optimal transport distances for anomaly detection, we find that the choices that best represent the background are not necessarily best for signal identification. These challenges with unsupervised anomaly detection bolster the case for additional exploration of semi-supervised or alternative approaches.