Bump Hunting in Latent Space

TL;DR

This paper introduces a physics-inspired variational autoencoder architecture for unsupervised anomaly detection in large datasets, demonstrating its effectiveness on LHC data and highlighting the benefits of embedding physical observables in the latent space.

Contribution

The paper presents a novel VAE architecture that incorporates physical observables into the latent space for improved anomaly detection in high-energy physics data.

Findings

Competitive performance on LHC Olympics datasets

Embedding physical observables aids anomaly characterization

Classifier remains agnostic to embedded physical features

Abstract

Unsupervised anomaly detection could be crucial in future analyses searching for rare phenomena in large datasets, as for example collected at the LHC. To this end, we introduce a physics inspired variational autoencoder (VAE) architecture which performs competitively and robustly on the LHC Olympics Machine Learning Challenge datasets. We demonstrate how embedding some physical observables directly into the VAE latent space, while at the same time keeping the classifier manifestly agnostic to them, can help to identify and characterise features in measured spectra as caused by the presence of anomalies in a dataset.