Parts | Days | Selection | Search | Updates | Downloads | Help

T: Fachverband Teilchenphysik

T 16: Data, AI, Computing 1 (anomaly detection)

T 16.1: Talk

Monday, March 4, 2024, 16:00–16:15, Geb. 30.33: MTI

Anomaly Detection Using Autoencoders in Belle II Data — •David Giesegh, Nikolai Hartmann, and Thomas Kuhr — Ludwig-Maximilians-Universität München

At Belle II the search for Beyond the Standard Model (BSM) Physics is an ongoing effort that concentrates mostly on dedicated searches inspired by specific BSM models. Since new effects might be hidden in unexpected observables or correlations thereof, these searches should be complemented by model agnostic methods. For this purpose we explore the application of machine learning models, especially autoencoders, for automated anomaly detection in Belle II data. The main idea is to train a model to compress preprocessed event data and use how well the compression works as an indication for how anomalous an event is. As proof of concept, preliminary results of this method on simulated data scenarios will be presented.

Keywords: Machine Learning; Anomaly Detection; Belle II; Autoencoder