SMuK 2023 – scientific programme

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

T: Fachverband Teilchenphysik

T 34: ML Methods II

T 34.1: Talk

Tuesday, March 21, 2023, 17:00–17:15, HSZ/0405

Equivariant Normalising Flows for Particle Jets — •Cedric Ewen — Institut für Experimentalphysik, Universität Hamburg

In high energy physics, current Monte Carlo simulations are time-consuming and the demand for fast computationally efficient simulations is rising. Therefore, generative machine learning models have become a major research interest due to their ability to speed up data generation. A data structure capable of describing collider events such as jets are variable-size point clouds. However, due to complex correlations between the points, a powerful architecture is needed for high generative fidelity. Continuously normalising flows (CNFs) can model these complex point processes while having traceable likelihood and straightforward sampling. We show an implementation of an architecture using CNFs with equivariant functions and compare its performance to multiple GAN approaches on benchmark datasets.