Mainz 2022 – scientific programme

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

HK: Fachverband Physik der Hadronen und Kerne

HK 19: Hadron Structure and Spectroscopy III

HK 19.5: Talk

Monday, March 28, 2022, 17:15–17:30, HK-H8

Multidimensional density estimation using Normalizing Flows — •Ellinor Eckstein — University of Bonn, Bonn, Germany

The investigation of multi-body hadronic decays of beauty and charm hadrons requires detailed estimates of efficiencies and background distributions in multidimensional phase space.

A fairly new approach for model independent density estimation are Normalizing Flows, a Machine Learning technique, which gained popularity in recent years. They provide a method to construct flexible probability density distributions by applying a series of trainable transformations on a simple base distribution. A special feature of these distributions is their invertibility. Consequently, the entire Normalizing Flow is invertible and, thus, a very transparent tool for parametrisations. Due to their straightforward structure NFs are easily expandable into multiple dimensions making them attractive for efficiency or background estimation. This talk gives a brief introduction to Normalizing Flows and demonstrates its performance on LHCb data.