Dortmund 2021 – scientific programme

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

T: Fachverband Teilchenphysik

T 38: Data analysis, information technology II

T 38.10: Talk

Tuesday, March 16, 2021, 18:15–18:30, Tm

Optimization of Selective Background Monte Carlo Simulation with Graph Neural Networks at Belle II — •Boyang Yu, Thomas Kuhr, and Nikolai Hartmann — Ludwig-Maximilians-Universität München

When measuring rare processes such as B → K^(∗)νν or B → lνγ, a huge luminosity is required, which means a large number of simulations are necessary to determine signal efficiencies and background contributions. However, this process demands high computation costs while most of the simulated data, in particular in case of background, are discarded by the event selection. Thus filters using neural networks are introduced after the Monte Carlo event generation to speed up the following processes of detector simulation and reconstruction.

In this work, we study optimizations of the performance of neural networks by implementing different architectures with graph neural networks based on moderner libraries and validate them on large datasets. Efficiency, accuracy, ROC curves and AUC values are considered as major criteria.