Master Thesis Open Access
Haide, Isabel
Quast, Günter; Husemann, Ulrich; Kahn, James; Goldenzweig, Pablo
The generation of Monte Carlo data is very time- and resource-consuming, which will only increase with higher luminosities of the particle physics experiments. The Time of Propagation detector at the Belle II experiment is the current bottleneck regarding simulation time. This thesis is exploring the concept of replacing the TOP simulation with a simulation done through machine learning methods, also called fast simulation. The focus here is on validating a possible fast simulation, thus ensuring correct modeling of the detector. The current simulation done by GEANT4 is evaluated and validation methods to ensure a correct simulation done by machine learning methods are developed. Furthermore, generative adversarial networks and conditional variational autoencoders are implemented and tested on reduced datasets as possible fast simulation algorithms.
Name | Size | |
---|---|---|
TopFastSim_Thesis.pdf
md5:50a3c0838d8bad64ee581ddb50457624 |
17.0 MB | Download |