Deep Learning for High Energy Physics Simulators [Duration: 2 terms]
Suggested Team Size: 2-3 students
Duration: 2 semesters
Berief Description:
PENELOPE is a Monte Carlo simulator for coupled electron-photon transport in arbitrary materials for a wide energy range. Photon, electron, and positron transport histories are generated considering the interactions according to material and the geometry properties. For developing radiation-protecting materials, time consuming simulations are carried out with different material combinations to optimize weight, cost, or some other metric while satisfying the requirements.
Main Goal:
In this project, we will be exploiting Artifical Neural Network (ANN) techniques -especially deep learning techniques- to eliminate the need of Monte Carlo simulations for selected materials and geometry that is in the scope of wearable protective materials. Hence, the required time for material design tasks will be improved significantly.
Required Tasks:
- Investigating Penelope/PenEasy simulations (there will be a walk-through for this task)
- Designing new simulations and collecting simulation outputs for various system settings
- Designing a deep learning model for estimating simulation outputs (requires literature survey and research discussions)
- For the second term, contributing to scientific paper preparation
Useful Links & Reading Materials:
- https://www.oecd-nea.org/jcms/pl_19590
- “Computational Many-Particle Physics,” H. Fehske, R. Schneider, A. Weisse