matteoguida / Belle-II-Analysis

Machine learning multiclassification task in particle physics experiment (Belle II) with deep neural networks (DNN) and gradient boosted decision trees (XGBoost).

Belle II analysis            

Prerequisites

Python versions supported:

Installing

You can recreate the conda environment used for this analysis with:

conda env create -f environment.yml

Goal

One of the studies carried out with the Belle II experiment is time-dependent CP asymmetry in the decay channel :

We want to train and test a Deep Neural Network (DNN) with Keras and a Boosted Decision Tree with XGBoost on Montecarlo samples with labeled data and use the best models to find our signal in unlabelled data (Data Challenge).

At the end the branching fraction for the process is calculated.

Authors:

• Valeria Fioroni (University of Padova)
• Philipp Zehetner (University of Padova, Ludwig Maximilian University of Munich)
• Matteo Guida (University of Padova)

Supervised by:

• Professor Marco Zanetti (University of Padova, CERN)
• Professor Stefano Lacaprara (University of Padova, BELLE2)

Useful External Links:

1. A high-bias, low-variance introduction to Machine Learning for physicists - Complete and continuously updated review provided with explanatory Jupyter notebooks.
2. Scikit-HEP project - Particle physics data analysis in Python.
3. CPV in the Standard Model - Some slides on the advanced physics behind the process considered.