This repository contains codes used to train 2D layered CNNs for performing signal-background classification on ATLAS RPV-SUSY data.
Tools for use for other datasets:
- Data visulizaiton
- Training CNN
- Viewing model results: learning curves and predictions
The dataset consists of simulation results for the ATLAS experiment. The signal is RPV-SUSY and background is QCD. For further information refer: https://arxiv.org/abs/1711.03573
- To see plots of raw data, use the jupyter notebook
data_visualization/1_view_data.ipynb
- The code containing codes to train data
atlas_cnn/main_code/code/
- Code that trains a set of CNNs on ATLAS SUSY data
The folder
main_code/jpt_notebooks/
: contains Jupyter notebooks that can read models and plot roc curves.
For example, main_code/jpt_notebooks/2_cnn_yaml_config.ipynb
has widgets to achieve this.
The notebook main_code/jpt_notebooks/2_cnn_yaml_config.ipynb
performs full training and testing.