This code uses the output file achieved from simulation of INO-ICAL detector using Geant4. Refer https://github.com/Hemalatanayak/ical_g4/ for more details on simulation. The output files are of four types such as mu+_up, mu-_up, mu+_down, mu-_down. Where,
- ical_output_mu+_up_10GeV_nt_ical_data.csv is generated by shooting a mu+ particle from top of the detector in negative z-direction having energy 10GeV
- ical_output_mu-_up_10GeV_nt_ical_data.csv is generated by shooting a mu- particle from top of the detector in negative z-direction having energy 10GeV
- ical_output_mu+_down_10GeV_nt_ical_data.csv is generated by shooting a mu+ particle from bottom of the detector in positive z-direction having energy 10GeV
- ical_output_mu-_down_10GeV_nt_ical_data.csv is generated by shooting a mu- particle from bottom of the detector in positive z-direction having energy 10GeV
All the input files are concatenated, digitized, labelled and then graph data structure is formed under the section graph.ipynb. Refer https://github.com/Hemalatanayak/ical_graph_data for more details on concatenation, digitization and labelling.
In our construted graph;
- Node feature = digitized hit coordinates
- Edge feature = time difference between two hits
The graphs are labelled as 0, 1, 2 and 3 i.e number of classes are four. Where,
- 0 = mu+ moving in negative z- direction
- 1 = mu- moving in negative z- direction
- 2 = mu+ moving in positive z- direction
- 3 = mu- moving in positive z- direction
The zip files are generated to create a muon classification dataset to which GNN will be applied for further analysis. The zip files contain graphs named as graph_0_1.dgl and so on, where 0 refers to the graphID and 1 refers to energy of the incident muon.