GNN classification of track/shower clusters for DUNE LArTPC detectors, developed by Aaron Meade as an Summer 2026 REU project at The University of Kansas, advised by Dr. Maria Brigida Brunetti
Runs off of data from MiniProdN5p1 including t0 info
SEE THE WIKI FOR MORE DETAILED INFORMATION
TrackShowerGNN.ipynb: Notebook that includes pretty much the entire GNN project from top to bottom. Includes loading data, preprocessing, model training, optimization, model inference, analysis.
OptimalInference.ipynb: Notebook designed only to run inferences on data and assign a probability score (pTrack) to clusters. Appends pTrack to root files organized the same way as the input data.
- OptimalModel-{tag}/checkpoints/best_model.pt: Saved weights for the optimal model. File can be replaced to load in new weights, as long as hyperparamters in the model architecture are set to match the model which generated the pt file.
environment.yml: The environment that will be needed to run the model. (Essentially just lists what packages you'll need to have downloaded)