LundNet

This repository contains the code and results presented in arXiv:2012.08526.

About

LundNet is a jet tagging framework to train graph-based jet tagging strategies.

Install LundNet

LundNet is tested and supported on 64-bit systems running Linux.

Install LundNet with Python's pip package manager:

git clone https://github.com/fdreyer/lundnet.git
cd lundnet
pip install -e .

To install the package in a specific location, use the "--target=PREFIX_PATH" flag.

This process will copy the lundnet program to your environment python path.

We recommend the installation of the LundNet package using a miniconda3 environment with the configuration specified here.

LundNet requires the following python 3 packages:

• torch
• dgl
• numpy
• fastjet (compiled with --enable-pyext)
• pandas
• json
• gzip
• argparse
• tqdm
• networkx
• uproot_methods
• scipy
• sklearn

Pre-trained models

The final models presented in arXiv:2012.08526 are stored in:

• models/LundNet3: contains the LundNet-3 models for each benchmark.
• models/LundNet5: contains the LundNet-5 models for each benchmark.

Input data

All data used for the final models can be downloaded from the git-lfs repository at https://github.com/JetsGame/data.

Running the code

To launch a test of the code, use

lundnet --demo --save test --device cpu --num-epochs 1

This will run the LundNet code on a sample of 5000 signal and background events and train a model on the CPU for one epoch, saving the results in a new test/ directory.

To train a full model, you can type:

lundnet --model lundnet5 --train-sig TRAIN_SIG --train-bkg TRAIN_BKG
        --val-sig VAL_SIG --val-bkg VAL_BKG --test-sig TEST_SIG --test-bkg TEST_BKG
        --save OUTPUT

where the first six filenames are the locations of the signal and background training, validation and testing samples, and the model is saved to an OUTPUT folder.

To apply an existing LundNet model to a new data set, you can use

lundnet --model lundnet5 --load PATH/TO/model_state.pt --test-sig TEST_SIG --test-bkg TEST_BKG --test-output OUTPUT

which loads the model given as input, before applying it to the TEST_SIG and TEST_BKG samples, with the results then saved to OUTPUT.pickle

To find more options on how to run full models, use

lundnet --help

References

• F. A. Dreyer and H. Qu, "Jet tagging in the Lund plane with graph networks," arXiv:2012.08526