LorentzLayerNetworks.jl

Julia implementation of Lorentz layer networks inspired by https://arxiv.org/abs/1707.08966

Files

script: all scripts used for variable extraction, model creation, training, plotting, etc.

Start Julia with julia --project=script to automatically get correct versions of all dependencies required.

• run.jl: main entry point for network training. Change basedir to location of the HDF5 files and output_dir inside the call to save_model to specify output location.
• variables.jl: all relevant variables for trining and plotting
• plots.jl: plot measures of the network. (change variable basedir to location of model output)
• likelihood.jl: perform likelihood analyses and plot results.
• plot_input_dists.jl/plot_recorded_measures.jl: plot input distributions/recorded measures

src: implementation of the LLN and utils

Package should be loaded automatically if specifying script as the project. Can be imported with using JuliaForHEP.

preprocessing

• hdf_preprocessing.py: quick and dirty Python script using Pandas + PyROOT, which extracts Cartesian four momenta from detector coordinates and filters out events with != 6 jets.