MLHad MLhad has taken part in the following codebases:
MLHad Public Repo
Conceptual image of the MLHad implementation of a Machine-Learning-based hadronization model to generate hadronization chains.
This repository contains the developped Machine-Learning-based hadronization models detailed in our published work. These models can be used to generate hadronization chains and compute high level observables. The public directory includes example files allowing the user to train and implement cSWAE and BNF models in full fragmentation chains. The programs are written in Python and extensively use the Pythia, PyTorch and Sci-kit learn libraries. Installation instructions can be found on the respective installation pages for each library.
MLHad Weights
Validation plot for the hadronization weight implementation, showing how the average time required to generate a single event evolves as a function of the number of alternative parameter values calculated during the generation.
Modified version of Pythia 8 to allow for hadronization weights as described in Reweighting Monte Carlo Predictions and Automated Fragmentation Variations in Pythia 8. This is intended to be a temporary repository for prototype reweighting, and will be deprecated once this code is included in an official Pythia release.
MLHad Weights Validation
Example of a validation plot produced with the code.
This code produces validation plots for the WeightsFragmentation class which can be used to reweight the Pythia hadronization model for different parameters, as described in Reweighting Monte Carlo Predictions and Automated Fragmentation Variations in Pythia 8.
Pythia 8 Contrib
This code builds a platform for user contributions to Pythia 8, based on the developped plugin implementation in Pythia 8.
Pythia 8
MLHad has helped develop the new Plugins implementation in Pythia 8.
EMD4CPV
A qualitative view of how the Earth Mover's Distance evaluates CP violation by comparing three-body decay distributions.
The code computes a test statistic for CP violation based on the Wasserstein (Earth) Mover's Distance, as introduced in Earth mover’s distance as a measure of CP violation, partly authored by members of the MLHad collaboration.
hsf23 tutorial
Example tuning Jupyter notebook produced in the tutorial.
Tutorial for Pythia 8 with a focus on hadronization parameter tuning.
Visualization
Work in progress.