ML4LHC-Tutorial

Data Preparation

• MadGraph: event generation. A summer school reference slides: https://indico.ihep.ac.cn/event/7822/contribution/19/material/slides/0.pdf

• Pythia: parton shower – numpythia: The interface between PYTHIA and NumPy

• Delphes: fast detector simulation
  • Run Delphes without Pythia : instructions here

• FastJet: jet clustering – PyJet: The interface between FastJet and NumPy

• Instructions on generating events on a cluster with pbs scheduler: [Running MadGraph on the cluster]

Public Datasets

• CMS Open Data

Data Analysis

• Framework: ROOT
  • c++: with ExRootAnalysis https://cp3.irmp.ucl.ac.be/projects/ExRootAnalysis/wiki/UserManual
  • Python: PyROOT Tutorial. And here is an example for PyROOT
• HSF
  • Reweighting: https://hsf-training.github.io/analysis-essentials/advanced-python/7DemoReweighting.html

Data Fromatting and Transformation

• root2hdf5: convert root TTrees into HDF5 tables.
• uproot (https://indico.cern.ch/event/686641/contributions/2894906/attachments/1606247/2548596/pivarski-uproot.pdf) /root_numpy
• h5py to create hdf5 files more “machine learning-friendly” than .root file

Deep Learning

• Deep Learning Course repository of Gilles Louppe: https://github.com/glouppe/info8010-deep-learning
• Introduction and basic workflow of DL4HEP: https://github.com/stwunsch/iml_tensorflow_keras_workshop

---

Exercises

Dataset

[to be setup. candidate: zenodo]

Playground

• Basic data generation and analysis: [Exercise01]
• advanced: tt~ generation
• Jet clustering
• Produce standard h5 files

---

Cluster Tips

• Basic usage
• Easily use jupyter notebook on a cluster: https://josephpcohen.com/w/jupyter-notebook-and-hpc-systems/

---

References

• Scikit-HEP: a toolset of Python packages for particle physics.
• A high-bias, low-variance introduction to Machine Learning for physicists, https://arxiv.org/abs/1803.08823