Git repo for Summer Projects 2024. We will be investigating various ways of improving the discrimination between observed and unobserved Standard Model processes, and improving the kinematic reconstruction for such processes.
Ethan Simpson - ethan.simpson@manchester.ac.uk
Zihan Zhang - zihan.zhang@manchester.ac.uk Skype name: live:exapples_3
Yvonne Peters - yvonne.peters@manchester.ac.uk
I wrote a quick physics primer, with overleaf here
Some computer skills like information about command-line interfaces can be found in TechnicalPrereqs.md
Possible requirements:
Best practice is to work in way that makes you feel most comfortable: this will probably be on your own laptop. Here you can run code in Jupyter / Google Colab notebooks (in a browser), or run scripts from the terminal. One of the best modern tools is VSCode: https://code.visualstudio.com/, which comes itself with a built in terminal, plus you cna build Jupyter notebooks in VSCode which is how I do most code development.
HEP software skills: https://hsf-training.github.io/analysis-essentials/#
These are used to analyse the data structures we store particle collision information in: reading/writing that data, processing it and applying transformations to it, making histograms and plotting results.
Particle physics data is generally stored in .root
files
https://root.cern/ - C++ and Python through PyROOT. ROOT is the main tool people use to do ATLAS analyses. ROOT installation guide: https://root.cern/install/#install-via-a-package-manager. ROOT is slightly harder to pick up from a Python background. If you want to stay in particle physics, you will probably have to use "proper ROOT" eventually.
https://scikit-hep.org/ - Python-based, "modern" alternative to ROOT. More pythonic syntax. More aligned with "data science" software stack, so arguably more applicable for more general data science. Machine-learning tools in general have to interface to this method.
Short example of doing a quick analysis using Scikit-HEP tools from Andy Pilkington available here. This uses uproot
to load the ROOT file, uses vector
to create 4-momentum objects which can manipulated, uses matplotlib
to create a histogram and plot it (and in the background uses awkward-arrays
as the array type).
In time we can store data on shared diskspace on Manchester CSF.
For now I have put some ttbar samples in the following Google Drive: https://drive.google.com/drive/folders/1qIEkxLa28mjkq9DHc2bAvHzpoY0LPj-I?usp=sharing
Currently it contains:
ttbar_200k_dilep.root
= ROOT file with 200,000 ttbar dilepton events. This sample contains final-state particles only (after everything decayed), no top quarks. ttbar_1M_parton_dilep.root
= ROOT file with 1million ttbar dilepton events. This sample contains the truth particles.We will use this to generate the simulated data MadLAD