heinertholen / hepml

hepML

A package for machine learning studies with HEP

When installing make sure to initialise and update the submodule:

• git submodule init

• git submodule update

Start by turning your root trees into numpy arrays! Then carry out some python based data analysis.

Setup with anaconda:

• Install anaconda v4 (I couldn't get it to work with v5) e.g. using

  • bash /nfs/dust/cms/user/elwoodad/Anaconda2-4.4.0-Linux-x86_64.sh

  • choose whether you want anaconda added to bashrc, overwriting system default packages (i didn't)

  • point the anacondaSetup.sh file to your install (see mine as example)

• Make a conda environment with all the software you'll need (hepML)

  • When running on the naf you can do:

  • conda env create -f environment.yml

  • Or you can do a generic install of all the relevant tools (e.g. on maxwell max-display):

  • conda create -n hepML -c nlesc root root_numpy keras pandas seaborn scikit-learn tensorflow tensorflow-gpu pydot

  • Only include tensorflow-gpu if a graphics card is available, otherwise there can be errors

  • NOTE: I've had problems on some systems getting root to install, so you can optionally leave out root and root_numpy if there is an error and install them separately

  • root and numpy are installed using instructions in https://nlesc.gitbooks.io/cern-root-conda-recipes/content/index.html , but the environment file should do everything

• Activate your environment and you're good to go

  • source activate hepML

An example script showing off some of the basic features is available:

• python exampleScript.py

A more sophisticated script that produced the results in https://arxiv.org/abs/1806.00322 is available:

• python run.py