michelif/hhbbgg_eth

HH->bbgg analysis

HHbbgg_ETH

Getting the code

Fork to your directory the code here: https://github.com/michelif/HHbbgg_ETH
Clone it locally:
git clone [email protected]:[YOURNAME]/HHbbgg_ETH.git HHbbgg_ETH_devel
You should create your own branch, then do pull requests when you are done.

Training, optimization of MVAs and ntuple production

In order to run the notebooks on the browser you need an ssh tunnel on a machine where jupyter is running. Opening it in a screen session allows you to detach it and not redoing it every time (you will only need to ssh).

One-time initialization of Jupyter:

• Start jupyter on t3:
  ssh -L 8002:localhost:8002 t3ui02.psi.ch; //this first line has to be done every time
  screen;
  cd HHbbgg_ETH_devel;
  jupyter notebook --port 8002 --no-browser;
  ctrl-a-d //detaches screen session
  In your browser open: http://localhost:8002
  

The code is organized in notebooks, the programs you execute directly in the browser. They are in the "notebooks" folder.

• Training/notebooks
  This folder contains all the notebooks needed
  -- trainMVAHHbbgg.ipynb
  Trains a Gradient Boost Classifier to separate signal/backgrounds. All re-weighting implemented, output saved.
  -- createTrees.ipynb
  It runs on ntuples, applies a given MVA and saves a reduced ntuples which includes the MVA scores
  -- optimizeClassifier.ipynb
  Optimize parameters of the classifier. In Training/scropts/optimizeClassifier.py there is a (probably not-updated) version of this optimization code wich runs without using notebooks.

• Training/python/
  This folder contains all the tools needed by the notebooks: data format, train/test splitting, plotting... These classes have to be loaded at the beginning of your notebook.

Workspaces and datacards production

The instructions for this part of the analysis are in the README in folder Limits/macros

Combine

The instructions for this part of the analysis are in the README in folder Limits/combine_macros