ML-2DHM

This repository contains the code used to produce results in the paper Signal mixture estimation for degenerate heavy Higgses using a deep neural network published in EPJC: link (open access).

@Article{Kvellestad2018,
    author="Kvellestad, Anders
    and Maeland, Steffen
    and Str{\"u}mke, Inga",
    title="Signal mixture estimation for degenerate heavy Higgses using a deep neural network",
    journal="The European Physical Journal C",
    year="2018",
    month="Dec",
    day="12",
    volume="78",
    number="12",
    pages="1010",
    issn="1434-6052",
    doi="10.1140/epjc/s10052-018-6455-z",
    url="https://doi.org/10.1140/epjc/s10052-018-6455-z"
}

Directory structure

• pythia: Contains code for generating Monte Carlo events and performing analysis.
• generate_samples: Scripts to facilitate event generation for specific THDM parameter. Creates train/test/validation data sets.
• plot: Scripts to plot feature distributions
• neuralnet: Scripts to train neural networks
• measure_yields: Create templates for likelihood fits, make network predictions on testing data and assess performance
• scan: Scan over THDM model parameters and calculate cross sections times branching ratios

Each directory has a README file explaining how to run the code.

Dependencies

The scripts are written for Python 2.7 and require

• Keras
• Tensorflow
• Numpy
• Scipy
• Scikit-learn
• h5py
• Matplotlib

In addition, the following physics software is needed:

• Pythia8.2, version 8.219
• 2HDMC, version 1.7.0
• HiggsBounds and HiggsSignals, versions 4.3.1 and 1.4.0 respectively
• SusHi, version 1.6.1
• ROOT, version 6.10.00

The version numbers are the ones used for the paper; other versions might work too.

Reproduce the results

The following steps reproduces the results and plots shown in the paper.

1. Download and compile the programs listed above.
2. Edit env.sh to point to the correct installation paths.
3. Do source env.sh
4. Go to the pythia directory, and follow instructions for compiling the analysis code against the Pythia installation.
5. Go to the generate_samples directory, and follow instructions to generate simulated events.
6. Go to the neuralnet directory, and follow instructions to train a neural network.
7. Go to the measure_yields directory, and follow instructions to create templates, run fits for the phistar and neural net methods, and obtain results.