A small framework for binned maximum likelihood analyses using mixture models (signal+background), i.e only fitting the signal fraction. An implementation of the signal subtracted likelihood is available as one of the likelihood formulations

The main code base is in C++ but this project is meant to be used through its python bindings which exposes all the needed functunality to perform most analyses.

To build MLSandbox the following dependencies are needed: gsl, boost, numpy and BoostNumpy (https://github.com/martwo/BoostNumpy)

At the moment only manual build is possible. The easiest way to build the project is to create a build directory, step into the build directory and execute:

  cmake path/to/MLSandbox/source
  make

Besides building the C++ library and the python module this will also create a env_shell.sh. Running this file with sh will create an enviroment where the MLSandbox python module is in the PYTHON_PATH. Therefore after running:

  sh env_shell.sh

you should be able to import MLSandbox in your python script simply with

import MLSandbox

Status of build: