Semi-discrete optimal transport algorithms in Python

• Detailed installation instructions
• Documentation

Dependencies

• cmake >= 2.8.12
• a recent C/C++ compiler (at least C++11 capable)
• Python 3
• CGAL >= 4.12
• Eigen

The other dependencies (pybind11, NNRVD, geogram) are included as git submodules and can be fetched using the following command:

git submodule update --init --recursive

Installation

Set the environment variable CGAL_DIR to point to the directory containing the file CGALConfig.cmake

Install with:

python3 setup.py install --user

Run the tests with:

pytest -p no:warnings

We ignore the warnings since scipy.sparse raises a PendingDeprecationWarning exception due to its usage of the numpy.matrix class, see: scipy/scipy#9093.

References

• http://www.benjack.io/2018/02/02/python-cpp-revisited.html

• An algorithm for optimal transport between a simplex soup and a point cloud, Quentin Mérigot, Jocelyn Meyron, Boris Thibert. arXiv preprint

• Initialization procedures for discrete and semi-discrete optimal transport, Jocelyn Meyron. preprint