Python code supplement for "Expectation propagation as a way of life"

Requirements (tested version):

• python (3.5.2)
• numpy (1.13.3)
• scipy (0.19.1)
• cython (0.27.3)
• pystan (2.17.0.0)
• matplotlib (2.1.0) (only for plotting the results)

Setup

Compile the Cython utilities with python setup.py build_ext --inplace. Test Scipy compatibility by running python test_scipy.py, see notes on the bottom.

Usage

The folder experiment contains simple hierarchical logistic regression examples. See python fit.py -h or the respective module docstring for help. For more information, see e.g. the class documentation of dep.method.Master.

License

Released under the 3-clause BSD license.

Notes

The program does not work correctly in some scipy builds because of an issue in the in-place operation of dpotri Lapack-routine with C- or F-order matrices. In some builds, the in-place operation works for F-order matrices but not for C-order matrices, whereas in some builds, it works the opposite way. The former behaviour is assumed in this program. This compatibility problem can be tested by running a simple test script by python test_scipy.py.