The data and experiment code can be found in datasets/wine_quality and datasets/formula1.
This library is provided as a companion to our "Favard Kernels" paper.
To install, clone the repository, and ensure that the location you install
favard_kernels is in your PYTHONPATH environment variable.
The current library depends on mercergp our sparse Gaussian process library,
and ortho, our orthogonal polynomials manipulation library. Links to these
are available in the supplementary material that accompanies the paper that
this library is connected to.
A superset of the dependencies can be found in dependencies.txt, which contains
a dump of pipdeptree on this project.
To run the wine data experiments, use datasets/wine_quality/wine_quality_analysis_2.py.
In the code, changed the pretrained and precompared flags to False, and run.
To run the wine data experiments, use datasets/formula1/analysis_2.py.
In the code, changed the pretrained and precompared flags to False, and run.
Code for generation of the eigenvalue consistency diagram can be found in
./eigenvalue_consistency/; run the file eigenvalue_consistency_diagram.py.
Code for generation of the posterior sampling diagram can be found in
./posterior_sampling/; run the file paper_example.py.
Code for generation of the posterior sampling diagram can be found in
./predictive_density/; run the file experiment.py to generate the data;
then use analysis.py to generate the diagram.