Repository for SKIM Bayesian sparse regression in Jax and NumPyro. Code adapted from this paper and this tutorial with major modifications to plotting routines and hyperparameter optimization. More verbose explanation coming soon. Check out the accompanying Google Colab tutorial highlighting the SKIM kernel technique to find predictors for COVID-19 mortality rates by country on data circa April 2020.
To install SKIM and its dependencies in your Python environment, first clone the repository:
git clone https://github.com/tlmakinen/pyskim.git
then,
cd ../pyskim
and install dependencies and the SKIM module via
python3 setup.py install
For an intuituve walkthrough and review of the math behind the kernel, see the browser-based Colab Notebook. To see if your module is working properly on your local OS, try
python _test/test.py
for a quick test of the module. Note for Windows subsystem for Linux users: the testing function should display a sample corner plot for the test covariates through matplotlib. Be sure to have X window forwarding enabled !