flowevidence is a Python package that provides evidence estimations from a set of MCMC samples and the associated unnormalized (log-)posterior values.
flowevidence estimates the posterior density by training a flow architecture directly on the samples. Then, for each of them, an evidence estimate can be computed as the ratio of the associated unnormalized posterior value and the flow pdf prediction.
The package documentation is available at this link. Check out the examples directory for more info (TODO).
flowevidence heavily depends on pytorch and normflows.
- Clone the repository:
git clone https://github.com/asantini29/flowevidence.git
cd flowevidence
- Run install:
python setup.py install
We use SemVer for versioning.
Current Version: 0.0.1
- Alesandro Santini
Get in touch if you would like to contribute!
This project is licensed under the MIT License - see the LICENSE.md file for details.
If you use flowevidence in your research, you can cite it in the following way:
(TODO)
A previous work in this context can be found in ArXiv:2404.12294. Please consider citing it as well.
The idea of translating the evidence-estimation problem in a density-estimation one can also be found in "Statistics, Data Mining, and Machine Learning in Astronomy", Željko, Andrew, Jacob, and Gray. Princeton University Press, 2012.