margarine: Posterior Sampling and Marginal Bayesian Statistics

Introduction

margarine:	Marginal Bayesian Statistics
Authors:	Harry T.J. Bevins
Version:	1.4.2
Homepage:	https://github.com/htjb/margarine
Documentation:	https://margarine.readthedocs.io/

Installation

The software should be installed via the git repository using the following commands in the terminal

git clone https://github.com/htjb/margarine.git # or the equivalent using ssh keys
cd margarine
pip install .

or via a pip install with

pip install margarine

Note that the pip install is not always the most up to date version of the code.

margarine versions >=1.3.0 will work with modern versions of tensorflow.

Details/Examples

margarine is designed to make the calculation of marginal bayesian statistics feasible given a set of samples from an MCMC or nested sampling run.

An example of how to use the code can be found on the github in the jupyter notebook notebook/Tutorial.ipynb or alternatively at here.

Documentation

The documentation is available at: https://margarine.readthedocs.io/

To compile it locally you can run

cd docs
sphinx-build source html-build

after cloning the repo and installing the relevant packages with

pip install sphinx numpydoc sphinx_rtd_theme

Licence and Citation

The software is available on the MIT licence.

If you use the code for academic purposes we request that you cite the following paper and the MaxEnt22 proceedings If you use the clustering implementation please cite the following preprint. You can use the following bibtex

@ARTICLE{2023MNRAS.526.4613B,
      author = {{Bevins}, Harry T.~J. and {Handley}, William J. and {Lemos}, Pablo and {Sims}, Peter H. and {de Lera Acedo}, Eloy and {Fialkov}, Anastasia and {Alsing}, Justin},
        title = "{Marginal post-processing of Bayesian inference products with normalizing flows and kernel density estimators}",
      journal = {\mnras},
    keywords = {methods: data analysis, methods: statistical, cosmic background radiation, dark ages, reionization, first stars, Astrophysics - Instrumentation and Methods for Astrophysics, Astrophysics - Cosmology and Nongalactic Astrophysics, Computer Science - Machine Learning},
        year = 2023,
        month = dec,
      volume = {526},
      number = {3},
        pages = {4613-4626},
          doi = {10.1093/mnras/stad2997},
archivePrefix = {arXiv},
      eprint = {2205.12841},
primaryClass = {astro-ph.IM},
      adsurl = {https://ui.adsabs.harvard.edu/abs/2023MNRAS.526.4613B},
      adsnote = {Provided by the SAO/NASA Astrophysics Data System}
}

and

@ARTICLE{2022arXiv220711457B,
     author = {{Bevins}, Harry and {Handley}, Will and {Lemos}, Pablo and {Sims}, Peter and {de Lera Acedo}, Eloy and {Fialkov}, Anastasia},
      title = "{Marginal Bayesian Statistics Using Masked Autoregressive Flows and Kernel Density Estimators with Examples in Cosmology}",
    journal = {arXiv e-prints},
   keywords = {Astrophysics - Cosmology and Nongalactic Astrophysics, Astrophysics - Instrumentation and Methods for Astrophysics},
       year = 2022,
      month = jul,
        eid = {arXiv:2207.11457},
      pages = {arXiv:2207.11457},
archivePrefix = {arXiv},
     eprint = {2207.11457},
primaryClass = {astro-ph.CO},
     adsurl = {https://ui.adsabs.harvard.edu/abs/2022arXiv220711457B},
    adsnote = {Provided by the SAO/NASA Astrophysics Data System}
}

and

@ARTICLE{2023arXiv230502930B,
      author = {{Bevins}, Harry and {Handley}, Will},
        title = "{Piecewise Normalizing Flows}",
      journal = {arXiv e-prints},
    keywords = {Statistics - Machine Learning, Computer Science - Machine Learning},
        year = 2023,
        month = may,
          eid = {arXiv:2305.02930},
        pages = {arXiv:2305.02930},
          doi = {10.48550/arXiv.2305.02930},
archivePrefix = {arXiv},
      eprint = {2305.02930},
primaryClass = {stat.ML},
      adsurl = {https://ui.adsabs.harvard.edu/abs/2023arXiv230502930B},
      adsnote = {Provided by the SAO/NASA Astrophysics Data System}
}

Contributing

Contributions and suggestions for areas of development are welcome and can be made by opening a issue to report a bug or propose a new feature for discussion.

Please see the contribution guidelines in CONTRIBUTING.md for more details.

Roadmap to 2.0.0

margarine is under active development with a roadmap to version 2.0.0 which will include:

• JAX support for KDE and MAF implementations to allow for GPU acceleration.
• A base class density estimator to make adding new density estimators easier.
• A Neural Spline Flow density estimator implementation.