A transdimensional, hierarchical, and BAyesian MCMC sampler to sample from a metamodel

mcmc, bayesian, transdimensional, catalog, hierarchical
pip install pcat==0.1


PCAT (Probabilistic Cataloger)

PCAT is a transdimensional, hierarchical, and Bayesian framework to sample from the posterior probability distribution of a metamodel (union of models with different dimensionality) given some Poisson-distributed data.

In addition to its previous use in the literature to sample from the point source catalog space consistent with the Fermi-LAT gamma-ray data, and the dark matter subhalo catalog space consistent with an Hubble Space Telescope (HST) optical image, it can also be used as a general-purpose Poisson mixture sampler.

During burn-in, it adaptively optimizes its within-model proposal scale to minimize the autocorrelation time. Furthermore, it achieves parallelism through bypassing Python's Global Interpreter Lock (GIL). It is implemented in python2.7 and its theoretical framework is introduced in Daylan, Portillo & Finkbeiner (2016). Refer to its webpage for an introduction.


You can install PCAT either by using pip

pip install pcat

or, by running its setup.py script.

python setup.py install

Note that PCAT depends on TDPY, a library of MCMC and numerical routines. The pip installation will install PCAT along with its dependencies.


PCAT user manual is on ReadTheDocs.