paper.md

title	tags	authors	affiliations	date	bibliography
emcee v3: A Python ensemble sampling toolkit for affine-invariant MCMC	Python astronomy	name orcid affiliation Daniel Foreman-Mackey 0000-0003-0872-7098 1 name orcid affiliation Will M. Farr 0000-0003-1540-8562 1, 2 name orcid affiliation Manodeep Sinha 0000-0002-4845-1228 3, 4 name orcid affiliation David W. Hogg 0000-0003-2866-9403 1, 5 name orcid affiliation Jeremy S. Sanders 0000-0003-2189-4501 6 name orcid affiliation Joe Zuntz 0000-0001-9789-9646 7 name orcid affiliation Peter K. G. Williams 0000-0003-3734-3587 8, 9 name orcid affiliation Andrew R. J. Nelson 0000-0002-4548-3558 10 name orcid affiliation Tobias Erhardt 0000-0002-6683-6746 11 name orcid affiliation Ilya Pashchenko 0000-0002-9404-7023 12 name orcid affiliation Oriol Abril Pla 0000-0002-1847-9481 13 name orcid affiliation Miguel de Val-Borro 0000-0002-0455-9384 14	name index Center for Computational Astrophysics, Flatiron Institute 1 name index Department of Physics and Astronomy, Stony Brook University, United States 2 name index Centre for Astrophysics & Supercomputing, Swinburne University of Technology, Australia 3 name index ARC Centre of Excellence for All Sky Astrophysics in 3 Dimensions (ASTRO 3D) 4 name index Center for Cosmology and Particle Physics, Department of Physics, New York University 5 name index Max Planck Institute for Extraterrestrial Physics 6 name index Institute for Astronomy, University of Edinburgh, Edinburgh, EH9 3HJ, UK 7 name index Center for Astrophysics | Harvard & Smithsonian 8 name index American Astronomical Society 9 name index Australian Nuclear Science and Technology Organisation, NSW, Australia 10 name index Climate and Environmental Physics and Oeschger Center for Climate Change Research, University of Bern, Bern, Switzerland 11 name index P.N. Lebedev Physical Institute of the Russian Academy of Sciences, Moscow, Russia 12 name index Universitat Pompeu Fabra, Barcelona 13 name index Planetary Science Institute, 1700 East Fort Lowell Rd., Suite 106, Tucson, AZ 85719, USA 14	17 October 2019	paper.bib

Summary

emcee is a Python library implementing a class of affine-invariant ensemble samplers for Markov chain Monte Carlo (MCMC). This package has been widely applied to probabilistic modeling problems in astrophysics where it was originally published [@Foreman-Mackey:2013], with some applications in other fields. When it was first released in 2012, the interface implemented in emcee was fundamentally different from the MCMC libraries that were popular at the time, such as PyMC, because it was specifically designed to work with "black box" models instead of structured graphical models. This has been a popular interface for applications in astrophysics because it is often non-trivial to implement realistic physics within the modeling frameworks required by other libraries. Since emcee's release, other libraries have been developed with similar interfaces, such as dynesty [@Speagle:2019]. The version 3.0 release of emcee is the first major release of the library in about 6 years and it includes a full re-write of the computational backend, several commonly requested features, and a set of new "move" implementations.

This new release includes both small quality of life improvements—like a progress bar using tqdm—and larger features. For example, the new backends interface implements real time serialization of sampling results. By default emcee saves its results in memory (as in the original implementation), but it now also includes a HDFBackend class that serializes the chain to disk using h5py.

The most important new feature included in the version 3.0 release of emcee is the new moves interface. Originally, emcee was implementation of the affine-invariant "stretch move" proposed by @Goodman:2010, but there are other ensemble proposals that can get better performance for certain applications. emcee now includes implementations of several other ensemble moves and an interface for defining custom proposals. The implemented moves include:

• The "stretch move" proposed by @Goodman:2010,
• The "differential evolution" and "differential evolution snooker update" moves [@Ter-Braak:2006; @Ter-Braak:2008], and
• A "kernel density proposal" based on the implementation in the kombine library [@Farr:2015].

emcee has been widely used and the original paper has been highly cited, but there have been many contributions from members of the community. This paper is meant to highlight these contributions and provide citation credit to the academic contributors. A full up-to-date list of contributors can always be found on GitHub.

References