rickyan / astroabc

Approximate Bayesian Computation Sequential Monte Carlo sampler for parameter estimation.

astroABC

Author: Elise Jennings

arxiv:1608.07606

Astronomy and Computing DOI:10.1016/j.ascom.2017.01.001

astroABC is a Python implementation of an Approximate Bayesian Computation Sequential Monte Carlo (ABC SMC) sampler for parameter estimation.

Key features

• Parallel sampling using MPI or multiprocessing
• MPI communicator can be split so both the sampler, and simulation launched by each particle, can run in parallel
• A Sequential Monte Carlo sampler (see e.g. [Toni et al. 2009], [Beaumont et al. 2009], [Sisson & Fan 2010]) [Toni et al. 2009]:https://arxiv.org/abs/0901.1925 [Sisson & Fan 2010]:http://arxiv.org/abs/1001.2058 [Beaumont et al. 2009]:https://arxiv.org/abs/0805.2256
• A method for iterative adapting tolerance levels using the qth quantile of the distance for t iterations ([Turner & Van Zandt (2012)])
• Scikit-learn covariance matrix estimation using [Ledoit-Wolf shrinkage] for singular matrices [Ledoit-Wolf shrinkage]:http://scikit-learn.org/stable/modules/covariance.html
• A module for specifying particle covariance using method proposed by [Turner & Van Zandt (2012)], optimal covariance matrix for a multivariate normal perturbation kernel, local covariance estimate using scikit-learn KDTree method for nearest neighbours ([Filippi et al 2013]) and a weighted covariance (Beaumont et al 2009) [Turner & Van Zandt (2012)]:http://link.springer.com/article/10.1007/s11336-013-9381-x [Filippi et al 2013]:https://arxiv.org/abs/1106.6280
• Restart files output frequently so an interrupted run can be resumed at any iteration
• Output and restart files are backed up every iteration
• User defined distance metric and simulation methods
• A class for specifying heterogeneous parameter priors
• Methods for drawing from any non-standard prior PDF e.g using Planck/WMAP chains
• A module for specifying a constant, linear, log or exponential tolerance level
• Well-documented examples and sample scripts

Wiki

For more information please read the wiki.

Installing

Install astroABC using pip

$ pip install astroabc

or git clone the repository using the url above. Check the dependencies listed in the next section are installed.

Dependencies

• numpy
• scipy
• mpi4py
• multiprocessing
• sklearn

Python distributions like Anaconda have most of what is needed. You can then conda install or pip install all of the required dependencies.

$ conda install  numpy scipy scikit-learn mpi4py
$ pip install numpy scipy scikit-learn mpi4py

License

Copyright 2016 Elise Jennings

astroABC is free software made available under the MIT License. For details see the LICENSE.txt file.