zeus is a Python implementation of the Ensemble Slice Sampling method.
For instance, if you wanted to draw samples from a 10-dimensional Gaussian, you would do something like:
import zeus
import numpy as np
def log_prob(x, ivar):
return - 0.5 * np.sum(ivar * x**2.0)
nsteps, nwalkers, ndim = 1000, 100, 10
ivar = 1.0 / np.random.rand(ndim)
start = np.random.randn(nwalkers,ndim)
sampler = zeus.EnsembleSampler(nwalkers, ndim, log_prob, args=[ivar])
sampler.run_mcmc(start, nsteps)
chain = sampler.get_chain(flat=True)
Read the docs at zeus-mcmc.readthedocs.io
To install zeus
using pip
run:
pip install zeus-mcmc
To install zeus
in a [Ana]Conda environment use:
conda install -c conda-forge zeus-mcmc
Please cite the following papers if you found this code useful in your research:
@article{karamanis2021zeus,
title={zeus: A Python implementation of Ensemble Slice Sampling for efficient Bayesian parameter inference},
author={Karamanis, Minas and Beutler, Florian and Peacock, John A},
journal={arXiv preprint arXiv:2105.03468},
year={2021}
}
@article{karamanis2020ensemble,
title = {Ensemble slice sampling: Parallel, black-box and gradient-free inference for correlated & multimodal distributions},
author = {Karamanis, Minas and Beutler, Florian},
journal = {arXiv preprint arXiv: 2002.06212},
year = {2020}
}
Copyright 2019-2021 Minas Karamanis and contributors.
zeus is free software made available under the GPL-3.0 License. For details see the LICENSE
file.