Overview

The taurex-emcee plugin enables the use of the emcee sampler in TauREx3.1 retrievals.

To make use of the plugin, you can simply replace your existing optimizer with the emcee sampler.

For example if we have a multinest optimizer defined as:

[Optimizer]
optimizer = multinest
multi_nest_path = ./multinest

The taurex-emcee enabled version is simply:

[Optimizer]
optimizer = emcee

The emcee sampler has a number of parameters that can be set. These are described in the table below.

Table 1 Parameters for the emcee sampler.

Name	Description	Default
num_global_samples	Number of samples to initially draw from the prior	10000
num_chains	Number of independent ensembles to run	4
num_walkers	Ensemble size	max(100, 4 * ndim)
max_ncalls	Maximum number of likelihood function evaluations	1000000
growth_factor	Factor by which to increase the number of steps	10
max_improvement_loops	Number of times MCMC should be re-attempted	4
num_initial_steps	Number of sampler steps to take in first iteration	100
min_autocorr_times	If > 0, sets autocorelation criterion to converge	0
rhat_max	Sets Gelman-Rubin diagnostic to converge	1.01
geweke_max	Sets Geweke diagnostic to converge	2.0
progress	If True, show progress bars	True

Tip

Find detailed information on convergence criteria at Introduction to Bayesian Analysis Procedures.