CONAN._classes.fit_setup.sampling

CONAN._classes.fit_setup.sampling#

CONAN._classes.fit_setup.sampling(sampler='dynesty', n_cpus=4, n_chains=64, n_steps=2000, n_burn=500, emcee_move='stretch', n_live=300, dyn_dlogz=0.1, force_nlive=False, nested_sampling='static', verbose=True)#

configure sampling

Parameters:

  • sampler (str;) – sampler to use. Default is “dynesty”. Options are [“emcee”,”dynesty”].

  • n_cpus (int;) – number of cpus to use for parallelization.

  • n_chains (int;) – number of chains/walkers

  • n_steps (int;) – length of each chain. the effective total steps becomes n_steps*n_chains.

  • n_burn (int;) – number of steps to discard as burn-in

  • emcee_move (str;) – sampler algorithm to use in traversing the parameter space. Options are [“demc”,”snooker”,stretch]. The default is stretch to use the emcee StretchMove.

  • n_live (int;) – number of live points to use for dynesty sampler. Default is 300.

  • dyn_dlogz (float;) – stopping criterion for dynesty sampler. Default is 0.1.

  • force_nlive (bool;) – force dynesty to use n_live even if less than the required ndim*(ndim+1)//2. Default is False.

  • nested_sampling (str;) – type of nested sampling to use. Default is “static”. Options are [“static”,”dynamic[pfrac]”] where pfrac is a float from [0, 1] that determine the posterior.evidence fraction. “dynamic[1.0]” performs sampling optimized for 100% posterior evaluation and “dynamic[0.8]” is 80% posterior, 20% evidence.

  • verbose (bool;) – print output. Default is True.

Contents