Class
NumCosmoMathFitESMCMC
Description [src]
final class NumCosmoMath.FitESMCMC : GObject.Object
{
/* No available fields */
}Ensemble sampler Markov Chain Monte Carlo analysis.
NcmFitESMCMC is a class that implements the Ensemble sampler Markov Chain
Monte Carlo analysis. The object requires a NcmFit object to be set before
running the analysis. The initial points are sampled from a NcmMSetTransKern
object. The walkers are defined by a NcmFitESMCMCWalker object.
The NcmFitESMCMC object can be run in parallel using MPI.
Constructors
ncm_fit_esmcmc_new_funcs_array
Creates a new NcmFitESMCMC object using the given parameters.
The funcs_array is used to compute extra columns in the catalog.
The functions must be scalar and constant.
Functions
ncm_fit_esmcmc_clear
Decreases the reference count of esmcmc by one. If the reference
count reaches zero, all memory allocated by the object is released.
The esmcmc pointer is set to NULL.
Instance methods
ncm_fit_esmcmc_end_run
Terminates the run. This method should be called after all run related methods.
ncm_fit_esmcmc_free
Decreases the reference count of esmcmc by one. If the reference
count reaches zero, all memory allocated by the object is released.
ncm_fit_esmcmc_get_accept_ratio
The acceptance ratio is the ratio of accepted proposals over the total number of proposals.
ncm_fit_esmcmc_get_accept_ratio_last_update
The acceptance ratio is the ratio of accepted proposals over the total number of proposals during the last update.
ncm_fit_esmcmc_get_offboard_ratio
The offboard ratio is the ratio of offboard proposals over the total number of proposals.
ncm_fit_esmcmc_get_offboard_ratio_last_update
The offboard ratio is the ratio of offboard proposals over the total number of proposals during the last update.
ncm_fit_esmcmc_get_skip_check
Get whether to skip the check of the last ensemble in the catalog when continuing a run.
ncm_fit_esmcmc_run
Runs the Monte Carlo until it reaches the n-th realization. Note that
if the first_id is non-zero it will run n - first_id realizations.
ncm_fit_esmcmc_run_burnin
Runs the ESMCMC algorithm until the prerun-th iterations are reached.
Then it runs ntimes times the estimated constant break.
ncm_fit_esmcmc_run_lre
Runs the ESMCMC algorithm until the least relative error is less than lre.
It runs at least prerun pre-runs before starting the algorithm.
ncm_fit_esmcmc_set_auto_trim
If enable is TRUE turns on the auto-trimming when performing a run_lre.
ncm_fit_esmcmc_set_nthreads
If nthreads is larger than nwalkers / 2, it will be set to
nwalkers / 2.
ncm_fit_esmcmc_set_skip_check
Set whether to skip the check of the last ensemble in the catalog when continuing a run.
ncm_fit_esmcmc_start_run
Starts the run. This method should be called before any other run related method.
ncm_fit_esmcmc_use_mpi
If use_mpi is TRUE then the parallelization will be accomplished
using MPI if any slaves are available. If no slaves are available
then it falls back to threads.
ncm_fit_esmcmc_validate
Recalculates the value of $-2\ln(L)$ and compares with the values found in the catalog. This function is particularly useful to check if any problem occurred during a multithread evaluation of the likelihood.
Signals
Signals inherited from GObject (1)
GObject::notify
The notify signal is emitted on an object when one of its properties has its value set through g_object_set_property(), g_object_set(), et al.