mc_reject

Header file for the Monte Carlo Rejection Sampler

Functions

void parallel_sample(int threadid, mcr_job job)

internal routine to draw N samples

Should not be called by the user

Just made this separate from the class for now for convenience

Parameters:: job – mcr job — see mc_reject.h

class mcr_sampler

#include <mc_reject.h>

Public Functions

mcr_sampler(void (*probability)(double *prob, void *parameters, int d, int threadid), void (*draw_parameters)(void *prop_parameters, int d, int threadid, gsl_rng *r), int dimension, double *param_ranges_max, double *param_ranges_min)

Class constructor for Monte Carlo Rejection Sampler.

Assumes no parallel threading

`probability’ is the function that should return the probability of a given sample

`draw_parameters’ — NULL, if default should be used. Otherwise, this should propose another parameter set

param_ranges_max/min are only used if the default draw_parameter is used, otherwise can be NULL

Parameters:

dimension – Dimension
param_ranges_max – Max allowed parameters — Shape [dimension]
param_ranges_min – Min allowed parameters — Shape [dimension]

mcr_sampler(void (*probability)(double *prob, void *parameters, int d, int threadid), void (*draw_parameters)(void *prop_parameters, int d, int threadid, gsl_rng *r), int dimension, double *param_ranges_max, double *param_ranges_min, int thread_num, bool thread_safe)

Class constructor for Monte Carlo Rejection Sampler.

Allows parallel threading

`probability’ is the function that should return the probability of a given sample

`draw_parameters’ — NULL, if default should be used. Otherwise, this should propose another parameter set

param_ranges_max/min are only used if the default draw_parameter is used, otherwise can be NULL

Parameters:

dimension – Dimension
param_ranges_max – Max allowed parameters — Shape [dimension]
param_ranges_min – Min allowed parameters — Shape [dimension]
thread_num – Thread number to use
thread_safe – Bool thread safe — true if parallel threading should be used

~mcr_sampler()

Destructor function for mcr_sampler.

Deallocates the gsl_rng memory

void sample_distribution(int N_samples, void **output)

Main driver — Samples from distribution that has been associated with a sampler object.

Parameters:: output – [out] Output array shape [N_samples][dimension] Caste to void **

void sample(void *output, int threadid)

Internal routine to return one accepted sample.

Should not be used directly by user — couldn’t make private though

Parameters:

output – [out] Output, accepted parameter set
threadid – Thread id — not currently used, but included for future proofing

Public Members

bool thread_safe

int thread_num

int dimension

int parallel_job_num

double *param_ranges_max

double *param_ranges_min

std::function<void(double*, void*, int, int)> pfn

std::function<void(void*, int, int, gsl_rng*)> draw_param

double **samples

Private Functions

void draw_param_standard(void *prop_params, int d, int threadid, gsl_rng *r)

Standard option to draw samples randomly.

Assumes the parameter vector is of type double *

Uses the ranges provided by the user to randomly draw samples uniformly

Parameters:

prop_params – [out] Proposed Param
d – Dimension of parameter space
threadid – Thread ID for parallel execution
r – GSL random number seed

void init(void (*probability)(double *prob, void *parameters, int d, int threadid), void (*draw_parameters)(void *prop_parameters, int d, int threadid, gsl_rng *r), int dimension, double *param_ranges_max, double *param_ranges_min, int thread_num, bool thread_safe)

Private Members

const gsl_rng_type *T

gsl_rng **rvec

struct mcr_job

#include <mc_reject.h>

Public Members

mcr_sampler *sampler

int starting_index = 0

int job_length = 0

void **output