meanwalker::GoodmanWeareWalkers Class Reference

This class implements the Goodman-Weare method for taking samples from a particular probability distribution. More...

#include <goodmanwearewalkers.h>

Inheritance diagram for meanwalker::GoodmanWeareWalkers:

List of all members.

Public Member Functions
	GoodmanWeareWalkers (RandomNumberGenerator *pRndGen)
	Constructor, in which you need to specify the random number generator to be used.
bool	init (Function *pProbFunction, const double *pStartCoordCenter, double startCoordRadius, double a=2.0, int Nwalkers=0)
	Initialize the sampler.
bool	init (Function *pProbFunction, const double *pStartCoordCenter, const double *pStartCoordRadii, double a=2.0, int Nwalkers=0)
	Initialize the sampler.
size_t	getNumberOfWalkers () const
	Returns the number of 'walkers' used.

---

Detailed Description

This class implements the Goodman-Weare method for taking samples from a particular probability distribution.

This class implements the Goodman-Weare method for taking samples from a particular probability distribution. Is does this by creating a number of 'walkers', items which perform a sort of biased random walk, which together sample the specified distribution. The probability distribution can be specified either by itself, or as the logarithm of the distribution. In the first case, multiplication by a constant factor doens't matter (i.e. it doesn't have to be normed), which in the second case means that adding a specific constant does not matter.

References

• J. Goodman and J. Weare, Ensemble samplers with affine invariance, Communications in Applied Mathematics and Computational Science, Vol 5 (2010), No. 1, p65-80

---

Constructor & Destructor Documentation

meanwalker::GoodmanWeareWalkers::GoodmanWeareWalkers

(

RandomNumberGenerator *

pRndGen

)

Constructor, in which you need to specify the random number generator to be used.

---

Member Function Documentation

size_t meanwalker::GoodmanWeareWalkers::getNumberOfWalkers

(

)

const [inline]

Returns the number of 'walkers' used.

bool meanwalker::GoodmanWeareWalkers::init	(	Function *	pProbFunction,
		const double *	pStartCoordCenter,
		const double *	pStartCoordRadii,
		double	a = 2.0,
		int	Nwalkers = 0
	)

Initialize the sampler.

Initialize the sampler.

Parameters:

pProbFunction	The probability density to be sampled.
isLog	Flag indicating whether the specified function returns a value proportional to the actual probability, or a value equal to the logarithm of the probability (plus some arbitrary constant).
pStartCoordCenter	If the number of dimensions is larger than one, the starting coordinates of the walkers will be chosen randomly on a hypersphere, or rather a hyperellipsoid, with this center. In the dimension is equal to one, the walkers will be placed randomly in the interval [center-radius, center+radius].
pStartCoordRadii	Radii of of the hyperellipsoid along each axis, or size of the interval (see previous parameter).
a	Scale factor for the stretch move from the Goodman-Weare algorithm.
Nwalkers	The number of walkers which will sample the probability distribution. Must be at least one more than the dimension, which is also the default value (specified when setting this parameter to 0).

bool meanwalker::GoodmanWeareWalkers::init	(	Function *	pProbFunction,
		const double *	pStartCoordCenter,
		double	startCoordRadius,
		double	a = 2.0,
		int	Nwalkers = 0
	)

Initialize the sampler.

Initialize the sampler.

Parameters:

pProbFunction	The probability density to be sampled.
isLog	Flag indicating whether the specified function returns a value proportional to the actual probability, or a value equal to the logarithm of the probability (plus some arbitrary constant).
pStartCoordCenter	If the number of dimensions is larger than one, the starting coordinates of the walkers will be chosen randomly on a hypersphere with this center. In the dimension is equal to one, the walkers will be placed randomly in the interval [center-radius, center+radius].
startCoordRadius	Radius of the hypersphere, or size of the interval (see previous parameter).
a	Scale factor for the stretch move from the Goodman-Weare algorithm.
Nwalkers	The number of walkers which will sample the probability distribution. Must be at least one more than the dimension, which is also the default value (specified when setting this parameter to 0).

---

The documentation for this class was generated from the following file:

• src/goodmanwearewalkers.h