fitthread Class Reference

thread class to perform fitting and error calculation tasks More...

#include <fitthread.h>

Inheritance diagram for fitthread:

Collaboration diagram for fitthread:

Public Slots
void	stop_calculations ()
void	paraboloid (double **x, double *y, int N, int points, double *fxy)
	fits a dim-dimensional paraboloid centered at zero at points x, UNUSED More...

Signals
void	refresh_progress (int ID, int citer, double chi2)
	progressbar in fitmodel should be updated
void	finished (int ID)
	calculation is finished
void	error_message (int runi, QString message)
	fitmodel instance should print error message

Public Member Functions
void	run ()
void	exit ()
	fitthread (double *para, int *changed, double *errors, int cpara, double(*f)(void *), void *instance, int MaxIter, double conv, bool forceSimplex)
	global fitting More...
	fitthread (double *para, int runi, double *errors, int cpara, double(*f)(void *, int), void *instance, int MaxIter, double conv)
	seriell fitting More...
	fitthread (double *para, int *changed, double *errors, int cpara, double(*chi2)(void *), void *instance, ErrMod mode, double(*fitting)(void *), int MonteCarloIterations, int crun, int cglobal, bool(*correlated)(void *, int, int))
	global error calculation More...
	fitthread (double *para, int runi, double *errors, int cpara, double(*chi2)(void *, int), void *instance, ErrMod mode, int MonteCarloIterations)
	global error calculation
void	errors_covariant ()
	see ErrMod
void	errors_MonteCarlo (double pdfmin, double sigmac)
	see ErrMod More...
void	errors_uncorrelated (bool EmitSignals)
	see ErrMod, EmitSignals may suppress Progressbar update at internal use
void	errors_MINOS ()
	see ErrMod

Detailed Description

thread class to perform fitting and error calculation tasks

Moessfit uses multithreading to efficiently fit and error calculate the data which usually consists of several runs. Multithreading is organized in the following way:
(1) seriell fitting: the model specified by the user uses run specific parameters only. Then every run can be calculated individually (=seriell), no data is shared. For every run a fitthread will be created. All fitthread can work paralell.
(2) global fitting: only one fitthread is used, multithreading is not used by using several fitthreads, but within fitmodel using QtConcurrent to calculate the spectra individually, if the global model has to be updated

A fitthread knows the data to manipulate (which is owned by a fitmodel) and a function to recalculate the related fitmodel and get an updated chi2 value.

Constructor & Destructor Documentation

fitthread::fitthread	(	double *	para,
		int *	changed,
		double *	errors,
		int	cpara,
		double(*)(void *)	f,
		void *	instance,
		int	MaxIter,
		double	conv,
		bool	forceSimplex
	)

global fitting

Parameters

para,:	parameters to be changed
changed,:	help array to indicate if a parameter was changed
errors,:	intial step sizes
cpara,:	parameter count
instance,:	fitmodel instance, is owner and used to recalculate
MaxIter,:	maximum iteration count
conv,:	convergence criterium
forceSimplex,:	use GSL Simplex minimizer instead internal one

fitthread::fitthread	(	double *	para,
		int	runi,
		double *	errors,
		int	cpara,
		double(*)(void *, int)	f,
		void *	instance,
		int	MaxIter,
		double	conv
	)

seriell fitting

Parameters

runi,:

run index in fitmodel for other parameters see first constructor

fitthread::fitthread	(	double *	para,
		int *	changed,
		double *	errors,
		int	cpara,
		double(*)(void *)	chi2,
		void *	instance,
		ErrMod	mode,
		double(*)(void *)	fitting,
		int	MonteCarloIterations,
		int	crun,
		int	cglobal,
		bool(*)(void *, int, int)	correlated
	)

global error calculation

Parameters

MonteCarloIterations,:	sample count to be probed
crun,:	run count
cglobal,:	global parameter count
correlated,:	for other parameters see first constructor

Member Function Documentation

void fitthread::errors_MonteCarlo	(	double	pdfmin,
		double	sigmac
	)

see ErrMod

Parameters

sigma,:

is the multiplier to uncorrelated error intervall. The region to probe is defined by these extended intervalls

void fitthread::paraboloid ( double **  x, double *  y, int  N, int  points, double *  fxy  )

slot

fits a dim-dimensional paraboloid centered at zero at points x, UNUSED

x is points*dim array holding the arguments, y are the function values at x (a points-dimensional vetctor and fxy returns the elements of the Hesse Matrix H in the following order: fx = (f11, f12, f13 ..., f22, f23,...,fNN)) The data can be used in a quadradic model: f(x) = 0.5*xHx

---

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

• fitthread.h
• fitthread.cpp