API_LowLvl::lib::fvg

Collaboration diagram for API_LowLvl::lib::fvg:

Data Structures
struct	SprStatsStats
	(this module ignores this pointer) More...

Enumerations
enum	{ SPR_STATS_MEAN, SPR_STATS_SIGMA, SPR_STATS_COVAR, SPR_STATS_MOMENTS_ONLY }

Functions
SprMathReal	spr_math_stats_cvt (int vlen, double *stats, SprMathReal *mu, SprMathReal *cov, SprMathReal Npt_lim)
SprMathReal *	spr_math_stats2fvg (SprMathReal *stats, int vlen, SprMathReal *tmp)
double	spr_math_fvgs_read (SprMathReal ***stats, char *fname, int Efmt, int *Vlen, int *Ngauss)
double	spr_math_fvgs_eval (double *pvec, SprMathReal **fvgs, int Ngauss, int vlen, SprMathReal *x, SprMathReal *mc_x)
SprStatsStats *	spr_stats_free (SprStatsStats *stats)
void	spr_stats_init (SprStatsStats *stats)
SprStatsStats *	spr_stats_alloc (SprStatsStats *stats, int what, int vlen)
void	spr_stats_update (SprStatsStats *stats, const SprMathReal *data, int what)
void	spr_stats_wupdate (SprStatsStats *stats, const SprMathReal *data, float weight, int what)
int	spr_stats_calc (SprStatsStats *stats, int what, const SprMathReal *const *Data, int ndata)

Detailed Description

Enumeration Type Documentation

anonymous enum

Enumerator
SPR_STATS_MEAN
SPR_STATS_SIGMA
SPR_STATS_COVAR
SPR_STATS_MOMENTS_ONLY

Function Documentation

SprMathReal spr_math_stats_cvt	(	int	vlen,
		double *	stats,
		SprMathReal *	mu,
		SprMathReal *	cov,
		SprMathReal	Npt_lim
	)

Conversion of the statistics (M0,M1,M2) -> (N,mu,cov). The number of points 'N' is returned as return value. The mean vector mu may be the NULL pointer. The covariance matrix cov is stored as an upper triangular matrix (first row, second 'partial' row, ...). Distributions that are estimated with less than Npt_lim are replaced with a standard normal distribution.

SprMathReal* spr_math_stats2fvg	(	SprMathReal *	stats,
		int	vlen,
		SprMathReal *	tmp
	)

Conversion of the statistics (N,mu,cov) to statistics that are ready for fast evaluation (Mu,Lr,log(|cov|)) format. The Lr matrix is the result of an LDL-decomposition of the covariance matrix, and had the following property:

 cov^(-1) = Lr(end:-1:1,:)*Lr(end:-1:1,:)'

with Lr(end:-1:1,:) a lower triangular matrix.
The scratch array tmp must be of the same size as the stats array.

double spr_math_fvgs_read	(	SprMathReal ***	stats,
		char *	fname,
		int	Efmt,
		int *	Vlen,
		int *	Ngauss
	)

Read a set of full covariance matrices. If the input matrix is NULL, a matrix of the appropriate size will be allocated. The values of Vlen and Ngauss are checked w.r.t. the the key-set, except for (-1) values (see read_matrix for more details). The format indicator Efmt is used to switch between: "standard format:" (N/Ntot,mu,cov) format with cov in upper triangular format. "evaluation format:" (mu,U,Lr,N/Ntot,log(|cov|)) format for fast evaluation.

The total number of points underneath all gaussians (Ntot) is returned (-1.0 on error).

double spr_math_fvgs_eval	(	double *	pvec,
		SprMathReal **	fvgs,
		int	Ngauss,
		int	vlen,
		SprMathReal *	x,
		SprMathReal *	mc_x
	)

Evaluate a set of full variance gaussians <stats,Ngauss,vlen> for the given point x. The variable mc_x is a scratch array of length vlen and is used to store the mean compensated x value. The total probability is returned.

SprStatsStats* spr_stats_free

(

SprStatsStats *

stats

)

void spr_stats_init

(

SprStatsStats *

stats

)

Parameters

stats

stat buffer to initialize

SprStatsStats* spr_stats_alloc	(	SprStatsStats *	stats,
		int	what,
		int	vlen
	)

Parameters

stats	buffer to allocate / complete
what	allocate for the requested statistics
vlen	vector length

void spr_stats_update	(	SprStatsStats *	stats,
		const SprMathReal *	data,
		int	what
	)

Parameters

stats

stat buffer to update

void spr_stats_wupdate	(	SprStatsStats *	stats,
		const SprMathReal *	data,
		float	weight,
		int	what
	)

Parameters

stats	stat buffer to update
weight	weighted updata

int spr_stats_calc	(	SprStatsStats *	stats,
		int	what,
		const SprMathReal *const *	Data,
		int	ndata
	)

Calculate statistics for a given data-set. If the data pointer Data equals the NULL pointer, the statistics accumulated using the UpdateStats() routine will be used.

Parameters

stats	buffer to store result
what	mean, variance, ...
Data	the data to process
ndata	number of vectors