core/vpdl/vpdt/vpdt_update_gaussian.h File Reference

Iterative updating of Gaussians. More...

#include <vpdl/vpdt/vpdt_field_traits.h>
#include <vpdl/vpdt/vpdt_gaussian.h>

Go to the source code of this file.

Classes
struct	vpdt_update_covariance< F, Covar, Disambiguate >
	Specialized classes to update covariance of different types. More...
struct	vpdt_update_covariance< F, typename vpdt_eigen_sym_matrix_gen< F >::type, typename vpdt_field_traits< F >::type_is_vector >
	Specialized classes to update covariance of different types. More...
struct	vpdt_update_covariance< F, typename vpdt_field_traits< F >::vector_type, typename vpdt_field_traits< F >::type_is_vector >
	Specialized classes to update covariance of different types. More...
struct	vpdt_update_covariance< F, typename vpdt_field_traits< F >::scalar_type, typename vpdt_field_traits< F >::type_is_vector >
	Specialized classes to update covariance of different types. More...
struct	vpdt_update_covariance< F, typename vpdt_field_traits< F >::scalar_type, typename vpdt_field_traits< F >::type_is_scalar >
	Specialized classes to update covariance of different types. More...
Functions
template<class F , class Covar >
void	vpdt_update_gaussian (vpdt_gaussian< F, Covar > &gaussian, typename vpdt_field_traits< F >::scalar_type rho, const F &sample)
	Update the statistics given a 1D Gaussian distribution and a learning rate.
template<class F , class Covar >
void	vpdt_update_gaussian (vpdt_gaussian< F, Covar > &gaussian, typename vpdt_field_traits< F >::scalar_type rho, const F &sample, typename vpdt_field_traits< F >::scalar_type min_var)
	Update the statistics given a Gaussian distribution and a learning rate.

---

Detailed Description

Iterative updating of Gaussians.

Author:

Matt Leotta (mleotta@lems.brown.edu)

Date:

March 7, 2009

Definition in file vpdt_update_gaussian.h.

---

Function Documentation

template<class F , class Covar >

void vpdt_update_gaussian	(	vpdt_gaussian< F, Covar > &	gaussian,
		typename vpdt_field_traits< F >::scalar_type	rho,
		const F &	sample
	)

Update the statistics given a 1D Gaussian distribution and a learning rate.

Note:

if rho = 1/(num observations) then this just an online cumulative average

Definition at line 145 of file vpdt_update_gaussian.h.

template<class F , class Covar >

void vpdt_update_gaussian	(	vpdt_gaussian< F, Covar > &	gaussian,
		typename vpdt_field_traits< F >::scalar_type	rho,
		const F &	sample,
		typename vpdt_field_traits< F >::scalar_type	min_var
	)

Update the statistics given a Gaussian distribution and a learning rate.

Parameters:

min_var

forces the variance to stay above this limit

Note:

If the same sample is observed repeatedly, the variances will converge to the minimum value parameter rather than zero.

Definition at line 171 of file vpdt_update_gaussian.h.