Public Member Functions
	GaussianMixtureModel (Eigen::MatrixXd samples, int num_components)
	Use this method to initialize a GMM fitting given data samples with the desired number of components. Training is done with Expectation-Maximization algorithm. More...

	GaussianMixtureModel (std::vector< Eigen::VectorXd > means, std::vector< Eigen::MatrixXd > covariances, std::vector< double > weights)
	Use this method to initialize a GMM starting from mean points and covariance matrices already known. Faster than fitting a GMM from a dataset. The number of component is equal to the number of elements in the input vectors. More...

	GaussianMixtureModel (int num_components)
	Use this method to initialize a GMM from random mean points and covariance matrices. DEPRECATED! The fitting procedure is very slow. It's better to initialize an empty GMM and then set the parameters with the setter methods. More...

	GaussianMixtureModel ()
	Use this method to initialize an empty GMM. Parameters must be set with the setter methods.

	~GaussianMixtureModel ()
	Destructor.

void	fitgmm (Eigen::MatrixXd samples, int num_components, int max_iterations, double tolerance, bool verbose)
	Expectation-Maximization algorithm. Fits a GMM to a dataset recursively iterating between E-step and M-step. Iterations stop when the log likelihood of the data does not change more than a specified tolerance or when the maximum number of iterations is reached. The GMM must be already initialized. More...

void	fitgmm2 (Eigen::MatrixXd samples, int num_components, int max_iterations, double tolerance, bool verbose)
	Expectation-Maximization algorithm. Fits a GMM to a dataset recursively iterating between E-step and M-step. Iterations stop when the log likelihood of the data does not change more than a specified tolerance or when the maximum number of iterations is reached. The GMM must be already initialized. More...

void	fitgmm (std::vector< Eigen::VectorXd > samples, int num_components, int max_iterations, double tolerance, bool verbose)
	Expectation-Maximization algorithm. Fits a GMM to a dataset recursively iterating between E-step and M-step. Iterations stop when the log likelihood of the data does not change more than a specified tolerance or when the maximum number of iterations is reached. The GMM must be already initialized. Result is the same as above, but convergence should be faster. More...

std::vector< Eigen::VectorXd >	getMeans ()
	Mean getter. More...

std::vector< Eigen::MatrixXd >	getCovariances ()
	Covariances getter. More...

std::vector< double >	getWeights ()
	Weights getter. More...

double	getLogLikelihood ()
	Log-likelihood getter. More...

void	setMeans (std::vector< Eigen::VectorXd > means)
	Mean points setter method. More...

void	setCovariances (std::vector< Eigen::MatrixXd > covariances)
	Covariance matrices setter method. More...

void	setWeights (std::vector< double > weights)
	Weights setter method. More...

bool	check ()
	Check dimensions consistency of the GMM parameters. More...

double	gauss_pdf_2d (Eigen::VectorXd q, Eigen::VectorXd mean, Eigen::MatrixXd cov)
	Use this method to calculate the probability of a given 2D point from a Bivariate Gaussian Distribution (single component). More...

Constructor & Destructor Documentation

◆ GaussianMixtureModel() [1/3]

GaussianMixtureModel::GaussianMixtureModel	(	Eigen::MatrixXd	samples,
		int	num_components
	)

Use this method to initialize a GMM fitting given data samples with the desired number of components. Training is done with Expectation-Maximization algorithm.

Parameters

[in]	samples	training set
[in]	num_components	desired number of clusters

◆ GaussianMixtureModel() [2/3]

GaussianMixtureModel::GaussianMixtureModel	(	std::vector< Eigen::VectorXd >	means,
		std::vector< Eigen::MatrixXd >	covariances,
		std::vector< double >	weights
	)

Use this method to initialize a GMM starting from mean points and covariance matrices already known. Faster than fitting a GMM from a dataset. The number of component is equal to the number of elements in the input vectors.

Parameters

[in]	means	vector of mean points, one for each cluster
[in]	covariances	vector of covariance matrices, one for each cluster
[in]	weights	mixture proportion of each cluster

Exceptions

return error if size of input vectors is not consistent

◆ GaussianMixtureModel() [3/3]

GaussianMixtureModel::GaussianMixtureModel ( int num_components )

Use this method to initialize a GMM from random mean points and covariance matrices. DEPRECATED! The fitting procedure is very slow. It's better to initialize an empty GMM and then set the parameters with the setter methods.

Parameters

[in] num_components number of clusters

Member Function Documentation

◆ check()

bool GaussianMixtureModel::check ( )

Check dimensions consistency of the GMM parameters.

Returns: true if dimensions are consistent, false otherwise

◆ fitgmm() [1/2]

void GaussianMixtureModel::fitgmm	(	Eigen::MatrixXd	samples,
		int	num_components,
		int	max_iterations = `1000`,
		double	tolerance = `1e-3`,
		bool	verbose = `false`
	)

Expectation-Maximization algorithm. Fits a GMM to a dataset recursively iterating between E-step and M-step. Iterations stop when the log likelihood of the data does not change more than a specified tolerance or when the maximum number of iterations is reached. The GMM must be already initialized.

Parameters

[in]	samples	matrix of data samples. Each column is a sample.
[in]	num_components	number of clusters
[in]	max_iterations	maximum number of iterations (default: 1000)
[in]	tolerance	convergence threshold (default: 1e-3)
[in]	verbose	print number of iterations, final log_likelihood and time required for computation (default: false)

◆ fitgmm() [2/2]

void GaussianMixtureModel::fitgmm	(	std::vector< Eigen::VectorXd >	samples,
		int	num_components,
		int	max_iterations = `1000`,
		double	tolerance = `1e-3`,
		bool	verbose = `false`
	)

Expectation-Maximization algorithm. Fits a GMM to a dataset recursively iterating between E-step and M-step. Iterations stop when the log likelihood of the data does not change more than a specified tolerance or when the maximum number of iterations is reached. The GMM must be already initialized. Result is the same as above, but convergence should be faster.

Parameters

[in]	samples	vector of samples. Each element of the std::vector contains coordinates of a sample.
[in]	num_components	number of clusters
[in]	max_iterations	maximum number of iterations (default: 1000)
[in]	tolerance	convergence threshold (default: 1e-3)
[in]	verbose	print number of iterations, final log_likelihood and time required for computation (default: false)

◆ fitgmm2()

void GaussianMixtureModel::fitgmm2	(	Eigen::MatrixXd	samples,
		int	num_components,
		int	max_iterations = `1000`,
		double	tolerance = `1e-3`,
		bool	verbose = `false`
	)

Expectation-Maximization algorithm. Fits a GMM to a dataset recursively iterating between E-step and M-step. Iterations stop when the log likelihood of the data does not change more than a specified tolerance or when the maximum number of iterations is reached. The GMM must be already initialized.

Parameters

[in]	samples	matrix of data samples. Each column is a sample.
[in]	num_components	number of clusters
[in]	max_iterations	maximum number of iterations (default: 1000)
[in]	tolerance	convergence threshold (default: 1e-3)
[in]	verbose	print number of iterations, final log_likelihood and time required for computation (default: false)

◆ gauss_pdf_2d()

double GaussianMixtureModel::gauss_pdf_2d	(	Eigen::VectorXd	q,
		Eigen::VectorXd	mean,
		Eigen::MatrixXd	cov
	)

Use this method to calculate the probability of a given 2D point from a Bivariate Gaussian Distribution (single component).

Parameters

[in]	q	considered point coordinates
[in]	mean	mean point of the Bivariate Gaussian Distribution
[in]	cov	covariance matrix of the Bivariate Gaussian Distribution

◆ getCovariances()

std::vector< Eigen::MatrixXd > GaussianMixtureModel::getCovariances ( )

Covariances getter.

Returns: covariances vector of covariance matrices, one for each cluster

◆ getLogLikelihood()

double GaussianMixtureModel::getLogLikelihood ( )

Log-likelihood getter.

Returns: logLikelihood log-likelihood value of the data

◆ getMeans()

std::vector< Eigen::VectorXd > GaussianMixtureModel::getMeans ( )

Mean getter.

Returns: means vector of mean points, one for each cluster

◆ getWeights()

std::vector< double > GaussianMixtureModel::getWeights ( )

Weights getter.

Returns: weights vector of weights, one for each cluster

◆ setCovariances()

void GaussianMixtureModel::setCovariances ( std::vector< Eigen::MatrixXd > covariances )

Covariance matrices setter method.

Parameters

[in] covariances vector of covariance matrices of each Gaussian component

◆ setMeans()

void GaussianMixtureModel::setMeans ( std::vector< Eigen::VectorXd > means )

Mean points setter method.

Parameters

[in] means vector of mean points of each Gaussian component

◆ setWeights()

void GaussianMixtureModel::setWeights ( std::vector< double > weights )

Weights setter method.

Parameters

[in] weights vector of weights of each Gaussian component

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

include/gaussian_mixture_model/gaussian_mixture_model.h
src/gaussian_mixture_model.cpp

Public Member Functions

Constructor & Destructor Documentation

◆ GaussianMixtureModel() [1/3]

◆ GaussianMixtureModel() [2/3]

◆ GaussianMixtureModel() [3/3]

Member Function Documentation

◆ check()

◆ fitgmm() [1/2]

◆ fitgmm() [2/2]

◆ fitgmm2()

◆ gauss_pdf_2d()

◆ getCovariances()

◆ getLogLikelihood()

◆ getMeans()

◆ getWeights()

◆ setCovariances()

◆ setMeans()

◆ setWeights()