#include <pca.h>

Collaboration diagram for Running_PCA:

Public Member Functions
	Running_PCA (int _J, int _d)

void	project (const Matrix1D< double > &input, Matrix1D< double > &output) const

void	get_eigenvector (int j, Matrix1D< double > &result) const
	Get a certain eigenvector. More...

double	get_eigenvector_variance (int j) const
	Get the variance associated to a certain eigenvector. More...

Public Attributes
int	J
	Total number of eigenvectors to be computed. More...

int	d
	Dimension of the sample vectors. More...

Matrix1D< double >	current_sample_mean
	Current estimate of the population mean. More...

long	n
	Current number of samples seen. More...

Matrix2D< double >	eigenvectors

Matrix1D< double >	sum_all_samples

Matrix1D< double >	sum_proj

Matrix1D< double >	sum_proj2

Detailed Description

Running PCA. Running PCA is an algorithm that estimates iteratively the principal components of a dataset that is provided to the algorithm as vectors are available.

See J. Weng, Y. Zhang, W.S. Hwang. Candid covariance-free incremental principal component analysis. IEEE Trans. On Pattern Analysis and Machine Intelligence, 25(8): 1034-1040 (2003).

Definition at line 227 of file pca.h.

Constructor & Destructor Documentation

◆ Running_PCA()

Running_PCA::Running_PCA	(	int	_J,
		int	_d
	)

Constructor. J is the number of eigenvectors to compute. d is the dimension of the sample vectors.

Definition at line 462 of file pca.cpp.

 {
     J = _J;
     d = _d;
     n = 0;
     sum_all_samples.initZeros(d);
     current_sample_mean.initZeros(d);
     sum_proj.initZeros(d);
     sum_proj2.initZeros(d);
     eigenvectors.initZeros(d, J);
 }

Member Function Documentation

◆ get_eigenvector()

void Running_PCA::get_eigenvector	(	int	j,
		Matrix1D< double > &	result
	)		const

inline

Get a certain eigenvector.

Definition at line 260 of file pca.h.

     {
         eigenvectors.getCol(j, result);
     }

◆ get_eigenvector_variance()

double Running_PCA::get_eigenvector_variance ( int j ) const

inline

Get the variance associated to a certain eigenvector.

Definition at line 266 of file pca.h.

     {
         if (n <= j)
             return 0.0;
         double mean_proj = sum_proj(j) / n;
         return sum_proj2(j) / n - mean_proj*mean_proj;
     }

◆ project()

void Running_PCA::project	(	const Matrix1D< double > &	input,
		Matrix1D< double > &	output
	)		const

Project a sample vector on the PCA space.

Definition at line 548 of file pca.cpp.

 {
     output.initZeros(J);
     for (int j = 0; j < J; j++)
         for (int i = 0; i < d; i++)
             output(j) += (input(i) - current_sample_mean(i)) * eigenvectors(i, j);
 }

Member Data Documentation

◆ current_sample_mean

Matrix1D<double> Running_PCA::current_sample_mean

Current estimate of the population mean.

Definition at line 237 of file pca.h.

◆ d

int Running_PCA::d

Dimension of the sample vectors.

Definition at line 234 of file pca.h.

◆ eigenvectors

Matrix2D<double> Running_PCA::eigenvectors

Current estimate of the eigenvectors. Each column is an eigenvector.

Definition at line 244 of file pca.h.

◆ J

int Running_PCA::J

Total number of eigenvectors to be computed.

Definition at line 231 of file pca.h.

◆ n

long Running_PCA::n

Current number of samples seen.

Definition at line 240 of file pca.h.

◆ sum_all_samples

Matrix1D<double> Running_PCA::sum_all_samples

Definition at line 275 of file pca.h.

◆ sum_proj

Matrix1D<double> Running_PCA::sum_proj

Definition at line 278 of file pca.h.

◆ sum_proj2

Matrix1D<double> Running_PCA::sum_proj2

Definition at line 281 of file pca.h.

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

xmipp/libraries/classification/pca.h
xmipp/libraries/classification/pca.cpp

Public Member Functions

Public Attributes