You are here: Start » Function Reference » Principal Component Analysis » CreatePCATransform

CreatePCATransform

Header:	AVL.h
Namespace:	avl

Performs the Principal Component Analysis (PCA) on provided data, creates the feature vector and normalization coefficients (mean and standard deviation of variables).

Syntax

void avl::CreatePCATransform
(
	const avl::Matrix& inMatrix,
	const int inDimensions,
	atl::Optional<float> inVarianceToLeave,
	avl::PCAModel& outPCAModel,
	avl::Matrix& outTransformedMatrix,
	avl::Matrix& diagCovarianceMatrix,
	avl::Matrix& diagNormalizedData,
	atl::Array<int>& diagUsedFeatureIndices
)

Parameters

Name	Type	Range	Default	Description
inMatrix	const Matrix&			Input data, where variables are in column, and examples are in rows.
inDimensions	const int	1 -		How many data dimensions (variables) to be left in transformed data.
inVarianceToLeave	Optional<float>	0.0 - 1.0	0.95f	How many of input data variance should be left in transformed data; overrides inDimensions input.
outPCAModel	PCAModel&			Resulting PCA model.
outTransformedMatrix	Matrix&			Transformed inMatrix with reduced dimensionality.
diagCovarianceMatrix	Matrix&			Covariance matrix of input data.
diagNormalizedData	Matrix&			Input data after normalization: scaling and centering.
diagUsedFeatureIndices	Array<int>&			Indices of columns in inMatrix, which were used as Principal Components.

Errors

List of possible exceptions:

Error type	Description
DomainError	Cannot conduct PCA on empty matrix in CreatePCATransform.
DomainError	Cannot conduct principal component analysis for 1-row data set in CreatePCATransform.
DomainError	Cannot reduce data to less than 1 dimension in CreatePCATransform.
DomainError	Could not compute eigenvalues and/or eigenvectors in CreatePCATransform.
DomainError	inDimensions has to be lesser then inMatrix column count in PCA filter in CreatePCATransform.
DomainError	The process did not converge in CreatePCATransform.
DomainError	The provided data did not satisfy the prerequisites in CreatePCATransform.