This class acts as the wrapper for all SVD factorizers which are incompatible with CF module. More...

Public Member Functions
	SVDWrapper (const Factorizer &factorizer=Factorizer())

double	Apply (const arma::mat &V, arma::mat &W, arma::mat &sigma, arma::mat &H) const
	Factorizer function which takes SVD of the given matrix and returns the frobenius norm of error. More...

double	Apply (const arma::mat &V, size_t r, arma::mat &W, arma::mat &H) const
	Factorizer function which computes SVD and returns matrices as required by CF module. More...

Detailed Description

template
<
class
Factorizer
=
DummyClass
>

class mlpack::cf::SVDWrapper< Factorizer >

This class acts as the wrapper for all SVD factorizers which are incompatible with CF module.

Normally SVD factrorizers implement Apply method which takes matrix V and factorizes it into P, sigma and Q where V = P * sigma * trans(Q). But CF module requires factrorization to be V = W * H. This class multiplies P and sigma and takes the first 'r' eigenvectors out where 'r' is the rank of factorization. Q matrix is transposed and trimmed to support the rank of factorization. The Factroizer class should implement Apply which takes matrices P, sigma, Q and V as their parameter respectively.

Definition at line 40 of file svd_wrapper.hpp.

Constructor & Destructor Documentation

◆ SVDWrapper()

SVDWrapper ( const Factorizer & factorizer = Factorizer() )

inline

Definition at line 44 of file svd_wrapper.hpp.

Member Function Documentation

◆ Apply() [1/2]

double Apply	(	const arma::mat &	V,
		arma::mat &	W,
		arma::mat &	sigma,
		arma::mat &	H
	)		const

Factorizer function which takes SVD of the given matrix and returns the frobenius norm of error.

Parameters

V	input matrix
W	first unitary matrix
sigma	eigenvalue matrix
H	second unitary matrix

Note: V = W * sigma * arma::trans(H)

◆ Apply() [2/2]

double Apply	(	const arma::mat &	V,
		size_t	r,
		arma::mat &	W,
		arma::mat &	H
	)		const

Factorizer function which computes SVD and returns matrices as required by CF module.