FastICA Class Reference

Class for running FastICA Algorithm. More...

Collaboration diagram for FastICA:

Public Member Functions
double	a1 ()
double	a2 ()
int	approach ()
int	DeflationFixedPointICA (bool stabilization_enabled, bool fine_tuning_enabled, double mu_orig, double mu_k, index_t failure_limit, int used_nonlinearity, int g_orig, int g_fine, double stroke, bool not_fine, bool taking_long, int initial_state_mode, Matrix X, Matrix B, Matrix W, Matrix *whitening_matrix)
	Run FastICA using Deflation approach.
int	DoFastICA (Matrix W, Matrix Y)
	Runs FastICA Algorithm on matrix X and Inits W to unmixing matrix and Y to independent components matrix, such that .
double	epsilon ()
	FastICA ()
	Default constructor does nothing special.
bool	fine_tune ()
int	FixedPointICA (Matrix X, Matrix whitening_matrix, Matrix *W)
	Verify the validity of some settings, set some parameters needed by the algorithm, and run the fixed-point FastICA algorithm using either the specified approach.
index_t	GetSamples (int max, double percentage, Vector *selected_indices)
	Select indices < max according to probability equal to parameter percentage, and return indices in a Vector.
int	Init (Matrix X_in, struct datanode *module_in)
	Initializes the FastICA object by obtaining everything the algorithm needs.
index_t	max_fine_tune ()
index_t	max_num_iterations ()
double	mu ()
int	nonlinearity ()
index_t	num_of_IC ()
double	percent_cut ()
index_t	RandomSubMatrix (index_t n, double percent_cut, Matrix X, Matrix *X_sub)
	Return Select indices < max according to probability equal to parameter percentage, and return indices in a Vector.
bool	stabilization ()
int	SymmetricFixedPointICA (bool stabilization_enabled, bool fine_tuning_enabled, double mu_orig, double mu_k, index_t failure_limit, int used_nonlinearity, int g_fine, double stroke, bool not_fine, bool taking_long, int initial_state_mode, Matrix X, Matrix B, Matrix W, Matrix *whitening_matrix)
	Run FastICA using Symmetric approach.
Matrix	X ()
Public Attributes
index_t	d
	number of dimensions (components) in original data
index_t	n
	number of samples of original data

Detailed Description

Class for running FastICA Algorithm.

This class takes in a D by N data matrix and runs FastICA, for number of dimensions D and number of samples N

Definition at line 102 of file fastica.h.

Constructor & Destructor Documentation

FastICA::FastICA ( ) [inline]

Default constructor does nothing special.

Definition at line 618 of file fastica.h.

Member Function Documentation

int FastICA::DeflationFixedPointICA	(	bool	stabilization_enabled,
		bool	fine_tuning_enabled,
		double	mu_orig,
		double	mu_k,
		index_t	failure_limit,
		int	used_nonlinearity,
		int	g_orig,
		int	g_fine,
		double	stroke,
		bool	not_fine,
		bool	taking_long,
		int	initial_state_mode,
		Matrix	X,
		Matrix *	B,
		Matrix *	W,
		Matrix *	whitening_matrix
	)			`[inline]`

Run FastICA using Deflation approach.

Definition at line 998 of file fastica.h.

References linalg__private::AddExpert(), GenVector< T >::Copy(), GenVector< T >::CopyValues(), d, GenVector< T >::Init(), Init(), linalg__private::MulOverwrite(), RandomSubMatrix(), linalg__private::RandVector(), linalg__private::Scale(), GenVector< T >::SetZero(), and linalg__private::SubInit().

Referenced by FixedPointICA().

int FastICA::DoFastICA	(	Matrix *	W,
		Matrix *	Y
	)			`[inline]`

Runs FastICA Algorithm on matrix X and Inits W to unmixing matrix and Y to independent components matrix, such that $ X = W * Y $ .

Definition at line 1388 of file fastica.h.

References linalg__private::Center(), FixedPointICA(), Init(), linalg__private::MulInit(), and linalg__private::WhitenUsingEig().

int FastICA::FixedPointICA	(	Matrix	X,
		Matrix	whitening_matrix,
		Matrix *	W
	)			`[inline]`

Verify the validity of some settings, set some parameters needed by the algorithm, and run the fixed-point FastICA algorithm using either the specified approach.

Precondition:: { X is a d by n data matrix, for d dimensions and n samples}

Definition at line 1280 of file fastica.h.

References d, DeflationFixedPointICA(), Init(), and SymmetricFixedPointICA().

Referenced by DoFastICA().

index_t FastICA::GetSamples	(	int	max,
		double	percentage,
		Vector *	selected_indices
	)			`[inline]`

Select indices < max according to probability equal to parameter percentage, and return indices in a Vector.

Precondition:: selected_indices is an uninitialized Vector, percentage in [0 1]

Definition at line 721 of file fastica.h.

References Init(), and GenVector< T >::Init().

Referenced by RandomSubMatrix().

int FastICA::Init	(	Matrix	X_in,
		struct datanode *	module_in
	)			`[inline]`

Initializes the FastICA object by obtaining everything the algorithm needs.

Definition at line 624 of file fastica.h.

References d, fx_param_bool(), fx_param_double(), fx_param_int(), and fx_param_str().

Referenced by DeflationFixedPointICA(), DoFastICA(), FixedPointICA(), GetSamples(), and SymmetricFixedPointICA().

index_t FastICA::RandomSubMatrix	(	index_t	n,
		double	percent_cut,
		Matrix	X,
		Matrix *	X_sub
	)			`[inline]`

Return Select indices < max according to probability equal to parameter percentage, and return indices in a Vector.

Precondition:: selected_indices is an uninitialized Vector, percentage in [0 1]

Definition at line 753 of file fastica.h.

References GetSamples(), and linalg__private::MakeSubMatrixByColumns().

Referenced by DeflationFixedPointICA(), and SymmetricFixedPointICA().

int FastICA::SymmetricFixedPointICA	(	bool	stabilization_enabled,
		bool	fine_tuning_enabled,
		double	mu_orig,
		double	mu_k,
		index_t	failure_limit,
		int	used_nonlinearity,
		int	g_fine,
		double	stroke,
		bool	not_fine,
		bool	taking_long,
		int	initial_state_mode,
		Matrix	X,
		Matrix *	B,
		Matrix *	W,
		Matrix *	whitening_matrix
	)			`[inline]`