<< fastica_gui Scilab ICA/BSS Toolbox icaplot >>

Scilab ICA/BSS Toolbox >> Scilab ICA/BSS Toolbox > fpica

fpica

Fixed point ICA. Main algorithm of FASTICA.

Calling Sequence

[A, W] = fpica(whitesig, whiteningMatrix, dewhiteningMatrix, approach, numOfIC, g, finetune, a1, a2, mu, stabilization, epsilon, maxNumIterations, maxFinetune, initState, guess, sampleSize, displayMode, displayInterval, verbose);

Parameters

whitesig :

the whitened data as row vectors

whiteningMatrix :

can be obtained with function whitenv

dewhiteningMatrix :

can be obtained with function whitenv

approach :

[ 'symm' | 'defl' ] the approach used (deflation or symmetric)

numOfIC :

[ 0 - Dim of whitesig ] number of independent components estimated

g :

[ 'pow3' | 'tanh' | 'gaus' | 'skew' ] the nonlinearity used

finetune :

[same as g + 'off'] the nonlinearity used in finetuning.

a1 :

parameter for tuning 'tanh'

a2 :

parameter for tuning 'gaus'

mu :

step size in stabilized algorithm

stabilization :

[ 'on' | 'off' ] if mu < 1 then automatically on

epsilon :

stopping criterion

maxNumIterations :

maximum number of iterations

maxFinetune :

maximum number of iteretions for finetuning

initState :

[ 'rand' | 'guess' ] initial guess or random initial state. See below

guess :

initial guess for A. Ignored if initState = 'rand'

sampleSize :

[ 0 - 1 ] percentage of the samples used in one iteration

displayMode :

[ 'signals' | 'basis' | 'filters' | 'off' ] plot running estimate

displayInterval :

number of iterations we take between plots

verbose :

[ 'on' | 'off' ] report progress in text format

Description

Perform independent component analysis using Hyvarinen's fixed point algorithm. Outputs an estimate of the mixing matrix A and its inverse W.

Examples

[E, D] = pcamat(vectors);
[nv, wm, dwm] = whitenv(vectors, E, D);
[A, W] = fpica(nv, wm, dwm);

See also

Authors


Report an issue
<< fastica_gui Scilab ICA/BSS Toolbox icaplot >>