MELGMM |
Maximizes Expectation of Log-Likelihood for Gaussian mixture.
Synopsis:
model = melgmm(X,Alpha)
model = melgmm(X,Alpha,cov_type)
Description:
model = melgmm(X,Alpha) maximizes expectation of log-likelihood
function for Gaussian mixture model
(Mean,Cov,Prior) = argmax F(Mean,Cov,Prior)
Mean,Cov,Prior
where
F = sum sum Alpha(j,i)*log(pdfgauss(X(:,i),Mean(:,y),Cov(:,:,y)))
y i
The solution is returned in the structure model with fields
Mean [dim x ncomp], Cov [dim x dim x ncomp] and Prior [1 x ncomp].
model = melgmm(X,Alpha,cov_type) specifies covariance matrix:
cov_type = 'full' full covariance matrix (default)
cov_type = 'diag' diagonal covarinace matrix
cov_type = 'spherical' spherical covariance matrix
Input:
X [dim x num_data] Data sample.
Alpha [ncomp x num_data] Distribution of hidden state given sample.
cov_type [string] Type of covariacne matrix (see above).
Output:
model [struct] Gaussian mixture model:
.Mean [dim x ncomp] Mean vectors.
.Cov [dim x dim x ncomp] Covariance matrices.
.Prior [1 x ncomp] Distribution of hidden state.
See also
EMGMM, MLCGMM.
About: Statistical Pattern Recognition Toolbox
(C) 1999-2003, Written by Vojtech Franc and Vaclav Hlavac
Czech Technical University Prague
Faculty of Electrical Engineering
Center for Machine Perception
Modifications:
30-apr-2004, VF
19-sep-2003, VF
27-feb-2003, VF