Mercurial > hg > camir-aes2014
annotate toolboxes/FullBNT-1.0.7/KPMstats/mixgauss_sample.m @ 0:e9a9cd732c1e tip
first hg version after svn
author | wolffd |
---|---|
date | Tue, 10 Feb 2015 15:05:51 +0000 |
parents | |
children |
rev | line source |
---|---|
wolffd@0 | 1 function [data, indices] = mixgauss_sample(mu, Sigma, mixweights, Nsamples) |
wolffd@0 | 2 % mixgauss_sample Sample data from a mixture of Gaussians |
wolffd@0 | 3 % function [data, indices] = mixgauss_sample(mu, Sigma, mixweights, Nsamples) |
wolffd@0 | 4 % |
wolffd@0 | 5 % Model is P(X) = sum_k mixweights(k) N(X; mu(:,k), Sigma(:,:,k)) or Sigma(k) for scalar |
wolffd@0 | 6 % data(:,i) is the i'th sample from P(X) |
wolffd@0 | 7 % indices(i) is the component from which sample i was drawn |
wolffd@0 | 8 |
wolffd@0 | 9 [D K] = size(mu); |
wolffd@0 | 10 data = zeros(D, Nsamples); |
wolffd@0 | 11 indices = sample_discrete(mixweights, 1, Nsamples); |
wolffd@0 | 12 for k=1:K |
wolffd@0 | 13 if ndims(Sigma) < 3 |
wolffd@0 | 14 sig = Sigma(k); |
wolffd@0 | 15 else |
wolffd@0 | 16 sig = Sigma(:,:,k); |
wolffd@0 | 17 end |
wolffd@0 | 18 ndx = find(indices==k); |
wolffd@0 | 19 if length(ndx) > 0 |
wolffd@0 | 20 data(:,ndx) = sample_gaussian(mu(:,k), sig, length(ndx))'; |
wolffd@0 | 21 end |
wolffd@0 | 22 end |