Mercurial > hg > camir-aes2014

annotate toolboxes/FullBNT-1.0.7/KPMstats/condgaussTrainObserved.m @ 0:e9a9cd732c1e tip

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
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 [mu, Sigma] = mixgaussTrainObserved(obsData, hiddenData, nstates, varargin);
wolffd@0 2 % mixgaussTrainObserved Max likelihood estimates of conditional Gaussian from raw data
wolffd@0 3 % function [mu, Sigma] = mixgaussTrainObserved(obsData, hiddenData, nstates, ...);
wolffd@0 4 %
wolffd@0 5 % Input:
wolffd@0 6 % obsData(:,i)
wolffd@0 7 % hiddenData(i) - this is the mixture component label for example i
wolffd@0 8 % Optional arguments - same as mixgauss_Mstep
wolffd@0 9 %
wolffd@0 10 % Output:
wolffd@0 11 % mu(:,q)
wolffd@0 12 % Sigma(:,:,q) - same as mixgauss_Mstep
wolffd@0 13
wolffd@0 14 [D numex] = size(obsData);
wolffd@0 15 Y = zeros(D, nstates);
wolffd@0 16 YY = zeros(D,D,nstates);
wolffd@0 17 YTY = zeros(nstates,1);
wolffd@0 18 w = zeros(nstates, 1);
wolffd@0 19 for q=1:nstates
wolffd@0 20 ndx = find(hiddenData==q);
wolffd@0 21 w(q) = length(ndx); % each data point has probability 1 of being in this cluster
wolffd@0 22 data = obsData(:,ndx);
wolffd@0 23 Y(:,q) = sum(data,2);
wolffd@0 24 YY(:,:,q) = data*data';
wolffd@0 25 YTY(q) = sum(diag(data'*data));
wolffd@0 26 end
wolffd@0 27 [mu, Sigma] = mixgauss_Mstep(w, Y, YY, YTY, varargin{:});