|
wolffd@0
|
1 function [data, label] = gmmsamp(mix, n)
|
|
wolffd@0
|
2 %GMMSAMP Sample from a Gaussian mixture distribution.
|
|
wolffd@0
|
3 %
|
|
wolffd@0
|
4 % Description
|
|
wolffd@0
|
5 %
|
|
wolffd@0
|
6 % DATA = GSAMP(MIX, N) generates a sample of size N from a Gaussian
|
|
wolffd@0
|
7 % mixture distribution defined by the MIX data structure. The matrix X
|
|
wolffd@0
|
8 % has N rows in which each row represents a MIX.NIN-dimensional sample
|
|
wolffd@0
|
9 % vector.
|
|
wolffd@0
|
10 %
|
|
wolffd@0
|
11 % [DATA, LABEL] = GMMSAMP(MIX, N) also returns a column vector of
|
|
wolffd@0
|
12 % classes (as an index 1..N) LABEL.
|
|
wolffd@0
|
13 %
|
|
wolffd@0
|
14 % See also
|
|
wolffd@0
|
15 % GSAMP, GMM
|
|
wolffd@0
|
16 %
|
|
wolffd@0
|
17
|
|
wolffd@0
|
18 % Copyright (c) Ian T Nabney (1996-2001)
|
|
wolffd@0
|
19
|
|
wolffd@0
|
20 % Check input arguments
|
|
wolffd@0
|
21 errstring = consist(mix, 'gmm');
|
|
wolffd@0
|
22 if ~isempty(errstring)
|
|
wolffd@0
|
23 error(errstring);
|
|
wolffd@0
|
24 end
|
|
wolffd@0
|
25 if n < 1
|
|
wolffd@0
|
26 error('Number of data points must be positive')
|
|
wolffd@0
|
27 end
|
|
wolffd@0
|
28
|
|
wolffd@0
|
29 % Determine number to sample from each component
|
|
wolffd@0
|
30 priors = rand(1, n);
|
|
wolffd@0
|
31
|
|
wolffd@0
|
32 % Pre-allocate data array
|
|
wolffd@0
|
33 data = zeros(n, mix.nin);
|
|
wolffd@0
|
34 if nargout > 1
|
|
wolffd@0
|
35 label = zeros(n, 1);
|
|
wolffd@0
|
36 end
|
|
wolffd@0
|
37 cum_prior = 0; % Cumulative sum of priors
|
|
wolffd@0
|
38 total_samples = 0; % Cumulative sum of number of sampled points
|
|
wolffd@0
|
39 for j = 1:mix.ncentres
|
|
wolffd@0
|
40 num_samples = sum(priors >= cum_prior & ...
|
|
wolffd@0
|
41 priors < cum_prior + mix.priors(j));
|
|
wolffd@0
|
42 % Form a full covariance matrix
|
|
wolffd@0
|
43 switch mix.covar_type
|
|
wolffd@0
|
44 case 'spherical'
|
|
wolffd@0
|
45 covar = mix.covars(j) * eye(mix.nin);
|
|
wolffd@0
|
46 case 'diag'
|
|
wolffd@0
|
47 covar = diag(mix.covars(j, :));
|
|
wolffd@0
|
48 case 'full'
|
|
wolffd@0
|
49 covar = mix.covars(:, :, j);
|
|
wolffd@0
|
50 case 'ppca'
|
|
wolffd@0
|
51 covar = mix.covars(j) * eye(mix.nin) + ...
|
|
wolffd@0
|
52 mix.U(:, :, j)* ...
|
|
wolffd@0
|
53 (diag(mix.lambda(j, :))-(mix.covars(j)*eye(mix.ppca_dim)))* ...
|
|
wolffd@0
|
54 (mix.U(:, :, j)');
|
|
wolffd@0
|
55 otherwise
|
|
wolffd@0
|
56 error(['Unknown covariance type ', mix.covar_type]);
|
|
wolffd@0
|
57 end
|
|
wolffd@0
|
58 data(total_samples+1:total_samples+num_samples, :) = ...
|
|
wolffd@0
|
59 gsamp(mix.centres(j, :), covar, num_samples);
|
|
wolffd@0
|
60 if nargout > 1
|
|
wolffd@0
|
61 label(total_samples+1:total_samples+num_samples) = j;
|
|
wolffd@0
|
62 end
|
|
wolffd@0
|
63 cum_prior = cum_prior + mix.priors(j);
|
|
wolffd@0
|
64 total_samples = total_samples + num_samples;
|
|
wolffd@0
|
65 end
|
|
wolffd@0
|
66
|
