|
wolffd@0
|
1 function net = rbfsetbf(net, options, x)
|
|
wolffd@0
|
2 %RBFSETBF Set basis functions of RBF from data.
|
|
wolffd@0
|
3 %
|
|
wolffd@0
|
4 % Description
|
|
wolffd@0
|
5 % NET = RBFSETBF(NET, OPTIONS, X) sets the basis functions of the RBF
|
|
wolffd@0
|
6 % network NET so that they model the unconditional density of the
|
|
wolffd@0
|
7 % dataset X. This is done by training a GMM with spherical covariances
|
|
wolffd@0
|
8 % using GMMEM. The OPTIONS vector is passed to GMMEM. The widths of
|
|
wolffd@0
|
9 % the functions are set by a call to RBFSETFW.
|
|
wolffd@0
|
10 %
|
|
wolffd@0
|
11 % See also
|
|
wolffd@0
|
12 % RBFTRAIN, RBFSETFW, GMMEM
|
|
wolffd@0
|
13 %
|
|
wolffd@0
|
14
|
|
wolffd@0
|
15 % Copyright (c) Ian T Nabney (1996-2001)
|
|
wolffd@0
|
16
|
|
wolffd@0
|
17 errstring = consist(net, 'rbf', x);
|
|
wolffd@0
|
18 if ~isempty(errstring)
|
|
wolffd@0
|
19 error(errstring);
|
|
wolffd@0
|
20 end
|
|
wolffd@0
|
21
|
|
wolffd@0
|
22 % Create a spherical Gaussian mixture model
|
|
wolffd@0
|
23 mix = gmm(net.nin, net.nhidden, 'spherical');
|
|
wolffd@0
|
24
|
|
wolffd@0
|
25 % Initialise the parameters from the input data
|
|
wolffd@0
|
26 % Just use a small number of k means iterations
|
|
wolffd@0
|
27 kmoptions = zeros(1, 18);
|
|
wolffd@0
|
28 kmoptions(1) = -1; % Turn off warnings
|
|
wolffd@0
|
29 kmoptions(14) = 5; % Just 5 iterations to get centres roughly right
|
|
wolffd@0
|
30 mix = gmminit(mix, x, kmoptions);
|
|
wolffd@0
|
31
|
|
wolffd@0
|
32 % Train mixture model using EM algorithm
|
|
wolffd@0
|
33 [mix, options] = gmmem(mix, x, options);
|
|
wolffd@0
|
34
|
|
wolffd@0
|
35 % Now set the centres of the RBF from the centres of the mixture model
|
|
wolffd@0
|
36 net.c = mix.centres;
|
|
wolffd@0
|
37
|
|
wolffd@0
|
38 % options(7) gives scale of function widths
|
|
wolffd@0
|
39 net = rbfsetfw(net, options(7)); |
