wolffd@0
|
1 function [data, ndata1, ndata2, targets]=gen_data(ndata, seed)
|
wolffd@0
|
2 % Generate data from three classes in 2d
|
wolffd@0
|
3 % Setting 'seed' for reproducible results
|
wolffd@0
|
4 % OUTPUT
|
wolffd@0
|
5 % data : data set
|
wolffd@0
|
6 % ndata1, ndata2: separator
|
wolffd@0
|
7
|
wolffd@0
|
8 if nargin<1,
|
wolffd@0
|
9 error('Missing data size');
|
wolffd@0
|
10 end
|
wolffd@0
|
11
|
wolffd@0
|
12 input_dim = 2;
|
wolffd@0
|
13 num_classes = 3;
|
wolffd@0
|
14
|
wolffd@0
|
15 if nargin==2,
|
wolffd@0
|
16 % Fix seeds for reproducible results
|
wolffd@0
|
17 randn('state', seed);
|
wolffd@0
|
18 rand('state', seed);
|
wolffd@0
|
19 end
|
wolffd@0
|
20
|
wolffd@0
|
21 % Generate mixture of three Gaussians in two dimensional space
|
wolffd@0
|
22 data = randn(ndata, input_dim);
|
wolffd@0
|
23 targets = zeros(ndata, 3);
|
wolffd@0
|
24
|
wolffd@0
|
25 % Priors for the clusters
|
wolffd@0
|
26 prior(1) = 0.4;
|
wolffd@0
|
27 prior(2) = 0.3;
|
wolffd@0
|
28 prior(3) = 0.3;
|
wolffd@0
|
29
|
wolffd@0
|
30 % Cluster centres
|
wolffd@0
|
31 c = [2.0, 2.0; 0.0, 0.0; 1, -1];
|
wolffd@0
|
32
|
wolffd@0
|
33 ndata1 = round(prior(1)*ndata);
|
wolffd@0
|
34 ndata2 = round((prior(1) + prior(2))*ndata);
|
wolffd@0
|
35 % Put first cluster at (2, 2)
|
wolffd@0
|
36 data(1:ndata1, 1) = data(1:ndata1, 1) * 0.5 + c(1,1);
|
wolffd@0
|
37 data(1:ndata1, 2) = data(1:ndata1, 2) * 0.5 + c(1,2);
|
wolffd@0
|
38 targets(1:ndata1, 1) = 1;
|
wolffd@0
|
39
|
wolffd@0
|
40 % Leave second cluster at (0,0)
|
wolffd@0
|
41 data((ndata1 + 1):ndata2, :) = data((ndata1 + 1):ndata2, :);
|
wolffd@0
|
42 targets((ndata1+1):ndata2, 2) = 1;
|
wolffd@0
|
43
|
wolffd@0
|
44 data((ndata2+1):ndata, 1) = data((ndata2+1):ndata,1) *0.6 + c(3, 1);
|
wolffd@0
|
45 data((ndata2+1):ndata, 2) = data((ndata2+1):ndata,2) *0.6 + c(3, 2);
|
wolffd@0
|
46 targets((ndata2+1):ndata, 3) = 1;
|
wolffd@0
|
47
|
wolffd@0
|
48 if 0
|
wolffd@0
|
49 ndata = 1;
|
wolffd@0
|
50 data = x;
|
wolffd@0
|
51 targets = [1 0 0];
|
wolffd@0
|
52 end
|