|
wolffd@0
|
1 function [relative, absolute, test_cn, train_cn] = sim_get_traintest_clip_overlap(datafile)
|
|
wolffd@0
|
2 %
|
|
wolffd@0
|
3 % get_traintest_clip_overlap(datafile)
|
|
wolffd@0
|
4 %
|
|
wolffd@0
|
5 % returns the percentage of overlapping constraints
|
|
wolffd@0
|
6 % with the corresponding training set for each test set
|
|
wolffd@0
|
7 %
|
|
wolffd@0
|
8 % how many percent of the test set are reappearing in the training set
|
|
wolffd@0
|
9
|
|
wolffd@0
|
10
|
|
wolffd@0
|
11 % simdata = load(datafile);
|
|
wolffd@0
|
12 if nargin < 1
|
|
wolffd@0
|
13 simdata = load('comp_partBinData_unclustered_cupaper_01');
|
|
wolffd@0
|
14 else
|
|
wolffd@0
|
15 simdata = load(datafile);
|
|
wolffd@0
|
16 end
|
|
wolffd@0
|
17 nTestSets = size(simdata.partBinTst, 2); % num cv bins
|
|
wolffd@0
|
18 ntrainsizes = size(simdata.partBinTrn, 2); % num increases of training
|
|
wolffd@0
|
19
|
|
wolffd@0
|
20
|
|
wolffd@0
|
21 absolute = zeros(nTestSets, ntrainsizes);
|
|
wolffd@0
|
22 relative = zeros(nTestSets, ntrainsizes);
|
|
wolffd@0
|
23 for k = 1:nTestSets % all test/training combinatios
|
|
wolffd@0
|
24
|
|
wolffd@0
|
25 % get clips of this test set
|
|
wolffd@0
|
26 test_clips = unique([simdata.partBinTst{k}(:,1); simdata.partBinTst{k}(:,2); simdata.partBinTst{k}(:,3)]);
|
|
wolffd@0
|
27 test_cn(k) = numel(test_clips);
|
|
wolffd@0
|
28 for m = 1:ntrainsizes
|
|
wolffd@0
|
29
|
|
wolffd@0
|
30 % get clips of this training set
|
|
wolffd@0
|
31 train_clips = unique([simdata.partBinTrn{k,m}(:,1); simdata.partBinTrn{k,m}(:,2); simdata.partBinTrn{k,m}(:,3)]);
|
|
wolffd@0
|
32
|
|
wolffd@0
|
33 % intersect both clip sets
|
|
wolffd@0
|
34 same = intersect(train_clips, test_clips);
|
|
wolffd@0
|
35
|
|
wolffd@0
|
36 % get stats
|
|
wolffd@0
|
37 absolute(k,m) = numel(same);
|
|
wolffd@0
|
38 relative(k,m) = absolute(k,m) / numel(test_clips);
|
|
wolffd@0
|
39 end
|
|
wolffd@0
|
40 train_cn(k) = numel(train_clips);
|
|
wolffd@0
|
41 end |
