Mercurial > hg > camir-aes2014

annotate core/magnatagatune/tests_evals/test_generic_display_results_absviolated.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 out = test_generic_display_results_absviolated(file)
wolffd@0 2 % out = test_generic_display_results_absviolated(file)
wolffd@0 3 %
wolffd@0 4 % displays the finalresults mat file and enables
wolffd@0 5 % further analysis and duiagnostics of the individual runs
wolffd@0 6
wolffd@0 7 global db_MTTAudioFeatureSlaney08;
wolffd@0 8 global db_magnamixedfeat_genrebasicsm;
wolffd@0 9 global db_MTTMixedFeatureGenreBasicSmPCA;
wolffd@0 10
wolffd@0 11
wolffd@0 12 global comparison;
wolffd@0 13 global comparison_ids;
wolffd@0 14
wolffd@0 15 if nargin < 1 || isempty(file) || isnumeric(file)
wolffd@0 16 u = dir();
wolffd@0 17 u = {u.name};
wolffd@0 18 [idx, strpos] = substrcellfind(u, '_finalresults.mat', 1);
wolffd@0 19
wolffd@0 20 if exist('file','var') && isnumeric(file)
wolffd@0 21 file = u{idx(file)};
wolffd@0 22 else
wolffd@0 23 file = u{idx(1)};
wolffd@0 24 end
wolffd@0 25 end
wolffd@0 26
wolffd@0 27 load(file);
wolffd@0 28
wolffd@0 29 % ---
wolffd@0 30 % % get statistics for feature parameters
wolffd@0 31 % Visualise the accuracy and variance
wolffd@0 32 % ---
wolffd@0 33 if isfield(out, 'inctrain')
wolffd@0 34 for i = 1:numel(out)
wolffd@0 35
wolffd@0 36 % ---
wolffd@0 37 % get training and test sample sizes
wolffd@0 38 % ---
wolffd@0 39 nData = [];
wolffd@0 40 n_train_data = zeros(1,numel(out(i).inctrain.dataPartition));
wolffd@0 41 n_test_data = zeros(1,numel(out(i).inctrain.dataPartition));
wolffd@0 42 for j = 1:numel(out(i).inctrain.dataPartition)
wolffd@0 43 n_train_data(j) = mean(out(i).inctrain.dataPartition(j).TrainSize);
wolffd@0 44 n_test_data(j) = mean(out(i).inctrain.dataPartition(j).TestSize);
wolffd@0 45 end
wolffd@0 46
wolffd@0 47 % ---
wolffd@0 48 % get lost percentages
wolffd@0 49 % ---
wolffd@0 50 mean_lost_test = 1 - [out(i).inctrain.mean_ok_test];
wolffd@0 51 mean_lost_test = mean_lost_test(1,:).* n_test_data;
wolffd@0 52
wolffd@0 53 var_ok_test = sqrt([out(i).inctrain.var_ok_test]);
wolffd@0 54
wolffd@0 55 mean_lost_train = 1 - [out(i).inctrain.mean_ok_train];
wolffd@0 56 mean_lost_train = mean_lost_train(1,:).* n_train_data;
wolffd@0 57
wolffd@0 58 % plot test results
wolffd@0 59 figure;
wolffd@0 60 subplot(2,1,1)
wolffd@0 61 plot(n_train_data, mean_lost_test,'r');
wolffd@0 62 hold;
wolffd@0 63 % plot training results
wolffd@0 64 plot(n_train_data, mean_lost_train,'m');
wolffd@0 65
wolffd@0 66 xlabel ('# training constraints');
wolffd@0 67 ylabel ('# constraints violated');
wolffd@0 68 legend ('test','training');
wolffd@0 69
wolffd@0 70 plot(n_train_data, mean_lost_test + var_ok_test(1,:).* n_test_data,'r:');
wolffd@0 71 plot(n_train_data, mean_lost_test - (var_ok_test(1,:).* n_test_data),'r:');
wolffd@0 72
wolffd@0 73 % ---
wolffd@0 74 % get percentage of unknown data examples learnt
wolffd@0 75 % ---
wolffd@0 76 lost_test_not_in_train = mean_lost_test - mean_lost_train;
wolffd@0 77 ntest_not_in_train = n_test_data - n_train_data;
wolffd@0 78
wolffd@0 79 lost_test_not_in_train = lost_test_not_in_train ./ ntest_not_in_train;
wolffd@0 80
wolffd@0 81 lost_test_not_in_train(isnan(lost_test_not_in_train)) = 0;
wolffd@0 82
wolffd@0 83 subplot(2,1,2)
wolffd@0 84 plot(n_train_data, lost_test_not_in_train);
wolffd@0 85
wolffd@0 86 xlabel ('# training constraints');
wolffd@0 87 ylabel ('% unknown constraints violated');
wolffd@0 88 end
wolffd@0 89 end
wolffd@0 90
wolffd@0 91
wolffd@0 92 % ---
wolffd@0 93 % write max. training success
wolffd@0 94 % ---
wolffd@0 95 mean_ok_test = [out.mean_ok_test];
wolffd@0 96 [val, idx] = max(mean_ok_test(1,:));
wolffd@0 97 fprintf(' --- Maximal training success: nr. %d, %3.2f percent. --- \n', idx, val * 100)
wolffd@0 98
wolffd@0 99 end