wolffd@0: function [num_compares] = get_comparison_stats()
wolffd@0: % [num_compares] = get_comparison_stats()
wolffd@0: % 
wolffd@0: % - retrieves number of comparisons for each pair of data clips
wolffd@0: % - extracts genre associations and statistics
wolffd@0: 
wolffd@0: global db_magnaclips;
wolffd@0: global comparison;
wolffd@0: global comparison_ids;
wolffd@0: 
wolffd@0: num_compares = sparse(numel(comparison_ids), numel(comparison_ids));
wolffd@0:  
wolffd@0: for i = 1:size(comparison,1)
wolffd@0:     abc = sort(comparison(i,1:3));
wolffd@0:     a = abc(1);
wolffd@0:     b = abc(2);
wolffd@0:     c = abc(3);
wolffd@0:     num_compares(a,b) = num_compares(a,b) +1;
wolffd@0:     num_compares(b,c) = num_compares(b,c) +1;
wolffd@0:     num_compares(a,c) = num_compares(a,c) +1;
wolffd@0: end
wolffd@0: 
wolffd@0: % ---
wolffd@0: % check for repeated pairs
wolffd@0: % ---
wolffd@0: % [i,j] = ind2sub(size(num_compares), find(num_compares > 1))
wolffd@0: % num = 300;
wolffd@0: % a = find((comparison(:,1) == i(num) | comparison(:,2) == i(num) | comparison(:,3) == i(num))...)
wolffd@0: % & (comparison(:,1) == j(num) | comparison(:,2) == j(num) | comparison(:,3) == j(num)))
wolffd@0: %
wolffd@0: % comparison(a,:)
wolffd@0: 
wolffd@0: 
wolffd@0: % ------------------------------------------------------------
wolffd@0: % compare genre frequency
wolffd@0: % ---
wolffd@0: clips = MTTClip(comparison_ids);
wolffd@0: 
wolffd@0: % comparison genres
wolffd@0: [genres, scores, gid] = clips.genres();
wolffd@0: 
wolffd@0: % all genres
wolffd@0: [genres2, scores2, gid2] = db_magnaclips.genredb.stats;
wolffd@0: 
wolffd@0: % ---
wolffd@0: % plot top genres
wolffd@0: % ---
wolffd@0: 
wolffd@0: figure;
wolffd@0: bar(scores(1:20))
wolffd@0: set(gca, 'XTick',1:20,'XTickLabel',genres(1:20))
wolffd@0: 
wolffd@0: figure;
wolffd@0: subplot(2,1,1)
wolffd@0: bar(scores(1:10) / max(scores))
wolffd@0: set(gca, 'XTickLabel',genres)
wolffd@0: title 'comparison subset'
wolffd@0: 
wolffd@0: subplot(2,1,2)
wolffd@0: bar(scores2(1:10) / max(scores2))
wolffd@0: set(gca, 'XTickLabel',genres2,'FontSize', 8)
wolffd@0: title 'Full MTT data set'
wolffd@0: 
wolffd@0: % ---
wolffd@0: % evaluate differences in distribution for all genres
wolffd@0: % shown are changes in relation to the whole database(genres2)
wolffd@0: %  ( relative to the summed tags )
wolffd@0: % ---
wolffd@0: genrediff = zeros(numel(genres2),1);
wolffd@0: for i=1:numel(genres2)
wolffd@0:     
wolffd@0: %     mgen = strcellfind(genres, genres2(i));
wolffd@0:     mgen = find(gid == gid2(i));
wolffd@0:     
wolffd@0:     if ~isempty(mgen)
wolffd@0:         % genrediff(i,1) = scores(mgen)/sum(scores) - scores2(i)/sum(scores2);
wolffd@0:         genrediff(i,1) = 1 - ( (scores2(i)/sum(scores2)) / (scores(mgen)/sum(scores)));
wolffd@0: 
wolffd@0:     else 
wolffd@0: %         genrediff(i) = inf;
wolffd@0:         error 'genre ids not consistent'
wolffd@0:     end
wolffd@0: end
wolffd@0: genrediff = genrediff * 100;
wolffd@0: 
wolffd@0: % visualise difference
wolffd@0: figure
wolffd@0: bar(genrediff);
wolffd@0: set(gca, 'XTick',1:44,'XTickLabel',genres2)
wolffd@0: axis([0 45 -200 100]);
wolffd@0: title 'relative loss for each genre considering comparison as an excerpt from MTT'
wolffd@0: ylabel 'loss(%)'
wolffd@0: 
wolffd@0: % ---
wolffd@0: % get distinc genre music sets:
wolffd@0: % we try the following sets of 'similar' genre tags
wolffd@0: % ---
wolffd@0: cgdb = db_magnaclips.genredb.subset(clips.id);
wolffd@0: 
wolffd@0: cids = cgdb.owner({'Classical', 'Baroque'}, 'or');
wolffd@0: cgdb = cgdb.exclude(cids);
wolffd@0: 
wolffd@0: eids = cgdb.owner({'Electronica', 'New Age', 'Ambient'}, 'or');
wolffd@0: cgdb = cgdb.exclude(eids);
wolffd@0: 
wolffd@0: rids = cgdb.owner({'Rock', 'Alt Rock', 'Hard Rock', 'Metal'}, 'or');
wolffd@0: cgdb = cgdb.exclude(rids);
wolffd@0: 
wolffd@0: % get reverse indexing for comparisons
wolffd@0: rev_compid = sparse(comparison_ids, 1, 1:numel(comparison_ids));
wolffd@0: 
wolffd@0: % 8 triples entirely classical 
wolffd@0: cfit = get_comparison_linfits(comparison, rev_compid(cids));
wolffd@0: 
wolffd@0: % 43 triples entirely electronic
wolffd@0: efit = get_comparison_linfits(comparison, rev_compid(eids));
wolffd@0: 
wolffd@0: % 6 triples entirely rock
wolffd@0: rfit = get_comparison_linfits(comparison, rev_compid(rids));
wolffd@0: 
wolffd@0: cgdb.stats
wolffd@0: 
wolffd@0: end
wolffd@0: % this function returns for each comparison line 
wolffd@0: % the num. of appearance of given selection of clip ids
wolffd@0: function out = get_comparison_linfits(comparison, goodset)
wolffd@0:     
wolffd@0:     out = zeros(1, size(comparison,1));
wolffd@0:     % for each line
wolffd@0:     for i = 1:size(comparison,1)
wolffd@0:     
wolffd@0:         out(i) = numel(intersect(comparison(i,1:3), goodset'));
wolffd@0:     end
wolffd@0: end
wolffd@0: 
wolffd@0: