Mercurial > hg > camir-aes2014

annotate toolboxes/MIRtoolbox1.3.2/somtoolbox/som_autolabel.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 [sTo] = som_autolabel(sTo, sFrom, mode, inds)
wolffd@0 2
wolffd@0 3 %SOM_AUTOLABEL Automatical labeling, or clearing of labels.
wolffd@0 4 %
wolffd@0 5 % sTo = som_autolabel(sTo, sFrom, [mode], [inds])
wolffd@0 6 %
wolffd@0 7 % sM = som_autolabel(sM,sD);
wolffd@0 8 % sD = som_autolabel(sD,sM);
wolffd@0 9 % sM = som_autolabel(sM,sD,'vote',[5]);
wolffd@0 10 %
wolffd@0 11 % Input and output arguments ([]'s are optional):
wolffd@0 12 % sTo (struct) data or map struct to which the labels are put,
wolffd@0 13 % the modified struct is returned
wolffd@0 14 % sFrom (struct) data or map struct from which the labels are taken
wolffd@0 15 % [mode] (string) labeling algorithm: 'add' (the default), 'freq'
wolffd@0 16 % or 'vote'
wolffd@0 17 % [inds] (vector) the column-indexes of the labels that are to be
wolffd@0 18 % used in the operation (e.g. [2] would mean to use
wolffd@0 19 % only the second column of labels array in sFrom)
wolffd@0 20 %
wolffd@0 21 % The modes:
wolffd@0 22 % 'add': all labels from sFrom are added to sTo (even multiple
wolffd@0 23 % copies of same)
wolffd@0 24 % 'add1': only one instance of each label is kept
wolffd@0 25 % 'freq': only one instance of each label is kept and '(#)', where
wolffd@0 26 % # is the frequency of the label, is added to the end of
wolffd@0 27 % the label. Labels are ordered according to frequency.
wolffd@0 28 % 'vote': only the label with most instances is kept
wolffd@0 29 %
wolffd@0 30 % NOTE: The operations are only performed for the new labels.
wolffd@0 31 % The old labels in sTo are left as they are.
wolffd@0 32 % NOTE: all empty labels ('') are ignored.
wolffd@0 33 %
wolffd@0 34 % For more help, try 'type som_autolabel' or check out online documentation.
wolffd@0 35 % See also SOM_LABEL, SOM_BMUS, SOM_SHOW_ADD, SOM_SHOW.
wolffd@0 36
wolffd@0 37 %%%%%%%%%%%%% DETAILED DESCRIPTION %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
wolffd@0 38 %
wolffd@0 39 % som_autolabel
wolffd@0 40 %
wolffd@0 41 % PURPOSE
wolffd@0 42 %
wolffd@0 43 % Automatically label to map/data structs based on given data/map.
wolffd@0 44 %
wolffd@0 45 % SYNTAX
wolffd@0 46 %
wolffd@0 47 % sTo = som_autolabel(sTo, sFrom)
wolffd@0 48 % sTo = som_autolabel(sTo, sFrom, 'add')
wolffd@0 49 % sTo = som_autolabel(sTo, sFrom, 'freq')
wolffd@0 50 % sTo = som_autolabel(sTo, sFrom, 'vote')
wolffd@0 51 % sTo = som_autolabel(..., inds)
wolffd@0 52 %
wolffd@0 53 % DESCRIPTION
wolffd@0 54 %
wolffd@0 55 % This function automatically labels given map/data struct based on an
wolffd@0 56 % already labelled data/map struct. Basically, the BMU of each vector in the
wolffd@0 57 % sFrom struct is found from among the vectors in sTo, and the vectors in
wolffd@0 58 % sFrom are added to the corresponding vector in the sTo struct. The actual
wolffd@0 59 % labels to add are selected based on the mode ('add', 'freq' or 'vote').
wolffd@0 60 %
wolffd@0 61 % 'add' : all labels from sFrom are added to sTo - even if there would
wolffd@0 62 % be multiple instances of the same label
wolffd@0 63 % 'add1' : only one instance of each label is kept
wolffd@0 64 % 'freq' : only one instance of each label is kept and '(#)', where
wolffd@0 65 % # is the frequency of the label, is added to the end of
wolffd@0 66 % the label. Labels are ordered according to frequency.
wolffd@0 67 % 'vote' : only the label with most instances is added
wolffd@0 68 %
wolffd@0 69 % Note that these operations do not effect the old labels of sTo: they
wolffd@0 70 % are left as they were.
wolffd@0 71 %
wolffd@0 72 % NOTE: empty labels ('') are ignored.
wolffd@0 73 %
wolffd@0 74 % REQUIRED INPUT ARGUMENTS
wolffd@0 75 %
wolffd@0 76 % sTo (struct) data or map struct to which the labels are put
wolffd@0 77 % sFrom (struct) data or map struct from which the labels are taken
wolffd@0 78 %
wolffd@0 79 % OPTIONAL INPUT ARGUMENTS
wolffd@0 80 %
wolffd@0 81 % mode (string) The mode of operation: 'add' (default),
wolffd@0 82 % 'add1', 'freq' or 'vote'
wolffd@0 83 % inds (vector) The columns of the '.labels' field in sFrom to be
wolffd@0 84 % used in operation
wolffd@0 85 %
wolffd@0 86 % OUTPUT ARGUMENTS
wolffd@0 87 %
wolffd@0 88 % sTo (struct) the given data/map struct with modified labels
wolffd@0 89 %
wolffd@0 90 % EXAMPLES
wolffd@0 91 %
wolffd@0 92 % To label a trained map based on (labelled) training data, just do
wolffd@0 93 %
wolffd@0 94 % sM = som_autolabel(sM,sD);
wolffd@0 95 %
wolffd@0 96 % This operation is sometimes called "calibration" in the literature.
wolffd@0 97 % You can also do this the other way around: use a labelled map to
wolffd@0 98 % label a data set:
wolffd@0 99 %
wolffd@0 100 % sD = som_autolabel(sD,sM);
wolffd@0 101 %
wolffd@0 102 % If you only want a single instance of each label, use the 'freq' mode:
wolffd@0 103 %
wolffd@0 104 % sM = som_autolabel(sM,sD,'freq');
wolffd@0 105 %
wolffd@0 106 % If you already have labels in the struct, and want to perform 'freq' on
wolffd@0 107 % them, do the following:
wolffd@0 108 %
wolffd@0 109 % sMtemp = som_label(sM,'clear','all'); % make a map struct with no labels
wolffd@0 110 % sM = som_autolabel(sMtemp,sM,'freq'); % add labels to it
wolffd@0 111 %
wolffd@0 112 % The third mode 'vote' votes between the labels and only adds the one
wolffd@0 113 % which is most frequent. If two labels are equally frequent, one or the
wolffd@0 114 % other is chosen based on which appears first in the list.
wolffd@0 115 %
wolffd@0 116 % sM = som_autolabel(sM,sD,'vote');
wolffd@0 117 %
wolffd@0 118 % The lat argument is useful if you have specific labels in each column
wolffd@0 119 % of the '.labels' field. For example, the first column might be an
wolffd@0 120 % identifier, the next a typecode and the last a year. In this case, you
wolffd@0 121 % might want to label the map based only on the typecode:
wolffd@0 122 %
wolffd@0 123 % sM = som_autolabel(sM,sD,'vote',2);
wolffd@0 124 %
wolffd@0 125 % SEE ALSO
wolffd@0 126 %
wolffd@0 127 % som_label Give/remove labels from a map/data set.
wolffd@0 128 % som_bmus Find BMUs from the map for the given set of data vectors.
wolffd@0 129 % som_show Show map planes.
wolffd@0 130 % som_show_add Add for example labels to the SOM_SHOW visualization.
wolffd@0 131
wolffd@0 132 % Copyright (c) 1997-2000 by the SOM toolbox programming team.
wolffd@0 133 % http://www.cis.hut.fi/projects/somtoolbox/
wolffd@0 134
wolffd@0 135 % Version 1.0beta juuso 101297
wolffd@0 136 % Version 2.0beta juuso 101199
wolffd@0 137
wolffd@0 138 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
wolffd@0 139 %% check arguments
wolffd@0 140
wolffd@0 141 error(nargchk(2, 4, nargin)); % check no. of input args is correct
wolffd@0 142
wolffd@0 143 % sTo
wolffd@0 144 todata = strcmp(sTo.type,'som_data');
wolffd@0 145
wolffd@0 146 % sFrom
wolffd@0 147 [dummy m] = size(sFrom.labels);
wolffd@0 148
wolffd@0 149 % mode
wolffd@0 150 if nargin<3 | isempty(mode), mode = 'add'; end
wolffd@0 151
wolffd@0 152 % inds
wolffd@0 153 if nargin<4, inds = 1:m; end
wolffd@0 154 inds = inds(find(inds>0 & inds<=m));
wolffd@0 155
wolffd@0 156 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
wolffd@0 157 %% get a list of the labels to be added to each vector
wolffd@0 158
wolffd@0 159 % calculate BMUs
wolffd@0 160 if todata, bmus = som_bmus(sFrom,sTo,1);
wolffd@0 161 else bmus = som_bmus(sTo,sFrom,1); end
wolffd@0 162
wolffd@0 163 % for each vector in sTo, make a list of all new labels
wolffd@0 164 Labels = cell(size(sTo.labels,1),1);
wolffd@0 165 for d=1:length(bmus),
wolffd@0 166 m = bmus(d);
wolffd@0 167 if todata, t = d; f = m; else t = m; f = d; end
wolffd@0 168 if ~isnan(m),
wolffd@0 169 % add the labels
wolffd@0 170 for j=1:length(inds),
wolffd@0 171 if ~isempty(sFrom.labels{f,inds(j)}),
wolffd@0 172 Labels{t}{length(Labels{t})+1} = sFrom.labels{f,inds(j)};
wolffd@0 173 end
wolffd@0 174 end
wolffd@0 175 end
wolffd@0 176 end
wolffd@0 177
wolffd@0 178 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
wolffd@0 179 %% insert the labels to sTo
wolffd@0 180
wolffd@0 181
wolffd@0 182 if strcmp(mode,'add1') | strcmp(mode,'freq') | strcmp(mode,'vote'),
wolffd@0 183
wolffd@0 184 % modify the Labels array apprpriately
wolffd@0 185
wolffd@0 186 for i=1:length(Labels),
wolffd@0 187
wolffd@0 188 % calculate frequency of each label in each node
wolffd@0 189 new_labels = {};
wolffd@0 190 new_freq = [];
wolffd@0 191 for j=1:length(Labels{i}),
wolffd@0 192 if isempty(Labels{i}{j}), % ignore
wolffd@0 193 elseif ~any(strcmp(Labels{i}{j},new_labels)), % a new one!
wolffd@0 194 k = length(new_labels) + 1;
wolffd@0 195 new_labels{k} = Labels{i}{j};
wolffd@0 196 new_freq(k) = sum(strcmp(new_labels{k},Labels{i}));
wolffd@0 197 else, % an old one, ignore
wolffd@0 198 end
wolffd@0 199 end
wolffd@0 200
wolffd@0 201 % based on frequency, select label(s) to be added
wolffd@0 202 if length(new_labels) > 0,
wolffd@0 203 if strcmp(mode,'add1'),
wolffd@0 204 Labels{i} = new_labels;
wolffd@0 205 else
wolffd@0 206
wolffd@0 207 % sort labels according to frequency
wolffd@0 208 [dummy order] = sort(1./(1+new_freq));
wolffd@0 209 new_labels = new_labels(order);
wolffd@0 210 new_freq = new_freq(order);
wolffd@0 211
wolffd@0 212 switch mode,
wolffd@0 213 case 'freq',
wolffd@0 214 % replace each label with 'label(#)' where # is the frequency
wolffd@0 215 for j=1:length(new_labels),
wolffd@0 216 labf = sprintf('%s(%d)',new_labels{j},new_freq(j));
wolffd@0 217 new_labels{j} = labf;
wolffd@0 218 end
wolffd@0 219 Labels{i} = new_labels;
wolffd@0 220 case 'vote',
wolffd@0 221 % place only the one with most votes
wolffd@0 222 Labels{i} = {new_labels{1}};
wolffd@0 223 end
wolffd@0 224 end
wolffd@0 225 end
wolffd@0 226
wolffd@0 227 end
wolffd@0 228
wolffd@0 229 end
wolffd@0 230
wolffd@0 231 sTo = som_label(sTo,'add',[1:length(Labels)]',Labels);
wolffd@0 232
wolffd@0 233 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
wolffd@0 234
wolffd@0 235
wolffd@0 236
wolffd@0 237