annotate core/tools/machine_learning/svmlight2weight.m @ 0:e9a9cd732c1e tip

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 w = svmlight2weight(modelfile)
wolffd@0 2 % returns the dual weight vecor for a given svm model
wolffd@0 3 %
wolffd@0 4 % All the script does is compute the weighted sum of the support vectors
wolffd@0 5 % (first element in line is alpha*y, what follows is the feature vector).
wolffd@0 6 % For further info, see the comment in the model file for its format.
wolffd@0 7
wolffd@0 8 tagstr = python('svmlight2weight.py', modelfile);
wolffd@0 9
wolffd@0 10 [firstidx] = strfind(tagstr, '1 :');
wolffd@0 11
wolffd@0 12 %get start of vector description
wolffd@0 13 tagstr = tagstr(firstidx:end);
wolffd@0 14
wolffd@0 15 % parse text function output
wolffd@0 16 vecs = textscan(tagstr,'%d %f','Delimiter',':');
wolffd@0 17
wolffd@0 18 % initialise
wolffd@0 19 w = zeros(max(vecs{1}),1);
wolffd@0 20
wolffd@0 21 % set values
wolffd@0 22 w(vecs{1}) = vecs{2};
wolffd@0 23