Hello, I'm trying to use svm as my classifier, and I tried using
train_svm.m and
test_svm.m instead of the
train_bp.m and
test_bp.m from the introductory tutorial.
However, I get an error because in running perfmet_maxclass, the acts argument is empty.
In fact, acts is just '' at that point.
After examining test_svm.m, I found out why. Following is taken from test_svm.m:
acts = zeros(size(testtargs));
sanity_check(testpats,testtargs,scratch);
test_max_idx(test_max_idx == 1) = 1;
test_max_idx(test_max_idx == 2) = -1;
% Now test generalization performance on the test data
[scratch.err, scratch.predictions] = svmclassify(testpats',test_max_idx',scratch.model);
Instead of setting acts as an nOuts x nTestTimepoints matrix that contains the activations of the output units at test as it is supposed to do,
test_svm.m simply does acts = '';
What needs to be done if I want to use test_svm.m right?
Thank you.