Basic errors

In order to evaluate one-class classifiers, one has to find out what the error of the first and second kind are (or the false positive and false negative rate). When you are only given target objects, life becomes hard and you cannot estimate these errors. But first assume you have some target and outlier test objects available. The false positive and false negative rates can be computed by dd_error:

  >> x = target_class(gendatb([50 0]),'1');
  >> w = gauss_dd(x,0.1);
  >> z = oc_set(gendatb(200),'1');
  >> e = dd_error(z,w)
  >> dd_error(z*w)   % other possibility
  >> z*w*dd_error    % other possibility

The first entry e(1) gives the false negative rate (i.e. the error on the target class) while e(2) gives the false positive rate (the error on the outlier class).

Question: can you imagine what would happen when you would replace oc_set in the third line by target_class?