I am studying for an exam that is later this week and I was wondering if someone could help me with this practice question, or explain what I need to do?
For a binary classification problem, a Boolean function learner reduced the hypothesis space down to two possible candidate hypotheses: hA and hB. That is, both hypotheses are completely consistent with all the training examples they’ve seen. You test them on a test set of 30 instances not used for training. You find that hA got 25 of the test instances right, and hB got 27 of them right. How confident are we to say that hB is statistically different from hA? How confident are we to say that hB better than hA?
[link][3 comments]