6.034 Artificial Intelligence - Recitations, fall 2004 online slides on learning
	Information Theory and Decision Trees What is the correct classification? Before splitting, estimate of probabilities of possible answers calculated as proportions of positive and negative examples If training set has $p$ positive examples and $n$ negative examples, then the information contained in a correct answer is I((p/p+n), (n/p+n)) Splitting on a single attribute does not usually answer entire question, but it gets us closer How much closer? (Note that the search is greedy, i.e., hill-climbing)