Jose Pimentel
Aug 31, 2018 · 1 min read

There is a second possible relation, where the output is “1” only if the first and the third column are “1”.

Under this second model (which is valid for the whole training set), the output would be “0”. I assume you’re basically just getting the average of these two possibilities, since we don’t have any real way of knowing which of the two are correct.