User talk:2601:645:8004:4C00:3491:F717:9A95:80FB

I think the equation for q_ij is wrong. The sum under the fraction should be over all pairs of y_l and y_k with l not equal to k. The way it's written here isn't symmetric in i and j. See http://www.jmlr.org/papers/volume9/vandermaaten08a/vandermaaten08a.pdf