Given a sparse matrix listing, what's the best way to calculate the cosine similarity between each of the columns (or rows) in the matrix? I would rather not iterate n-choose-two times. Say the input