“ Thus, sketches comprising just a few hundred values can be utilized to approximate the similarity of arbitrarily large datasets. ”