Scaling and Renormalization in Statistical Learning