We consider the broad framework of supervised learning, where one gets examples of objects together with some labels (such as tissue samples labeled as cancerous or non-cancerous, or images of handwritten digits labeled ...