Boosting is an approach to machine learning based on the idea of creating a highly accurate predictor by combining many weak and inaccurate "rules of thumb." A remarkably rich theory has evolved around boosting, with connections to a range of topics, including statistics, game theory, convex optimiz[...]