NIPS
2004
14 years 6 months ago
2004
We study a method of optimal data-driven aggregation of classifiers in a convex combination and establish tight upper bounds on its excess risk with respect to a convex loss funct...