S. Merled, B. Caprile, and C. Furlanello
Giving Adaboost a Parallel Boost
Abstract
Adaboost is one of the most successful classification methods in
use. Differently from other popular ensemble methods (e.g., Bagging),
however, Adaboost is inherently sequential. In many data intensive,
real world applications this may represent a fatal limitation.
In this paper, a method is presented for the parallelization of the
Adaboost. The procedure builds upon earlier results concerning the
dynamics of Adaboost weights, and yields approximations to the
standard Adaboost models that can be easily and efficiently
distributed over a network of computing nodes.
Submitted.