boostree 0.1

Xavier Carreras — Sat, 01 Dec 2007 03:16:14 -0000

This package provides an implementation Schapire and Singer's AdaBoost.MH for multi-label classification. As a main feature, the package provides decision-tree weak learning, a generalization of decision stumps that allows AdaBoost to learn conjunctions of original features. The implementation is specially suited for Natural Language classification taks, as it relies on sparse representations of examples and high-dimensional binary-feature spaces.

mloss.org boostree

boostree 0.1