shark::RFTrainer< LabelType > Class Template Reference

Random Forest. More...

#include <shark/Algorithms/Trainers/RFTrainer.h>

Detailed Description

template<class LabelType>
class shark::RFTrainer< LabelType >

Random Forest.

Random Forest is an ensemble learner, that builds multiple binary decision trees. The trees are built using a variant of the CART methodology

Typically 100+ trees are built, and classification/regression is done by combining the results generated by each tree. Typically the a majority vote is used in the classification case, and the mean is used in the regression case

Each tree is built based on a random subset of the total dataset. Furthermore at each split, only a random subset of the attributes are investigated for the best split

The node impurity is measured by the Gini criteria in the classification case, and the total sum of squared errors in the regression case

After growing a maximum sized tree, the tree is added to the ensemble without pruning.

For detailed information about Random Forest, see Random Forest by L. Breiman et al. 2001.

Definition at line 72 of file RFTrainer.h.

The documentation for this class was generated from the following file: