Adaboost package consists of two multi-class adaboost classifiers:
* AdaBoost_samme.m - a class implementing multi-class extension to classic adaboost.M1
algorithm (which was for two-class problems) which was first described in a paper
by Ji Zhu, Saharon Rosset, Hui Zou and Trevor Hastie, “Multi-class
AdaBoost”, January 12, 2006.
* AdaBoost_mult.m - solves same problem using a bank of two-class adaboost
classifiers. A three class problem will use three 2-class classifiers
solving class 1 vs. 2 & 3, class 2 vs. 1 & 3 and class 3 vs. 1 and 2
problems, than each sample is tested with each of the three classifiers
and class is assigned based on the one with the maximum score.
Boosting classifiers work by using a multiple "weak-learner" classifiers.
In this package we provide two weak-learner classifiers:
* decision_stump.m - a class implementing single node decision "tree".
* two_level_decision_tree.m - a class implementing three nodes in two
layers decision "tree" class.
Several helper functions:
* train_stump_2.m - fast low level decision stump function for 2-class problems
* train_stump_N.m - fast low level decision stump function for N-class problems
* save_adaboost_model.m - saves classifier to a CSV file
* load_adaboost_model.m - loads classifier from a CSV file
There are also four demo scripts:
* demo_adaboost_mult_with_decision_stumps.m - demo and testing of AdaBoost_mult classifier with decision_stump weak-learners
* demo_adaboost_mult_with_decision_trees.m - demo and testing of AdaBoost_mult classifier with two_level_decision_tree weak-learners
* demo_adaboost_sammy_with_decision_stump.m - demo and testing of AdaBoost_samme classifier with decision_stump weak-learners
* demo_adaboost_sammy_with_decision_trees.m - demo and testing of AdaBoost_samme classifier with two_level_decision_tree weak-learners
Create scripts with code, output, and formatted text in a single executable document.