changelog

MultiBoost 是c++实现的多类adaboost酸法。与传统的adaboost算法主要解决二类分类问题不同

MultiBoost 0.71:
NEW: Added output of the classification prediction.
FIXED: Bad bug which made the -d option useless.
CHANGED: Now multiple declarations of the same argument (with a different number of values) is allowed, and the error message is MUCH more meaningful!

MultiBoost 0.7:
NEW: AdaBoost.MH with real valued predictions
NEW: HaarMultiStumpLearner, which inherit from MultiStumpLearner.
NEW: Added Confusion Matrix output.
CHANGED: Now the data files can use any white characters for the separation of the columns. If
         a specific separator is needed, the option -d can be used (see multiboost -h io).
CHANGED: Several performance improvements.
FIXED: Fixed loading-related bugs.
FIXED: A lot of small and big bugs.. :)

MultiBoost 0.61:
NEW: Added confusion matrix
FIXED: Assert bug when using -test

MultiBoost 0.6:

NEW: Added Haar-like features and a weak learner (HaarSingleStumpLearner) that uses them.
(see Viola&Jones 2001 "Robust Real-time Object Detection" for more informations)

FIXED: Cleaned the code
FIXED: Rearranged part of the code
FIXED: New model for the parameters

MultiBoost 0.5:

First public release.