|
Hello , everyone. Can anybody explain to me why AdaBoost algorithm work just for binary class problem? Thank You |
|
AdaBoost in its original, most basic form requires that there is always a weak learner capable of classifying slightly more than half the weighted instances correctly. When you have a multiclass problem, even getting 50% + epsilon accuracy can be non-trivial. So we have to choose the model which has error rate not more than 0.5, don't we? Btw, how Adaboost.M1 can handle multiclass problem? And what is the difference between AdaBoost.M1 with original AdaBoost. Thank you
(May 22 '12 at 00:43)
tiopramayudi
|
