|
if we use a complex model for week learner in adaboost,does it increase the performance of adaboost? If yes then how? |
|
I haven't had very good experirence with improving the weak learners. I am not sure about theoretical dis-advantages of using a weak-learner compared to a strong learner. And certainly stumps as weak learners are better than a linear counterpart since you can build a non-linear classifier. However going with more complex classifiers might not give you a great advantage. |
