Hello,
I have a question that may sound basic but I didn't find a way to understand it clearly.
I don't understand why it is problematic for adaboost. if we set 2 weak learners one horizontal and one vertical wouldn't it be okay?
Thanks
Hello,
I've added a visualization to make it clearer.
If we were to do it step by step, we would first draw the horizontal classifier which classifies the top as green and the bottom as red. Then we add the vertical classifier that classifies the right as green and the left as red. The bottom-left is always incorrect. The top-right is always correct. Either bottom-right or top-left will be correct, depending on which classifier has higher weight (this depends on your training set. In this case, since mine has the exact same number of greens and reds distributed equally, the two classifiers will have equal weight.)
In the end, with these two classifiers our accuracy is 50%. Therefore, Adaboost does not work.
Best,
Sena
It helped me a lot,
Thank you very much!
