In the AdaBoost algorithm, if the final strong classifier makes no mistakes on the training data, which of the following is correct?
You need to implement gradient boosting to generalize your model.
You would no longer have any weak learners in future iterations because all sample weights would be 0 for the next iteration.
You can test generalization of your final strong classifier model by only using the first subset of trees.
You have potentially overfitted to your training data.