Suppose you’re using depth 0 decision trees, which simply return the weighted majority class of the data points as the classification, as the weak learner. Imagine that, at the first iteration, 80% of the data points were positive and 20% of the data points were negative. What would the cumulative weights of positive and negative data points be after one round of boosting)? From your result, do you think whether using depth 0 decision trees as weak learners is a good idea?