Figure $(3)$: Sample training data for boosting algorithm I. At the end of the first iteration, indicate on the figure the points that would have increased$/$higher weights. II$.$ During training, what is the lowest number of iterations that the model could need to attain zero error? III. Is it possible to add one more data point to the set in Figure $(3)$ so that the model can attain zero error in two steps? Figure $(3)$: Sample training data for boosting algorithm

I. At the end of the first iteration, indicate on the figure the points that would have increased$/$higher weights.

II$.$ During training, what is the lowest number of iterations that the model could need to attain zero error?

III. Is it possible to add one more data point to the set in Figure $(3)$ so that the model can attain zero error in two steps?

$2.$ Ensembles are successful in generating supervised learning systems with high accuracies. How do you comment on that? Why is it better to use a set of diverse base classifiers rather than using a single classifier? and does it matter whether to use weak learners or strong learners in boosting?