Consider the following algorithm for boosting a regression tree : (in attachment) a. Set and for all in the training set. b. For b = 1, 2, …, B , repeat: i. Fit a tree with d splits ( d + 1 terminal...

a. Set and for all in the training set.

b. For
b = 1, 2, …, B, repeat:

i. Fit a tree with
d
splits (d
+ 1 terminal nodes) to the training data (X, r)

ii. Update by adding a shrunken version of the new tree: