#xgboost

Question 1: Single choice

You want to speed up training on many CPU cores. Which statement about parallelizing tree ensembles is correct?

Gradient boosting trains all its trees at once on separate cores, since each tree sees its own bootstrap sample1
Neither method can use more than one core, because tree growing is inherently sequential2
Gradient boosting adds trees one by one, so it parallelizes the work inside each tree, e.g. the split search3
A random forest cannot be parallelized because every tree depends on the previous one4