Which of the following statements is true?

a. A bagging ensemble is more computationally expensive than a boosting ensemble.

b. An ensemble method requires a collection of strong, stable individual learners to be effective.

c. Random forests try to decorrelate their constituent regression trees by basing them on slightly different training sets and by randomly restricting the variables they can use to generate rules.

d. A boosting ensemble can be constructed in a parallel because the individual models do not rely on each other.