GBDT Compared with the random forest algorithm, which of the following statements is wrong?

Question

Options

• AGBDT Algorithm is easier to underfit than random forest
• BRandom forest is calculated in parallel, while GBDT Can't
• CGBDT Algorithm is easier to overfit than random forest
• DGBDT And random forest are built on CART Based on the tree

• A
  14% (4)
• B
  4% (1)
• C
  75% (21)
• D
  7% (2)

Explanation

Option C is the wrong statement because GBDT does not necessarily overfit more easily than Random Forest - in fact, GBDT incorporates built-in regularization (a learning rate/shrinkage, deliberately shallow weak learners, and optional row subsampling) that actively limits overfitting; Random Forest, which grows deep unpruned CART trees individually, can have higher per-tree variance even though bagging partially corrects this at the ensemble level.

Why the distractors are correct (not wrong):

• A is true: GBDT's weak learners are shallow and require many sequential iterations to build expressive power, making early-stopped GBDT more susceptible to underfitting than Random Forest's deep, fully-grown trees.
• B is true: Random Forest trees are independent of each other and can be trained in parallel; GBDT trees are strictly sequential - each tree depends on the residuals of the previous one.
• D is true: both algorithms use CART (Classification and Regression Trees) as their base learner.

Memory tip: Think of the contrast as slow and careful vs. fast and noisy - GBDT corrects mistakes gradually with a small learning rate (regularized → lower overfit risk), while Random Forest casts a wide net with deep trees that individually overfit, relying on the crowd to average out the noise.

No community discussion yet for this question.