Decision Trees and Random Forests Quiz

Answers and explanations appear as you take the quiz.

1. How does a decision tree make a prediction?
   • By averaging all pixels in an image
   • By translating text between languages
   • By randomly guessing each time
   • By following a series of yes/no questions down to a leaf
2. What is the main weakness of a single, deep decision tree?
   • It cannot handle numbers
   • It tends to overfit and memorize noise in the training data
   • It can only output text
   • It requires a GPU to run
3. How does a random forest combine the predictions of its constituent trees?
   • By aggregating predictions, such as class votes or probabilities
   • By using exactly one tree and discarding every other output
   • By converting all trees into one fixed training row
   • By guaranteeing that all individual errors cancel
4. What sampling operation does bagging use when constructing conventional bootstrap training sets?
   • Sorting rows only by filename
   • Permanently deleting all numerical features
   • Drawing training rows with replacement
   • Assigning exactly one unique row to every tree
5. What metric do classification trees commonly use to choose a split?
   • Pixel brightness
   • Gini impurity or entropy
   • Audio frequency
   • Word count
6. Which named system is described in the paper A Scalable Tree Boosting System?
   • Photoshop
   • FFmpeg
   • Kubernetes
   • XGBoost
7. Which random-forest mechanism varies candidate input features at a split rather than resampling training rows?
   • Bootstrap row sampling alone
   • Alphabetical label sorting
   • Random feature-subset selection
   • A guarantee of identical splits in every tree
8. For a training row in a bootstrap forest, which trees should contribute to its out-of-bag prediction?
   • Trees whose bootstrap samples omitted that row
   • Only trees whose samples included that row at least twice
   • Every tree regardless of whether it trained on the row
   • A required newly trained tree using only that row