Machine Learning MCQs - 5 (Ensemble Models)

 Machine Learning MCQs - 5 
(Ensemble Models)

---------------------------------------------------------------------

1. The model which consists of a group of predictors is called a

1. Group
2. Entity
3. Ensemble
4. Set

Ans: 3

2. A Random forest is an ensemble of Decision Trees

1. True
2. False

Ans: 1

3. The steps involved in deciding the output of a Random Forest are

1. Obtain the predictions of all individual trees
2. Predict the class that gets the most votes
3. Both of the above

Ans: 3

4. A hard voting classifier takes into consideration

1. The probabilities of output from each classifier
2. The majority votes from the classifiers
3. The mean of the output from each classifier
4. The sum of the output from each classifier

Ans: 2

5. If each classifier is a weak learner, the ensemble can still be a strong learner?

1. True
2. False

Ans: 1

6. Ensemble methods work best when the predictors are

1. Sufficiently diverse
2. As independent from one another as possible
3. Making very different types of errors
4. All of the above

Ans: 4

7. To get diverse classifiers we cannot train them using different algorithms

1. True
2. False

Ans: 2

8. Training the classifiers in an ensemble using very different algorithms increases the chance that they will make very different types of errors, improving the ensemble’s accuracy

1. True
2. False

Ans: 1

9. When we consider only the majority of the outputs from the classifiers then it is called

1. Hard Voting
2. Soft Voting

Ans: 1

10. Soft voting takes into consideration

1. The majority of votes from the classifiers
2. The highest class probability averaged over all the individual classifiers

Ans: 2

11. In soft voting, the predicted class is the class with the highest class probability, averaged over all the individual classifiers

1. True
2. False

Ans: 1

12. Soft voting achieves higher performance than hard voting because

1. Majority votes classifications are often wrong
2. It gives more weight to highly confident votes
3. Finding majority is computationally expensive
4. This statement is false

Ans: 2

13. When sampling is performed with replacement, the method is

1. Bagging
2. Pasting

Ans: 1

14. When sampling is performed without replacement, it is called

1. Pasting
2. Bagging

Ans: 1

15. Both bagging and pasting allow training instances to be sampled several times across multiple predictors, but only bagging allows training instances to be sampled several times for the same predictor

1. True
2. False

Ans: 1

16. In bagging/pasting training set sampling and training, predictors can all be trained in parallel, via different CPU cores or even different servers

1. True
2. False

Ans: 1

17. To use the bagging method, the value of the bootstrap parameter in the BaggingClassifier should be set to

1. True
2. False

Ans: 1

18. Overall, bagging often results in better models

1. True
2. False

Ans: 1

19. With bagging, it is not possible that some instances are never sampled

1. True
2. False

Ans: 2

20. Features can also be sampled in the BaggingClassifier

1. True
2. False

Ans: 1

21. The hyperparameters which control the feature sampling are

1. max_samples and bootstrap
2. max_features and bootstrap_features

Ans: 2

22. Random forest is an ensemble of Decision Trees generally trained via ______

1. Bagging
2. Pasting

Ans: 1

23. We can make the trees of a Random Forest even more random by using random thresholds for each feature rather than searching for the best possible thresholds?

1. No
2. Yes, and these are called Extremely Randomised Trees ensemble

Ans: 2

24. If we look at a single Decision Tree, important features are likely to appear closer to

1. Leaf of the tree
2. Middle of the tree
3. Root of the tree

Ans: 3

25. One of the drawbacks of AdaBoost classifier is that

1. It is slow
2. It cannot be parallelized
3. It cannot be performed on larger training sets
4. It requires a lot of memory and processing power

Ans: 2

26. A Decision Stump is a Decision Tree with

1. More than two leaf nodes
2. Max depth of 1, i.e. single decision node with two leaf nodes
3. Having more than 2 decision nodes

Ans: 2

27. In Gradient Boosting, instead of tweaking the instance weights at every iteration like AdaBoost does, it tries to fit the new predictor to the residual errors made by the previous predictor.

1. True
2. False

Ans: 1

28. The learning_rate hyperparameter of GradientBoostingRegressor scales the contribution of each tree ?

1. True
2. False

Ans: 1

29. The ensemble method in which we train a model to perform the aggregation of outputs from all the predictors is called

1. Boosting
2. Bagging
3. Stacking
4. Pasting

Ans: 3