Ensembles

Question 1

What is the main purpose of an ensemble of classifiers?

Answer 1

Improve predictive accuracy

Question 2

How accuracy different for an ensemble compared to classification?

Answer 2

If classification is performed by a single classifier, it should have high predictive accuracy
For ensembles it is not crucial that each classifier has high predictive accuracy

Question 3

An error of 0.4 is equal to what accuracy?

Answer 3

0.6

Question 4

What can Bagging reduce?

Answer 4

Classification error (it reduce variance)

Question 5

When is Bagging useful?

Answer 5

Useful when base classification algorithm is unstable (classifier is sensitive to small variations in training set)

Question 6

How does Boosting work?

Answer 6

Assign a weight to each training example and adaptively change these weights.

Question 7

In ensembles what suffers when improving accuracy?

Answer 7

Comprehensibility (more difficult to interpret an ensemble of classification models)