Supervised Learning Assessment

Back to Home

01. Supervised Learning Assessment

Back to Home

01. Supervised Learning Assessment

Why is the term "Naive" used in the Naive Bayes Classifier?

The resulting probabilities are too small, and need to be normalized in order to make the algorithm useful.

We are assuming the features are independent events, when they may not be so.

The algorithm's takes too long to train when the data is large.

SOLUTION:

We are assuming the features are independent events, when they may not be so.

What is an advantage of L1 regularization over L2 regularization?

It's much faster.

It works better when the data is dense, namely, when not many of the columns are zero.

It tends to use all the weights equally.

It tends to turn a lot of weights into zero, while leaving only the most important ones, thus helping in feature selection.

SOLUTION:

It tends to turn a lot of weights into zero, while leaving only the most important ones, thus helping in feature selection.

Which one of the following could reduce overfitting in a random forest classifier?

Increase the number of trees

Increase the depth of the trees

Decrease the threshold in the leaves

SOLUTION:

Increase the number of trees

Which models can be used for non-binary classification?(select all that apply)

Logistic Regression

Decision Trees

Random Forests

Linear Regression

Boosted Trees

SOLUTION:

Decision Trees
Random Forests
Boosted Trees

udacimak v1.4.0