Question-17

You train a linear soft-margin SVM on a dataset with two features \(x_1\) and \(x_2\). If the dataset is linearly separable, which of the following is true about the margin?

• The margin increases as the regularization parameter \(C\) increases.

• The margin is independent of the regularization parameter \(C\).

• The margin increases as the regularization parameter \(C\) decreases.

• The margin decreases as the data points are moved farther from the decision boundary.

Answer

• The margin increases as the regularization parameter \(C\) increases.

• The margin is independent of the regularization parameter \(C\).

• The margin increases as the regularization parameter \(C\) decreases.

• The margin decreases as the data points are moved farther from the decision boundary.

Solution

As \(C\) decreases, the margin becomes wider, allowing the hinge-loss to keep growing.