6 / 8 Perceptron & SVM

Turning the Dial

In soft-margin SVM training, as you increase the regularization parameter CC toward a very large value, what happens to the resulting margin and the model's tolerance for slack/misclassified points?

A

Neither the margin nor slack tolerance changes; CC only controls which kernel is used

B

A very large CC turns the SVM into an unregularized least-squares regression classifier

C

The margin narrows and slack is penalized more heavily, so the model tolerates fewer training errors, approaching hard-margin behavior

D

The margin widens and the model tolerates more slack, behaving more like a simple majority-vote classifier

E

The model ignores misclassified points entirely no matter how large CC gets