5

Question 1

What is the fundamental idea behind SVM

Answer 1

Fit the widest possible street between classes

Question 2

What is a Support Vector

Answer 2

An instance that is located on the street. The decision boundary is entirely determined by support vectors

Question 3

Why do inputs need to be scaled for SVM

Answer 3

If not scaled SVM will neglect small features

Question 4

Can SVM classifiers give a confidence score or probability score

Answer 4

Confidence - It can give the distance between the test instance and decision boundary

Probability - Found by applying a linear regression on the SVM scores

Question 5

Should you use primal or duel SVM when training on sets with hundreds of instances

Answer 5

If there are millions of instances primal should be sued as its faster

Question 6

What should you do with the Gamma and C hyper parameters of an SVM if it is underfitting

Answer 6

Increase both gamma and C to reduce regularzation