SVM Regression

Question 1

How do we measure error with a tube regression?

Answer 1

There is no error within the tube, and the error outside the tube is measured by the distance to the tube

Question 2

What is ε in SVM regression?

Answer 2

The insensitive error. We ignore error as long as it is less than a given threshold value ε away from the decision boundary.

Question 3

What is the squared error?

Answer 3

Squared error calculates the error between all points and the regression function

Question 4

What is a pair of slack variables?

Answer 4

The distance from the upper bound and the lower bound

Question 5

What is the difference between Primal and Dual complexity for SVM Regression?

Answer 5

Primal has the variables w for each feature dim, so the complexity depends on the dimensions of the input space. Dual has ai and a*i for each data point, so the complexity depends on the number of support variables.