SVM

Question 1

• When a training set is linearly separable?
• What is the distance between a point and a hyperplane?
• What is a margin? What are support vectors?

Answer 1

…

Question 2

What is Hard-SVM? Equivalent formulations.

Equivalent formulation as quadratic optimization problem

(Equivalent formulation) Homogenous halfspaces

What is the dual problem for hard-SVM? When we use this notation?

Answer 2

…

Question 3

How Soft-SVM works?
Optimization problem, hinge formulation, homogeneus halfspaces.

Answer 3

…

Question 4

How to solve Soft-SVM?
What is gradient descent?
What is SGD?
How to use SGD to solve Soft-SVM?

Answer 4

…

Question 5

How we can use SVM if we have non linearly separable data? (slide 16.6)

What are the 2 issues generated by this procedure? How we can solve it?

Answer 5

…

First part of kernels_Basic paradigm

Question 6

What is a kernel function?

What is the kernel trick?

What are the most common kernels?

How do we choose the kernel?

What the mercer condition says?

Answer 6

…

Question 7

How to use SVM for regression?

What is the function to minimize?

What is the final model produced?

Answer 7

…