Describe basic idea behind Maximum margin classifier.
separate linearly separable data with hyperplane.
What properties does this hyperplane have? What are the support vectors? Formulate quadratic program for finding the hyperplane. Is the solution for Max MC unique? Is the solution of quadratic program formulation unique?
Linearly inseparable What is a slack variable? Reformulate quadratic program using slack variables. Name hyperparameters for soft margin SVM with slack variable. What kind of loss should be used in soft margin SVM? Which loss provide convex upper bound?
Structural Risk minimization j0