Week 6 ( (Fishers/)Linear Discriminant)

Question 1

3 main methods for classification

Answer 1

Question 2

Motivate decision boundary

Answer 2

Question 3

Geometry of decision boundary

Answer 3

Question 4

Show that W is orthogonal to vectors in the boundary

Answer 4

Question 5

Show that w0 determines the distance of the boundary from the original

Answer 5

Question 6

Show that y(x)/||W|| determines the distance of h the point x from the boundary

Answer 6

Question 7

Multi class classification decision boundary formulation

Answer 7

Question 8

Putting multi class classification decision boundary into matrix form

Answer 8

Question 9

Pros and cons of least square loss function for multi class linear classification

Answer 9

Question 10

What is Fisher’s Linear Discriminant

Answer 10

Binary classification method that projects high dimensional data onto a line y = W^TX to best separate 2 classes

Here y >= -w0 for class C1