SVM

Question 1

Describe a way to extend the capabilities of a linear model which works which has as decision boundary limited capacity since it produces only lines or hyperplanes

Answer 1

Just add more features meaning adding polynomials of the features or interactions between them

Question 2

In numpy how to add a new feature aka column to an existing array

Answer 2

Simply use the method np.hstack(X,array to add)

Question 3

What the following method actually does?

plt.figure()

Answer 3

It creates a figure object, In this object we can add multiple axes, do other operations as resizing etc.

Question 4

What is a mask at numpy array?

Answer 4

Lets say we want to keep all the elements os an array that are bigger than 0.
For that reason what we are going to do is the following:
mask=y==0
The mask is a table tat takes values which are true or false.
Applying this mask to the original array, give as a result to keep only the values that are zero.
mask.

Question 5

When we have more than two features it is quite difficult to map all the interactions among them.For that reason, which are the method for the svms to pick the appropriate feature in higher dimensions?

Answer 5

a) Polynomial kernel which computes all possible polynomials upon a specific power
b) Radial basis function

Question 6

What is support vectors and why they are important?

Answer 6

Support vectors are the points that lie on the boundary between two classes. These points are important since they determine the boundary line between the classes.

Question 7

How a prediction is made at SVM?

Answer 7

It is done by measuring the distance between the points and the support vectors and the importance of the supporting vectors

Question 8

Which are the main parameters of SVM?

Answer 8

a) gamma
b) feature selection type (Gaussian kernel)
c) the C parameter