Linear Models

Question 1

Definition of Linear (affine) functions and its hypothesis classes. Equivalent notation too.

Answer 1

2 / 2-4

Question 2

Definition of halfspace. When it is used?

Answer 2

2 / 5

Question 3

When data are linearly separable?

Answer 3

2 / 6 or 7 / 4

Question 4

Perceptron for halfspaces (describe the algorithm). When the algorithm stop?

Answer 4

2 / 7-8

Question 5

For Linear Regression what is hypothesis class, loss function and empirical risk

Answer 5

2 / 20-21

Question 6

In Linear regression, what is least squares? Also write the equivalent formulation RSS. What the acronym means?

Answer 6

2 / 22

Question 7

RSS in matrix form. How to find the solution of that minimizes RSS? What if the matrix is not invertible?

Answer 7

2 / 23-27 no 26

Question 8

What is feature normalization? Why is it important?

Answer 8

2 / 38

Question 9

Logistic regression : What it is? For what is used for? What is its hypothesis class?

Answer 9

2 / 39-41

Question 10

Logistic regression : What are the differences with halfspaces?

Answer 10

2 / 41

Question 11

Which is the loss function used in logistic regression models?

Answer 11

2 / 42-43

Question 12

What is the ERM problem for the Logistic regression? How can be solved?

Answer 12

2 / 44

Question 13

Only define what is the MLE?

Answer 13

2 / 45

Question 14

Describe the general approach to find the MLE? (NO part on logistic regression)

Answer 14

2 / 45

Question 15

Logistic regression and MLE, describe it. The MLE found at the end, is it similar to another approach we have studied?

Answer 15

2 / 46-47

Question 16

Coefficient of determination R^2 : definition and interpretation

Answer 16

2 / 31

Question 17

GD algortihm (pseudocode)

Answer 17

2 / 12

Question 18

Describe the stochastic gradient descent (SGD) algorithm (in general).

Answer 18

2 / 13

Question 19

What is the
main advantage of SGD with respect to the gradient descent algorithm?

Answer 19

2 / 15

Question 20

SGD for linear classification

Answer 20

2 / 16-18

Question 21

Compare the perceptron and the SGD perceptron. How can the SGD perceptron be speed up?

Answer 21

2 / 19