ML | Linear models | Priority

Question 1

What is linear regression? What do the terms p-value, coefficient, and r-squared value mean? What is the significance of each of these components?

Statistics q4 p9

Nouri 800

Answer 1

(See source material.) mu = sum_k[b_k*x_k]

Nouri 800

Question 2

What are the assumptions required for linear regression?

Statistics q5 p10

Nouri 800

Answer 2

(See source material.) MRLH pnemonic

Nouri 800

Question 3

What is linear regression?

ML q15 p42

Nouri 800

Answer 3

(See source material.) mu = sum_k[b_k*x_k]

ML q15 p42

Nouri 800

Question 4

What are the drawbacks of the linear model?

ML q16 p43

Nouri 800

Answer 4

(See source material.)

ML q16 p43

Nouri 800

Question 5

Say you are running a multiple linear regression and believe there are several predictors that are correlated. How will the results of the regression be affected if they are indeed correlated? How would you deal with this problem?

Google: Linear regression with correlated predictors

Answer 5

(See source material.)

Google: Linear regression with correlated predictors

Question 6

Equation for correlation coefficient r.

Machine Learning with PyTorch and Scikit-Learn Chapter 9 p276

Answer 6

(See source material.)

Machine Learning with PyTorch and Scikit-Learn Chapter 9 p276

Question 7

How does naive Bayes assign a class c to a document d (basic equation)? Why is this an example of a generative model?

Answer 7

(See source material.) Eq. (5.1). Uses a likelihood term, expresses how to generate the features of a document if we knew it was of class c.

Jurafsky SLP3E 5. Logistic regression

Question 8

3 equations to standardize input features.

Answer 8

(See source material.) Eq. (5.8)

Jurafsky SLP3E 5. Logistic regression

Question 9

Equation to normalize input features.

Answer 9

(See source material.) Eq. (5.9)

Jurafsky SLP3E 5. Logistic regression

Question 10

Equation for softmax.

Answer 10

(See source material.) Eq. (5.15)

Jurafsky SLP3E 5. Logistic regression: 5.2.4 Choosing a classifier

Question 11

Equation for L2 regularized objective.

Answer 11

(See source material.) Eq. (5.37)

Jurafsky SLP3E 5. Logistic regression: 5.2.4 Choosing a classifier

Question 12

Equation for L1 regularized objective.

Answer 12

(See source material.) Eq. (5.39)

Jurafsky SLP3E 5. Logistic regression: 5.2.4 Choosing a classifier

Question 13

Equation for a Gaussian prior on weights. How does this relate to regularization?

Answer 13

(See source material.) Eq. (5.40). L2 regularization corresponds to assuming that weights are distributed according to a Gaussian distribution with mean m = 0.

Jurafsky SLP3E 5. Logistic regression: 5.2.4 Choosing a classifier

Question 14

Multinomial logistic regression: The loss function for a single example x

Answer 14

Sum of the logs of the K output classes, each weighted by their probability yk (Eq. 5.44). This turns out to be just the negative log probability of the correct class c (Eq. 5.45). (See source material.)

Jurafsky SLP3E 5. Logistic regression: 5.2.4 Choosing a classifier