linear regression

Question 1

What is supervised learning?

Answer 1

Learning from labelled examples.

Question 2

What is the key characteristic of supervised learning?

Answer 2

It uses input-output pairs to learn a mapping.

Question 3

What is unsupervised learning?

Answer 3

Learning patterns from unlabelled data.

Question 4

What is the goal of unsupervised learning?

Answer 4

To find structure in data.

Question 5

What is reinforcement learning?

Answer 5

Learning by trial and error using rewards.

Question 6

What makes reinforcement learning different?

Answer 6

It uses feedback from actions instead of fixed labels.

Question 7

What are the three components of a supervised learning model?

Answer 7

Model type, cost function, optimizer.

Question 8

What does the model type determine?

Answer 8

How predictions are made.

Question 9

What does the cost function do?

Answer 9

Measures prediction error.

Question 10

What is the role of the optimizer?

Answer 10

Updates parameters to reduce error.

Question 11

What is a parametric model?

Answer 11

A model with a fixed number of parameters.

Question 12

What is a non-parametric model?

Answer 12

A model that adapts its complexity to the data.

Question 13

What is the first step in training a model?

Answer 13

Initialize the model.

Question 14

What happens after making predictions?

Answer 14

Calculate the cost.

Question 15

When does training stop?

Answer 15

When a stopping condition is met.

Question 16

What kind of task is linear regression used for?

Answer 16

Regression (predicting continuous values).

Question 17

What is the simple linear regression formula?

Answer 17

h(x) = θ₀ + θ₁x

Question 18

What is the vector form of linear regression?

Answer 18

h(x) = θᵀx

Question 19

What is the purpose of the bias term in x?

Answer 19

To allow the model to shift the line vertically.

Question 20

What are the features in the gem price example?

Answer 20

Weight and purity.

Question 21

What is the output in the gem price example?

Answer 21

Price of the gem.

Question 22

How do we compute the prediction?

Answer 22

Multiply features by weights and add the bias.

Question 23

What is the cost function used in linear regression?

Answer 23

Mean Squared Error (MSE).

Question 24

What is the formula for MSE?

Answer 24

J(θ) = (1/2m) ∑(h(xᵢ) - yᵢ)²

Question 25

Why is there a 1/2 in the MSE formula?

Answer 25

To simplify differentiation.

Question 26

What does the cost function measure?

Answer 26

How far predictions are from actual values.

Question 27

What is gradient descent used for?

Answer 27

Minimizing the cost function.

Question 28

What is the gradient descent update rule?

Answer 28

θⱼ := θⱼ - α * ∂J/∂θⱼ

Question 29

What does the learning rate α control?

Answer 29

How big the update steps are.

Question 30

What happens if α is too small?

Answer 30

Slow learning.

Question 31

What happens if α is too large?

Answer 31

Overshooting or divergence.

Question 32

When do we stop gradient descent?

Answer 32

When cost stabilizes or after max iterations.

Question 33

What were the initial θ values in the example?

Answer 33

[2, 2, 2]

Question 34

What were the initial predictions in the example?

Answer 34

32, 106, 32

Question 35

How was cost calculated in the example?

Answer 35

Using MSE with predictions vs. actual prices.

Question 36

What was the result of the first update?

Answer 36

Parameters changed slightly to reduce cost.

Question 37

What are the key training steps?

Answer 37

Define hypothesis, choose cost, optimize.

Question 38

Why use vectorization in ML?

Answer 38

To speed up computation.

Question 39

What is the goal of training the model?

Answer 39

Minimize the cost function.