Machine Learning Quiz week 2

Typically, linear regression tends to underperform compared to k-nearest neighbor algorithms when dealing with high-dimensional input spaces.

• True

• False

Find the uni-variate regression function that best fits the dataset:

• f(x) = 1 × x + 4

• f(x) = 1 × x + 5

• f(x) = 1.5 × x + 3

• f(x) = 2 × x + 1

The dimensions of the design matrix used in linear regression with a dataset of 500 instances and 6-dimensional input are:

• 500 × 6

• 500 × 7

• 500 × 62

• None of the above

Assertion A: Binary encoding is preferred over One-hot encoding to represent categorical data. Reason R: Binary encoding is more memory efficient.

• Both A and R are true, and R is the correct explanation of A

• Both A and R are true, but R is not the correct explanation

• A is true, but R is false

• A is false, but R is true

Subset selection methods are more likely to improve test error by:

• Focusing on important features and reducing variance in fit

• Improving train error by focusing on the most important features

• Improving both test and train error

• Not helping performance

Rank the 3 subset selection methods in terms of computational efficiency:

• Forward stepwise, best subset, forward stagewise

• Forward stepwise, forward stagewise, best subset

• Best subset, forward stagewise, forward stepwise

• Best subset, forward stepwise, forward stagewise

Select the TRUE statements about ridge and lasso regression:

• Ridge regression makes the final fit more interpretable

• Lasso regression is easier to optimize

• Ridge has stable optimization

• Lasso is better for interpretability

Which statement is TRUE for principal components (PCA)?

• v = argmax ∑(v^T x_i)^2

• v = argmin ∑(v^T x_i)^2

• Scaling affects numerical stability, not final components

• PCA vectors vary with dataset scale

What does the design matrix represent in linear regression?

• Data rows by predictor columns

• Predictor rows by data columns

• Only data rows

• Only predictor rows

What is the main benefit of binary encoding over one-hot encoding?

• Simplicity

• Better model interpretability

• Memory efficiency

• No specific benefit