Regression Modelling Quiz

One disadvantage of using the closed-form solution for linear regression is:

• It cannot find the exact minimum of the cost function

• It requires hyperparameter tuning like the learning rate

• It is computationally expensive and memory-intensive with large datasets

• It may get stuck in local minima for convex problems

Which of the following are disadvantages of using a closed-form solution for linear regression?

• It is computationally expensive for large datasets due to the need to compute the inverse of the feature matrix, ( high time complexity )

• It has limited applicability, as it only works for linear regression or models with a closed-form solution.
  
  

• It is memory-efficient and ideal for high-dimensional data.

• It requires storing the entire dataset in memory, making it problematic for high-dimensional data.

Which of the following is an advantage of gradient descent over a closed-form solution for linear regression?

• It provides an exact solution without iterations

• It is more suitable for high-dimensional data and large datasets

• It does not require setting a learning rate

• It always converges to the exact minimum in a single step

What is assumed about the errors made by a linear regression model?

• They are normally distributed with a mean equal to the mean squared error and a constant variance

• That the errors are the same for every data point

• That the errors are all zero

• They are normally distributed with a mean of zero and their variance is constant across the output range

Which of these scenarios would be suitable for modelling with multiple linear regression?

• Predicting a numeric output from a number of input variables

• Finding the average values of a set of multiple variables

• Predicting a categorical output from a number of input variables

• Plotting the relationship between two variables on a scatter graph

Which of these scenarios would be suitable for modelling with multiple logistic regression?

• Finding the average values of a set of multiple variables

• Plotting the relationship between two variables on a scatter graph

• Predicting a numeric output from a number of input variables

• Predicting a categorical output from a number of input variables

What quality makes the logistic function ideal for classification?

• Its output range is between zero and one, so it can be interpreted as a probability

• It is a non-continuous function

• It is able to output categorical values

• Its outputs are discrete

How do you treat categorical input variables for use in a regression model?

• Recode them using dummy variables or one-hot encoding

• You cannot use categorical inputs in a regression model

• Recode them with integers: 1, 2, 3, …, one for each class

• You don't need to do anything special, just include them as inputs

Which of these correctly describes Ridge Regression?

• Applies L1 regularization, which penalizes the squared size of the coefficients

• Applies L1 regularization, which penalizes the absolute size of the coefficients

• Applies L2 regularization, which penalizes the squared size of the coefficients

• Applies L2 regularization, which penalizes the absolute size of the coefficients

Which of these correctly describes Lasso Regression?

• Applies L2 regularization, which penalizes the absolute size of the coefficients

• Applies L2 regularization, which penalizes the squared size of the coefficients

• Applies L1 regularization, which penalizes the squared size of the coefficients

• Applies L1 regularization, which penalizes the absolute size of the coefficients