Artificial Intelligence for Economics Nptel Week 3 Answers
Are you looking for Nptel Artificial Intelligence for Economics Week 3 Answers ? All weeks solutions of this Swayam course are available here.
Table of Contents
Nptel Artificial Intelligence for Economics Week 3 Answers (July-Dec 2025)
Course link: Click here to visit course on Nptel Website
Question 1. A node in a decision tree has 20 datapoints with 2 features. 12 of these points are from class A and 8 are from class B. If the dataset is split based on feature 1, we find two subpopulations: i) 8 from A, 4 from B, ii) 4 from A, 4 from B. If split according to feature 2, we have two subpopulations: i) 6 from A, 0 from B, ii) 6 from A, 8 from B. How will you proceed?
a) Split based on feature 2 because A dominates one sub-population, B dominates the other
b) Split based on feature 1 because both sub-populations are more balanced in size
c) Split based on feature 1 because it reduces size-weighted entropy
d) Split based on feature 2 because it reduces size-weighted entropy
These are Nptel Artificial Intelligence for Economics Week 3 Answers
Question 2. Which is of the following statements about ordinary linear regression, LASSO regression and ridge regression are wrong?
a) In all cases, least square loss function is used
b) For ordinary linear regression, the only aim is to minimize the loss function
c) LASSO regression reduces to ordinary linear regression in some cases
d) Ordinary regression can be solved analytically while LASSO and ridge regression can be solved numerically
These are Nptel Artificial Intelligence for Economics Week 3 Answers
Question 3. I want to predict the GDP growth rate in the next quarter. You have GDP growth values in the past quarters, and the values of some economic indicators in the past and present quarters. I also want to include an upper and lower bound of my estimates. What should be my approach?
a) Fit autoregressive models where upper and lower bounds of GDP are predicted based on past values of GDP growth
b) Estimate the mean value ‘m’ of GDP based on past values and present indicators, also estimate the standard deviation ‘s’ separately
c) Consider GDP growth as Gaussian random variable, estimate its mean and standard deviation from its past values
d) Train a neural network that predicts both upper and lower bounds as its output
These are Nptel Artificial Intelligence for Economics Week 3 Answers
Question 4. When do we prefer SVM or logistic regression over Neural Networks for classification?
a) When the number of training datapoints is too high
b) When the number of classes is very high
c) When we want a probability distribution over the classes
d) When we already have feature vectors which are known to be linearly separable
Question 5. A node in a decision tree has 20 datapoints with 2 features. 10 of these points are from class A and 10 are from class B. If the dataset is split based on feature 1, we find two subpopulations: i) 8 from A, 4 from B, ii) 2 from A, 6 from B. If split according to feature 2, we have two subpopulations: i) 4 from A, 0 from B, ii) 6 from A, 10 from B. How will you proceed?
a) Split based on feature 2 because it creates a homogeneous subpopulation
b) Split based on feature 1 because both sub-populations are more balanced in size
c) Split based on feature 1 because it reduces size-weighted entropy
d) Split based on feature 2 because it reduces size-weighted entropy
These are Nptel Artificial Intelligence for Economics Week 3 Answers
Question 6. How can we measure the Bayes Error of a Bayes classifier for binary classifier?
a) Area under the curve of p(X|C1), where C1 is the more likely class
b) Area under the curve of p(X|C2) where C2 is the less likely class
c) Area under the curve of p(C1|X), where C1 is the more likely class at each point X
d) Area under the curve of p(C2|X) where C2 is the less likely class at each point X
Question 7. If perceptron algorithm already gives us a linear classifier, why do we need SVMs?
a) SVM can offer both linear and non-linear classifiers
b) Perceptron gives us any linear classifier, but SVM gives us the best linear classifier
c) SVM is computationally simpler than perceptron
d) Perceptron often fails to converge
These are Nptel Artificial Intelligence for Economics Week 3 Answers
Question 8. Consider 5 observations with label A: [-4.4, -3.8, -3.1, -2.1, -1.6], and 5 more observations with label B: [0.7, 1.2, 1.9, 6.8, 10.7]. There is a test observation 0.2. For what value of K should K-NN classify it as A?
a) K = 1
b) K = 3
c) K = 5
d) K = 7
Question 9. Consider a set of observations [-4.4, -3.8, -2.7, -2.1, -1.6, 0.4, 1.2, 1.9, 7.1, 10.0]. First 5 points have label -1, next 5 have label +1. Which linear classifier do you prefer most?
a) Y=sign(x)
b) Y=sign(x+0.6)
c) Y=sign(x-0.6)
d) All of these are equivalent
Question 10. Which of the following is true about neural networks for image recognition?
a) They relieve us from the need to design image features
b) They are necessarily more accurate than other non-neural network approaches
c) They always use convolutions and poolings
d) All of the above
Click here for all nptel assignment answers
These are Nptel Artificial Intelligence for Economics Week 3 Answers