B.

TECH/CSE/6TH SEM/CSEN 3233/2022

B.TECH/CSE/6TH SEM/CSEN 3233/2022

MACHINE LEARNING
(CSEN 3233)

Time Allotted : 3 hrs Full Marks : 70

Figures out of the right margin indicate full marks.
Candidates are required to answer Group A and
any 5 (five) from Group B to E, taking at least one from each group.
Candidates are required to give answer in their own words as far as practicable.

Group – A
(Multiple Choice Type Questions)

1. Choose the correct alternative for the following: 10 × 1 = 10

(i) An artificial neuron receives n inputs x1, x2, ........, xn with weights w1, w2, .........., wn
attached to the input links. The weighted sum __________ is computed to be passed
on to a non-linear filter Φ called activation function to release the output.
(a) Σ wi (b) Σ xi (c) Σ wi + Σ xi (d) Σ wi* xi
(ii) Which of the following methods do we use to best fit the data in Logistic
Regression?
(a) Least Square Error (b) Maximum Likelihood
(c) Jaccard distance (d) Both (a) and (b).
(iii) When a model performs well on training data (the data on which the algorithm
was trained) but does not perform well on test data (new or unseen data), we
say that the model is
(a) Overfitting (b) Generalizing (c) Regularizing (d) None of the above.
(iv) Which of the following statement(s) is / are true for Gradient Decent (GD) and
Stochastic Gradient Decent (SGD)?
1. In GD and SGD, you update a set of parameters in an iterative manner to
minimize the error function.
2. In SGD, you have to run through all the samples in your training set for a
single update of a parameter in each iteration.
(a) Only 1 (b) Only 2 (c) Both 1 and 2 (d) None of 1 and 2.
(v) What is generalization?
(a) The ability of a pattern recognition system to approximate the desired
output values for pattern vectors which are not in the test set.
(b) The ability of a pattern recognition system to approximate the desired
output values for pattern vectors which are not in the training set.
(c) Can be either way
(d) None of the mentioned.
CSEN 3233 1
B.TECH/CSE/6TH SEM/CSEN 3233/2022
(vi) Perceptron cannot learn
(a) AND (b) XOR (c) both (a) and (b) (d) none of these.
(vii) “Convolutional networks have generally more parameters than their equivalent
fully connected networks”- This Statement is ______.
(a) always true (b) always false (c) mostly true (d) mostly false
(viii) The back-propagation algorithm learns a globally optimal neural network with
hidden layers.
(a) Always True (b) Always False (c) Mostly True (d) Mostly False.
(ix) Suppose your model is overfitting. Which of the following is NOT a valid way to
try and reduce the overfitting?
(a) Increase the amount of training data
(b) Improve the optimization algorithm being used for error minimization
(c) Decrease the model complexity
(d) Reduce the noise in the training data.
(x) Which of the following gives non-linearity to a neural network.
(a) Convolution operator (weighted sum of the inputs)
(b) Stochastic gradient descent
(c) Sigmoid activation
(d) Non-zero bias.

Group- B
2. (a) Discuss with example the in-sample error and out-of-sample error.
[(CO2)(Remember/LOCQ)]
(b) Briefly explain error and noise with example. [(CO3)(Understand/LOCQ)]
(c) Classes attended by 10 students Table 1
in machine learning and marks Sl No Attendance Marks
obtained in the examination are 1 28 43
provided in the Table 1. 2 27 39
Estimate the marks a student 3 23 27
may obtain in the examination 4 27 36
when she attended 20 classes, 5 24 34
using linear regression. 6 28 39
[(CO4)(Analyse/IOCQ)] 7 26 36
8 21 36
9 22 31
10 28 37
4 + 6 + 2 = 12

3. (a) Derive the linear regression formula for single dependent variables.
[(CO4)(Remember/LOCQ)]
(b) What is error in prediction? Explain, with appropriate examples, three different
error measures used to deal with different prediction problems.
[(CO5)(Describe/HOCQ)]
CSEN 3233 2
B.TECH/CSE/6TH SEM/CSEN 3233/2022
(c) Write the Perceptron Learning Algorithm (PLA) and briefly explain the working
principle of the algorithm. [(CO2)(Analyze/IOCQ)]
4 + 4 + 4 = 12

Group - C
4. (a) Explain the importance of VC dimension in machine learning?
[(CO4)(Remember/HOCQ)]
(b) Find the VC dimension for the following hypotheses:
(i) Positive intervals F(x) = +1 for a ≤ x ≤ b; -1 otherwise.
(ii) Perceptron in R2. [(CO2)(Analyze/LOCQ)]
(c) Explain the Bias-Variance trade off in the context of learning.
[(CO3)(Analyze/IOCQ)]
3 + 3 + 6 = 12

5. (a) Define Dichotomy, Growth function and break point.

[(CO3)(Understand/LOCQ)]
(b) Calculate growth function and break point for positive intervals for N points.
[(CO4)(Analyze/IOCQ)]
(c) You are given 3 points X1, X2 and X3. Calculate the number of dichotomies when
(i) break point is 3 and (ii) break point is 2. [(CO6)(Apply/HOCQ)]
3 + 3 + 6 = 12

Group - D
6. (a) What does the learning rate do in Back-Propagation training?
[(CO4)(Remember/LOCQ)]
(b) Describe what is likely to happen when a learning rate is used that is too large,
and when one is used that is too small. How can one optimize the learning rate?
[(CO3)(Understand/LOCQ)]
(c) Describe the importance of using bias in neural network.
[(CO2)(Remember/LOCQ)]
(d) Explain the main reasons why a Back-Propagation training algorithm might not
find a set of weights which minimizes the training error for a given feed-forward
neural network. [(CO1)(Analyze/IOCQ)]
(e) Explain the purpose of the momentum term that is often included in the Back-
Propagation learning algorithm. [(CO1)(Analyze/IOCQ)]
2 + 3 + 2 + 3 + 2 = 12

7. (a) An input of volume 48×48×3 is fed to a Convolutional Neural Network. What

would be the output volume when you apply?
(i) Four 5×5×3 filters with stride 1 without padding.
(ii) Three 5×5×3 filters with stride 2 and a zero (0) padding of size 1.
[(CO3)(Understand/IOCQ)]
(b) What is the difference between Recurrent Neural Network and Convolutional
Neural Network? [(CO1)(Remember/LOCQ)]

CSEN 3233 3
B.TECH/CSE/6TH SEM/CSEN 3233/2022
(c) You are asked to simulate the Boolean function x1 ∧ x2 ∨ (￢𝑋3) using a multi-
layer perceptron. Construct the network and explain how your network is able
to model the said function. [(CO1)(Analyze/IOCQ)]
(2 + 2) + 2 + 6 = 12

Group - E
8.

Fig.1
(a) Support vector machines learn a decision boundary leading to the largest margin
from both classes. You are training SVM on a tiny dataset with 4 points shown in
Fig.1. This dataset consists of two examples with class label -1 (denoted with
plus), and two examples with class label +1 (denoted with triangles).
(i) Find the weight vector w and bias b. What’s the equation corresponding to
the decision boundary?
(ii) Circle the support vectors and draw the decision boundary.
[(CO4)(Analyse/HOCQ)]
(b) What is overfitting? How to deal with this situation?
[(CO4)(Remember/IOCQ)]
(c) What is weight decay? [(CO4)(Remember/LOCQ)]
(3 + 1) + (2 + 4) + 2 = 12

9. Write short notes on any three of the followings: (4 × 3) = 12

(i) Logistic Regression
(ii) Kernel trick of SVM
(iii) Regularization in Machine Learning
(iv) Validation in Machine Learning. [(CO1)(Remember/IOCQ)]

Cognition Level LOCQ IOCQ HOCQ

Percentage distribution 32.29 19.79 47.92

Course Outcome (CO):

After the completion of the course students will be able to

1. Learn and understand the basics of machine learning approaches and paradigm.
2. Understand and describe various machine learning algorithms.
3. Understand complexity of Machine Learning algorithms and their limitations.
CSEN 3233 4
B.TECH/CSE/6TH SEM/CSEN 3233/2022

4. Mathematically Analyse various machine learning approaches and paradigms

5. Analyse various machine learning techniques to get an insight of when to apply a
particular machine learning approach.
6. Apply common Machine Learning algorithms in practice and implementing their own
using real-world data.

*LOCQ: Lower Order Cognitive Question; IOCQ: Intermediate Order Cognitive Question;
HOCQ: Higher Order Cognitive Question.

CSEN 3233 5
B.TECH/CSE/6TH SEM/CSEN 3233/2022

CSEN 3233 6