50% found this document useful (2 votes)
1K views7 pages

Deep Learning - Question Papers

The document is an examination paper for a Deep Learning course at KIET Group of Institutions, covering various topics such as neural networks, machine learning, and applications of deep learning. It includes multiple sections with questions that require definitions, explanations, and comparisons of concepts like CNN, GAN, PCA, and LSTM. The exam is structured to assess students' understanding of both theoretical and practical aspects of deep learning within a specified duration.

Uploaded by

akshatkishorem
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
50% found this document useful (2 votes)
1K views7 pages

Deep Learning - Question Papers

The document is an examination paper for a Deep Learning course at KIET Group of Institutions, covering various topics such as neural networks, machine learning, and applications of deep learning. It includes multiple sections with questions that require definitions, explanations, and comparisons of concepts like CNN, GAN, PCA, and LSTM. The exam is structured to assess students' understanding of both theoretical and practical aspects of deep learning within a specified duration.

Uploaded by

akshatkishorem
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 7

KIET Group of Institutions

(Department of CompUtcr Science & Engineering)


CSE/ B.Tech Semester
Pre-University Examination. (2021-22)ODD Semester
Deep Learning (KCS078)
Duration: 3 hrs Max. Marks: 100
Note: -Attempt all the Questions.
Section-A IOX2=20
Q. No. Question Marks co
a Define neural network. COI I/F
b Explain how deep Learning is more efficient than ML in case of feature extraction? COI
c Discuss the difference between Shallow network and Deep Neural network. C02
d State any two practical applications of GAN in real life. C02
e Discuss in brief how LDA differs from PCA C03
f Discuss relationship of metric learning to deep learning. C03 I/F
g What is dimensionality reduction ? C04 I/F
State the concept of LSTM. coa
i What is NLP ? cos
cos 2/p
j Discuss bioinformatics with respect to its relation with Deep learning.
Section-B
Q. No. Question Marks co
Discuss in brief about Universal Function Apäroximation
OR COI 2/c
Write a short note on logistic regression.
Explain the probabilistic theory of Deep Learning.
OR C02 4/c
Explain the concept of Regularization. Explain its different types.
Explain the various hyper parameters required for deep learning optimization.
OR C03
Explain the various techniques of weights initialization in deep learning.
Illustrate the concept of deep reinforcement learning.
OR C04
Illustrate the concept of Artificial Neuroscience.
Analyze the face recognition mechanism in deep learning using some case study.
OR cos
Analyze the NLP mechanism in deep learning using some case study.
Section-C (5X10=50)
Question Marks co
Q. No.

o. 0663

10 COI

neural network. Assume learning rate be


Apply BPN for a single iteration on the following
0.8

Relate SVM, Logistic regression, ANN and CNN ?


Justify the concepts and use of GAN in Al domain? 10 C02
OR
the
knowledge, skill set that a student attains after completing
CO -Course Outcome generally refer to traits,
course successfully.
Justify the conce ts and use of Semi Su ervised learning in Al domain.
Consider the two dimensional patterns (5, 6), (6, 7), (7, 8).Compute the principal
component using PCA Algorithm. 4/p
9 10 cog
OR
Compare PCA and LDA. How do LDA works for dimension reduction.
Design a recurrent networks for time series prediction, elaborating it's architecture as
well. 10 C04 4/p
10 OR
Explain LSTM network and elaborating it's architecture as well.
Analyze the Wave Net deep generative model of raw audio waveforms.
OR 10 cos
11
Analyze the Word2vec technique for NLP.
KIETGroup of Institutions. Delhi - NCR.Ghaziabad

Dc artmcnt of Com uter Science and En ineerin


B. Tech. in Computer Science andEngineering 47t Semester
CT-I. (2021-22)Odd Semester
Coursc:l)ccp Learning (KCS 078)
Duration: 2hrs Max. Marks: 60

Section-A
Attempt all the questions of this section (2X10=20)
. No. ucstion Marks co BL/ KC
a Discuss the role of bias in ANN? 2 COI
b Ex lain the conce t of Machine Learnin ? 2 COI
c What are su ort vectors in SVM Al orithm? 2 co I
d Define feed forward Multila er Perce tron. 2 COI
e Outline the demerits of ML. 2 co 1
1.
f Discuss the difference between Shallow network and Dee Neural network. 2 C02
Define SVM Kernels? Where it is used ? 2 co I
h Define artificial neural network. 2 co 1
i State the need of activation functions in neural networks? 2 COI
Discuss the role of dro out in overfittin 2 C02
Section-B
Attempt all the questions of this Section(5X4=20)
. No. uestion Marks co BW KC
Com are and Contrast Al, ML and DL.
2 OR 5 COI 4/C
Classi the various learning strateoy of ML.
Correlate RNN throu h Multila er Perce tron?
3 OR 5 co I 4/C
Ex lain loss functions? Illustrate about various loss functions.
Discuss Re ularization. Ex lain its different ty es.
4 OR 5 C02
Elaborate the robabilistic theo of Dee Learnin
Discuss Universal Function A roximation theorem for neural network.
5 OR 5 COI
Describe SVM and it's limitations
Section-C
Attem t all the uestions of this Section 10X2=20
Q. No. Question Marks co BW KC
6
0.4
042

—092

6 10 C02
0.2

In ut la Or Hidden layer Output layer

Apply Forward Propagation and Backward propagation for a single iteration on the
followin neural network. Assume learnin rate be 0.6
OR
Illustrate the differences among Gradient Descent Algorithm, Stochastic Gradient
Descent Learnin , Mini-batch radient descent
Apply perceptron Learning Algorithm for AND function having two inputs and a bias
denoted as Xo —— I Assume the learning rate be 0.1 and the initial weights be

7
0.5 0.5 0.5 10 COI

OR
Apply Logistic Regression using sigmoid function through a suitable example for
classification roblem.

• CO —CourseOutcome, generally, refer to traits, knowledge,skill set that a student attains after completing
the course
successfully.
• BL—Asper Revised Bloom's Taxonomy, Bloom's Levels (BLS) are the cognitive process levels
viz. 1. Remember, 2.
Understand, 3. Apply, 4. Analyze 5. Evaluate and 6. Create
• KC —As per Revised Bloom's Taxonomy, Knowledge Categories (KCs) are F-Factual,
C-Conceptual, P—Procedural, M
Metacognitive
KIETGroup of Institutions. Delhi - NCR.Ghaziabad
(Roll Numbnr•. I Q

De artmcnt of Com uter Science and En ineerin


B. Tech. 7t 'Semester
CT-2. (2021-22)OddSemester
Coursc:Deep Learning (KCS 078)
Duration: 2hrs Max. Marks: 60

Section-A
Attempt all the questions of this section (2X10=20)
. No. uestion Marks co BL/ KC
a Discuss the role of CNN? 2 C02
b Ex lain the conce t of GAN? 2 C02
c What do ou understand b the term dimensionalit reduction? 2 C03
d State an two uses of Semi su ervised learnin in real life. 2 C02
e Outline the merits of Semi su ervised learnin 2 C02
1.
f Discuss the difference between Shallow network and Convolution Neural network. 2 C02
Define Generative modelin ? Where it is used ? 2 C03
h Define LDA? 2 C02
i State the need of Mahalanobis distance in metric learnin 2 C03
Discuss the role of central oint in LDA. 2 C03
Section-B
Attempt all the questions of this Section(5X4=20)
Q• No. Question Marks co BV KC
Ex lain Batch Normalization with a suitable exam
2 OR 5 C03
Classi the various es of autoenocoders.
Ex lain the role of of max oolin and min oolin in CNN.
OR
3
Explain VC dimension with example? Illustrate about various applications of VC
5 C02 4/C
dimension.
Ex lain Generative Model in GAN
4 OR 5 C02
Ex lain the role of convolution la er in CNN.
Discuss the role of auto encoders and it's com onents with a suitable dia ram.
5. OR 5 C03
Elaborate the conce t of metric learnin in reference to dee learnin
Section-C
Attem t all the uestions of this Section 10X2=20
. No. uestion Marks co BW KC
Given below is an input matrix named I, kernel matrix, calculate the Convoluted matrix
C using stride =l also apply max pooling on C.

Input Matrix I
0 0 0
o O Kernel Matrix
1 t o
6 o t o o o
10 C02
1 1
1
1 o O
0

OR
Illustrate the various steps involved in CNN for feature learning and classification of
an ical ima e. Elaborate with a suitable dia ram.
(4, 3), (5, 6). Compute the principal
Consider the two dimensional patterns (2, l), (3, 5),
com onentsusin PCA Al orithm. 10 C03
7 OR
of LDA in dimensionality reduction with a suitable example.
Illustrate the various steps

knowledge, skill set that a student attains after completing the course
e CO —CourseOutcome, generally, refer to traits,
successfully.
cognitive process levels viz. 1. Remember, 2.
• BL—Asper Revised Bloom's Taxonomy, Bloom's Levels (BLS) are the
Understand, 3. Apply, 4. Analyze 5. Evaluate and 6. Create
C-Conceptual, P—Procedural, M -
• KC —As per Revised Bloom's Taxonomy, Knowledge Categories (KCs) are F-Factual,
Metacognitive
B.TECØ.
(SEM Vil) THEORY EXAMINATION2021-22
DEEP LEARNING
Time: 3 Hours Total Marks: 100
Note: 1. Attempt all Sections. If require any missingdata; then choose suitably.

SECTION A
1. Attempt all questions in brief. 2 x 10 = 20
a. What are the applications of MachineLearning? When it is used.
b. What is deep leaming , Explain its usesand application and history.
c. What is Visual Geometry Group(VGG)?
d. What is SVMs and Perceptrons?
e. What is Neural networks as universalfunction approximates?
f. What are batch normalization?
g. Define Generative Adversarial Networks(GAN).
h. What are dimensionalityreductiontechniques?
i. What is deep learning techniques?
j. Define Optimization in deep learning.

SECTION B
2. Attempt any three of the following: 10x3=30
a. Draw and explain the architecture of corfyu{ion network
Difference between Deep and Sha o Network.
C. What is semi —supervised learnA..
d. Explain Back propagation w•tq algorithm.
e. Explain different comppne ts of SOA and their fitnctionalities.

SECTION
c
3. Attempt any one part ofYe following: 10 x 1=10
(a) Explain LSTM (Long Short Term Memory
(b) What is PCA (Principle ComponentAnalysis ) and R

4. Attempt any one part of the following: 10 x 1—


-10
(a) What are some common problemswith LSTM?
(b) How Al and neuroscience drive each other forwards?

5. Attempt any one part of the following: 10 x 1=10


(a) Explain Fractal Structureand GeneralizationProperties of Stochastic
Optimization
Algorithms.
(b) What are the applications of a recurrentneural network RNN?

6. Attempt any one part of thqto lowing:


(a) Explain RecurrentNeural Networksand Natural Language 10 x 1=10
Processing.
(b) What are good quality
word embeddingand how to generate them?
7. Attempt any one part of the
(a) following:
How will the use of facial 10 x 1=10
recognitionby private companies affect privacy?
(b) Explain Image Captioning in
DeepLearning.

You might also like