End Semester Examinations

Course Code / Name 20AIPC403/ ADVANCED MACHINE LEARNING

Regulations 2020
Department AI&DS
Year / Semester II / IV
Course Moderator R.RADHIKA/AP/AI&DS

Course Outcomes with Knowledge Level

Upon the completion of this course the students will be able to
CO 1 Explain the fundamentals of distinct graphical models and its applications (K2)
CO 2 Implement the essential inference and parameters in graphical models(K3)
Discuss about the objects and sampling, based on the high dimensional generative
CO 3
models(K2)
CO 4 Describe uncertainty estimation and to define density and distribution estimation (K2)
CO 5 Demonstrate the performance of diverse graph model in machine learning (K3)
CO 6 Examine various real time forecasting model in advanced machine learning (K4)

Distribution of COs (Percentage wise)

Exam / CO CO1 CO2 CO3 CO4 CO5 CO6
End Semester Exam 16.66 16.66 16.66 16.66 16.66 16.66

Unit – I

CO1 / Part - A / 1 Mark/ MCQ

Sl. Marks K–
Questions CO
No. Split-up Level
1. 1 K1 CO1

2. 1 K1 CO1
3. 1 K1 CO1
4. 1 K1 CO1
5. 1 K1 CO1
6. 1 K1 CO1
7. 1 K1 CO1
8. 1 K1 CO1
9. 1 K1 CO1
10. 1 K1 CO1
11. 1 K1 CO1
12. 1 K1 CO1
13. 1 K1 CO1
14. 1 K1 CO1
15. 1 K1 CO1
16. 1 K1 CO1
17. 1 K1 CO1
18. 1 K1 CO1
19. 1 K1 CO1
20. 1 K1 CO1
21. 1 K1 CO1
22. 1 K1 CO1
23. 1 K1 CO1

Unit - I / Part - B / 2 Marks

S.No Questions Marks, K -
Level,CO
1. Define Graphical model and its types. K1, CO1
2. Define polytree with an example. K1, CO1
 Unit - I / Part - B / 2 Marks
3. Find the degree of all the nodes in the given graph.

K2, CO1

4. Distinguish between subgraph and complete graph. K1, CO1

5. Describe the impact of clique. K2, CO1
6. What are the steps involved in undirected graphs? K1, CO1
7. Illustrate the terminologies for directed graphs. K2, CO1
8. When to use potts model. K2, CO1
9. Analyze the benefits of the Bayesian model. K4, CO1
10. What do you mean by markov random fields. K1, CO1
11. Differentiate interpolation and extrapolation with a neat diagram. K4, CO1
12. Illustrate the features of CRF. K2, CO1
13. What do you mean by moralization? K1,CO1

Unit - I / Part - C/ 10 Marks

S.No Questions Marks, K -
Level,CO
1. Design and explain the directed graphical model with an example. K2, CO1
2. Discuss about undirected graphical model and its terminologies. K2, CO1
3. Analyze and discuss the potts and ising model. K2, CO1
4. Explain your understanding of CRF and list out the advantages. K2, CO1
5. Illustrate the markov random field with an example. K2, CO1
6. Summarize the log linear model and write down the limitations of K2, CO1
the log linear model.
7. Generalize the factor models. K2, CO1
8. What is factorization? Explain the factor analysis in brief. K2, CO1
9. Describe the concept of tree. K2, CO1
10. State Bayesian networks. K2, CO1
11. How to convert directed graphs to undirected graphs. K2, CO1
12. Describe the impact of gibbs distribution. K2, CO1
 Unit - I / Part - C/ 10 Marks
13 Analyze any two undirected graphical models and point out the K3, CO1
features.
14 Explain in detail on ising and potts model with neat diagram K2, CO1
15 Why is the undirected graphical model more familiar than the K2, CO1
directed graphical model? Discuss one model for each.

Unit – II

CO2 : Implement the essential inference and parameters in graphical models(K3)

CO1 / Part - A / 1 Mark/ MCQ
Sl. Marks K–
Questions CO
No. Split-up Level
1. 1 K1 CO1

2. 1 K1 CO1
3. 1 K1 CO1
4. 1 K1 CO1
5. 1 K1 CO1
6. 1 K1 CO1
7. 1 K1 CO1
8. 1 K1 CO1
9. 1 K1 CO1
10. 1 K1 CO1
11. 1 K1 CO1
12. 1 K1 CO1
13. 1 K1 CO1
14. 1 K1 CO1
15. 1 K1 CO1
16. 1 K1 CO1
17. 1 K1 CO1
18. 1 K1 CO1
19. 1 K1 CO1
20. 1 K1 CO1
21. 1 K1 CO1
22. 1 K1 CO1
23. 1 K1 CO1

Unit - II / Part - B / 2 Marks

S.No Questions Marks, K-
Level,CO
1. What do you mean by belief propagation? K1,CO2
2. How to do variable elimination in a graphical model? K1,CO2
3. How to learn the model with partially observed data? K1,CO2
4. What do you mean by running intersection property? K1,CO2
5. What is commutative semi-ring? K1,CO2
6. List some commutative semi-ring functions. K1,CO2
7. What is the elimination order? K1,CO2
8. Define junction tree calibration. K1,CO2
9. Define Gradient Ascent. K1,CO2
10. What is Expectation Maximization? K1,CO2
11. What do you mean by moment matching? K1,CO2
12. List the two versions of junction tree inference. K1,CO2

Unit - II / Part - C / 10 Marks

S.No Questions Marks, K-
Level,CO
1. Discuss about various variable elimination methods. K2,CO2
2. Discuss about learning conditional graphical models with examples. K2,CO2
3. Explain junction tree algorithm K2,CO2
4. How can the jtree algorithm be applied on chain structured graphs? K3,CO2
5. How will you eliminate the variable while computing the marginal K2,CO2
probability?
6. Write the Expectation-maximization algorithm for BN with table- K2,CO2
CPDs.
7. Demonstrate the Maximum Likelihood Estimation. K3,CO2
8. K2,CO2

Discuss Conditionally Trained Models with

9 K2,CO2

How will you perform message passing in a junction tree?

 Unit - II / Part - C / 10 Marks
10 K2,CO2

Explain Conditional Likelihood Parameter Estimation

11 K3,CO2

Explain the use of Variable elimination algorithms in graphical

model inference.

Unit – III

CO3 : Discuss about the objects and sampling, based on the high dimensional generative
models(K2)

CO1 / Part - A / 1 Mark/ MCQ

Sl. Marks K–
Questions CO
No. Split-up Level
1. 1 K1 CO1

2. 1 K1 CO1
3. 1 K1 CO1
4. 1 K1 CO1
5. 1 K1 CO1
6. 1 K1 CO1
7. 1 K1 CO1
8. 1 K1 CO1
9. 1 K1 CO1
10. 1 K1 CO1
11. 1 K1 CO1
12. 1 K1 CO1
13. 1 K1 CO1
14. 1 K1 CO1
15. 1 K1 CO1
16. 1 K1 CO1
17. 1 K1 CO1
18. 1 K1 CO1
19. 1 K1 CO1
20. 1 K1 CO1
21. 1 K1 CO1
22. 1 K1 CO1
23. 1 K1 CO1

Unit - III / Part - B / 2 Marks

S.No Questions Marks, K - Level,
CO
1. What do you mean by generative model? K1,CO3
2. Describe the impact of auto encoder. K1,CO3
3. Draw the functional diagram for variation auto encoder. K2,CO3
4. What do you mean by GAN? K1,CO3
5. What is loss function? K1,CO3
6. List some real time examples for GAN. K1,CO3
7. Summarize cost function. K2,CO3
8. Define sampling. K1,CO3
9. Distinguish between forward and importance sampling. K4,CO3
10. What is MCMC sampling? K1,CO3
11. What do you mean by neural network? K1,CO3
12. Write the role of latent space. K1,CO3

Unit - III / Part - C / 10 Marks

S.No Questions Marks, K - Level,
CO
1. Discuss about generative model for text. K2,CO2
2. Explain in detail about GAN with suitable examples. K2,CO2
3. Explain variational auto encoders with neat diagram. K2,CO2
4. Analyze loss function and its types in GAN. K4,CO2
 Unit - III / Part - C / 10 Marks
5. Explain in detail about forward sampling. K2,CO2
6. Write types of sampling and demonstrate the importance sampling. K2,CO2
7. Discuss the significant role of MCMC sampling. K2,CO2
8. K2,CO2

Why is the GAN mechanism more familiar than the auto encoders.

9 K2,CO3

How will you perform loss function in GAN and write the impact of
generator and discriminator?

10 K2,CO3

Explain MCMC sampling with suitable examples.

11 K2,CO3

Explain the use of generative model and write any one real time
examples.

Unit – IV

CO4: Describe uncertainty estimation and to define density and distribution estima-
tion (K2)

CO5: Demonstrate the performance of diverse graph model in machine learning (K3)
 CO1 / Part - A / 1 Mark/ MCQ
Sl. Marks K–
Questions CO
No. Split-up Level
1. 1 K1 CO1

2. 1 K1 CO1
3. 1 K1 CO1
4. 1 K1 CO1
5. 1 K1 CO1
6. 1 K1 CO1
7. 1 K1 CO1
8. 1 K1 CO1
9. 1 K1 CO1
10. 1 K1 CO1
11. 1 K1 CO1
12. 1 K1 CO1
13. 1 K1 CO1
14. 1 K1 CO1
15. 1 K1 CO1
16. 1 K1 CO1
17. 1 K1 CO1
18. 1 K1 CO1
19. 1 K1 CO1
20. 1 K1 CO1
21. 1 K1 CO1
22. 1 K1 CO1
23. 1 K1 CO1

Unit - IV/ Part - B / 2 Marks

S.No Questions Marks, K - Level,
CO
1. Define Bayesian Neural Network. K1,CO5
 Unit - IV/ Part - B / 2 Marks
2. Differentiate between Standard neural network and Bayesian neural K1,CO5
network.
3. What do you mean by over fitting? K1,CO5
4. Define Posterior inference. K1,CO5
5. List the advantages of Bayesian Neural Network. K1,CO5
6. What are the types of uncertainties in parameter estimation? K1,CO5
7. What is Aleatory uncertainty? K1,CO5
8. What is Epistemic uncertainty? K1,CO5
9. List the levels of causality based on Judea Pearl’s work. K1,CO5
10. Define Meta learning. K1,CO5
11. List the two new methods for density estimation using multilayer K1,CO4
networks.
12. What do you mean by autoregressive model in density estimator? K1,CO4
13. What is the use of normalizing flows? K1,CO4
14. How to modify the autoencoder to satisfy the autoregressive K1,CO4
property?
15. What is the use of variational autoencoder algorithm? K1,CO4
16. Give the general pattern for training a variational auto-encoder. K1,CO4
17. List the abstractions defined by Tensor Flow Distribution. K1,CO4
18. What do you mean by approximate inference? K1,CO4
19. What are the three partitions of Tensor’s shape? K1,CO4
20. What are the applications of Tensor Flow Distribution? K1,CO4

Unit - IV/ Part - C / 10 Marks

S.No Questions Marks, K - Level,
CO
1. Explain Bayesian Neural Networks in detail. K2,CO5
2. Explain in detail about Meta-Learning. K2,CO5
3. Explain in detail about Counterfactual reasoning. K2,CO5
4. Explain in detail about Causality . K2,CO5
5. Describe briefly about Masked Autoregressive Flow for Density K2,CO4
Estimation.
6. Explain in detail Masked Autoencoder for Distribution Estimation. K2,CO4
7. Explain in detail Density estimation using real NVP K2,CO4
8. Explain in detail Few-shot Autoregressive Density Estimation. K2,CO4
9. Explain in detail TensorFlow Distributions K2,CO4
10. Explain Neural models for density estimation. K2,CO4
11 K2,CO5

Explain in briefly about Uncertainties in Parameters Estimated with

Neural Networks.
 Unit - IV/ Part - C / 10 Marks
12 K2,CO5

Explain briefly about Meta-Learning, Counterfactual reasoning,

Causality

13 K2,CO4

Explain autoregressive method for density estimation.

14 K2,CO4

Explain masked method for density estimation.

Unit – V

CO6 : Examine various real time forecasting model in advanced machine learning
(K4)

CO1 / Part - A / 1 Mark/ MCQ

Sl. Marks K–
Questions CO
No. Split-up Level
1. 1 K1 CO1

2. 1 K1 CO1

3. 1 K1 CO1

4. 1 K1 CO1

5. 1 K1 CO1

6. 1 K1 CO1

7. 1 K1 CO1

8. 1 K1 CO1

9. 1 K1 CO1

10. 1 K1 CO1
 CO1 / Part - A / 1 Mark/ MCQ

11. 1 K1 CO1

12. 1 K1 CO1

13. 1 K1 CO1

14. 1 K1 CO1

15. 1 K1 CO1

16. 1 K1 CO1

17. 1 K1 CO1

18. 1 K1 CO1

19. 1 K1 CO1

20. 1 K1 CO1

21. 1 K1 CO1

22. 1 K1 CO1

23. 1 K1 CO1

Unit - V / Part - B / 2 Marks

S.No Questions Marks, K - Level,

CO
1. Define Gaussian Process. K1,CO6

2. What do you mean by interpolator? K1,CO6

3. Define Multivariate forecasting K1,CO6

4. What do you mean by forecasting? K1,CO6

5. List the issues that prevent the estimation of high-dimensional K1,CO6

multivariate time series models.
6. What is a copula function? K1,CO6

7. Give any four public dataset for time series analysis. K1,CO6

8. List two baseline RNN models. K1,CO6

9. List the error metrics used in RNN. K1,CO6

10. What is lead time? K1,CO6

 CO1 / Part - A / 1 Mark/ MCQ

11. Define auto encoder. K1,CO6

12. Give the block diagram for time series data pre processing K1,CO6

Unit - III / Part - C/ 10 Marks

S.No Questions Marks, K - Level,
CO
1. Explain Probabilistic Forecasting in detail. K2,CO6
2. Describe briefly about how Autoregressive Recurrent Networks is K2,CO6
used in probabilistic forecasting.
3. Explain in detail High-Dimensional Multivariate Forecasting. K2,CO6
4. Explain in detail Low-Rank Gaussian Copula Processes K2,CO6
5. Explain Encoder-Decoder Model for Multivariate Time Series K2,CO6
Forecasting.
6. K2,CO6

Explain Deep AR methodology for probabilistic forecasting.

7. K2,CO6

How will perform forecasting in Correlated time series?