Skip to content
EndsemMay 20242019 patternBE AI&DS

Reinforcement Learning SPPU Endsem Question Paper May 2024

All 8 questions from this SPPU endsem paper, as printed. Read them here or open the PDF.

Open the PDF
Subject code
417533 D
Pattern
2019 pattern
Exam
Endsem
Session
May 2024
Max. marks
70
Time
2½ Hours
Semester
VIII
Paper code
[6263]-399

Printed for: B.E. (Artificial Intelligence and Data Science)

Instructions to the candidates

  1. Answer Q.1 or Q.2, Q.3 or Q.4, Q. 5 or Q. 6, Q. 7 or Q. 8.
  2. Neat diagrams must be drawn whenever necessary.
  3. Figures to the right side indicate full marks.
  4. Assume suitable data, if necessary.

Questions

  1. Q1

    1. a)

      What is dynamic programming, And how does it apply to solving Markov Decision Processes?

      [9]
    2. b)

      State the Banach Fixed point Theorem and its significance in dynamic programming.

      Also asked:May 2025

      [9]

  2. Q2

    1. a)

      Explain the contraction mapping property of Bellman expectation and optimality operators.

      Also asked:May 2025

      [9]
    2. b)

      State and explain the principle of optimality in the context of MDPs.

      [9]

  3. Q3

    1. a)

      What are monte Carlo Methods, and how are they used in reinforcement learning.

      [9]
    2. b)

      Explain the idea behind per-decision Importance Sampling and its significance in off-policy learning.

      [8]

  4. Q4

    1. a)

      What is the difference between On-policy and Off-policy learning in reinforcement learning.

      [8]
    2. b)

      What is Monte Carlo Tree Search (MCTS), and where is it commonly used? 2

      [9]

  5. Q5

    1. a)

      Enlist the advantages and disadvantages of using model-based and model- free approaches in reinforcement learning.

      [9]
    2. b)

      Describe the Q-learning algorithm and its main components.

      [9]

  6. Q6

    1. a)

      Discuss the double DQN algorithm and its advantages over traditional DQNs.

      [9]
    2. b)

      Explain the concept of Temporal difference (TD) learning in reinforcement learning.

      [9]

  7. Q7

    1. a)

      How can an agent adapt when the model used for planning is inaccurate?

      [9]
    2. b)

      How do Rollout Algorithms help in approximating the value function and improving decision-making?

      [8]

  8. Q8

    1. a)

      Explain the Dyna architecture and how it integrates planning, acting, and learning.

      [8]
    2. b)

      Discuss the advantages and limitations of using real-time Dynamic programming. 

      [9]