Skip to content
EndsemMay 20252019 patternBE AI&DS

Reinforcement Learning SPPU Endsem Question Paper May 2025

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

Open the PDF
Subject code
417533D
Pattern
2019 pattern
Exam
Endsem
Session
May 2025
Max. marks
70
Time
2½ Hours
Semester
VIII
Paper code
[6404]-394

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

Instructions to the candidates

  1. Answer Q1 or Q2, Q3 or Q4, Q5 or Q6, Q7 or Q8
  2. Neat diagrams must be drawn wherever necessary.
  3. Figures to the right side indicate full marks.
  4. Assume Suitable data if necessary.

Questions

  1. Q1

    1. a)

      Discuss dynamic programming for the Markov decision process and define formulation of planning in MDPs.

      [6]
    2. b)

      List and Explain the principles of optimality.

      [5]
    3. c)

      Elaborate Iterative policy evaluation and policy iteration.

      [6]

  2. Q2

    1. a)

      Elaborate proof of contraction mapping property of Bellman expectation and optimality operators.

      Also asked:May 2024

      [7]
    2. b)

      Discuss Banach fixed point theorem.

      Also asked:May 2024

      [5]
    3. c)

      Explain a proof of convergence of policy evaluation and value iteration algorithms.

      [5]

  3. Q3

    1. a)

      Discuss the role of Monte Carlo methods for model-free Reinforcement Learning and Monte Carlo control.

      [8]
    2. b)

      Explain on-policy and off-policy learning techniques .

      [4]
    3. c)

      Elaborate first visit and every visit to Monte Carlo in Reinforcement Learning.

      [6]

  4. Q4

    1. a)

      Discuss Discounting-aware Importance Sampling and Per-decision Importance Sampling with examples.

      [12]
    2. b)

      Elaborate Monte Carlo tree search along with examples. 2

      [6]

  5. Q5

    1. a)

      Elaborate Deep Q-networks with convolution neural networks and single- layer neural networks.

      [12]
    2. b)

      Compare Model based learning and model-free learning with applications.

      [6]

  6. Q6

    1. a)

      Elaborate Temporal difference learning technique.

      [6]
    2. b)

      Explain the Separate target network and discuss the role of the Separate target network in computing the target Q-values.

      [6]
    3. c)

      Elaborate Double DQN and Dueling DQN in reinforcement learning.

      [6]

  7. Q7

    1. a)

      Elaborate Multi-agent Reinforcement Learning with Rollout and Policy Iteration.

      [6]
    2. b)

      Discuss Trajectory Sampling and Real-time Dynamic Programming with respect to learning.

      [6]
    3. c)

      Comment on Planning at Decision Time along with its importance.

      [5]

  8. Q8

    1. a)

      Elaborate Heuristic Search and Rollout Algorithms in reinforcement learning.

      [6]
    2. b)

      Discuss Integrated Planning, Acting and Learning in Planning and Learning.

      [6]
    3. c)

      Explain Trajectory Sampling, and Real-time Dynamic Programming. ♦♦♦

      [5]