Department of Computer Science and Engineering
UNIT-VII
Q. PO /
Question Title Marks CO BL PSO
No.
Reinforcement learning has been successfully
applied to:
A) Game playing, robotics, finance, and
healthcare
1. 1 CO4 L1 PO1
B) Only supervised learning tasks
C) Problems with no sequential decision-making
D) Fixed-rule optimization
One of the earliest examples of reinforcement
learning in games was:
A) Samuel’s Checkers Player
2. B) AlphaGo 1 CO4 L1 PO1
C) TD-Gammon
D) Deep Blue
TD-Gammon was developed to play:
A) Backgammon
B) Chess
3. 1 CO4 L1 PO1
C) Go
D) Poker
TD-Gammon used reinforcement learning
by:
A) Learning from self-play using TD(λ)
4. B) Using only human expert moves 1 CO4 L1 PO1
C) Following handcrafted rules
D) Relying on random moves
TD-Gammon demonstrated that:
A) Neural networks combined with
5. reinforcement learning can achieve expert-level 1 CO4 L1 PO1
play
B) Pure search-based approaches are better
C) Supervised learning is the only effective
training method
D) Training with human games leads to stronger
AI
Arthur Samuel’s Checkers Player was one of
the first programs to:
A) Use self-learning techniques to improve its
performance
6. B) Follow a fixed set of expert rules 1 CO4 L1 PO1
C) Learn exclusively from human
demonstrations
D) Play randomly without improvement
The learning technique used in Samuel’s
Checkers Player was based on:
A) Temporal-Difference learning
7. B) Supervised deep learning 1 CO4 L1 PO1
C) Decision trees
D) Bayesian networks
IBM’s Watson used reinforcement learning
to:
A) Optimize its wagering strategy in Jeopardy!
8. 1 CO4 L1 PO1
B) Play against Deep Blue
C) Improve supervised learning models
D) Solve algebra problems
Watson’s decision-making in wagering was
based on:
A) Probabilistic reasoning and game state
evaluation
9. 1 CO4 L1 PO1
B) Random number generation
C) Fixed betting amounts
D) Human expert advice
The impact of Samuel’s Checkers Player was:
A) Demonstrating that machines could improve
performance through experience
B) Showing that handcrafted rules are the best
10. approach 1 CO4 L1 PO1
C) Proving that game-playing AI cannot surpass
human players
D) Limiting AI to deterministic games
Watson’s success in Jeopardy! showed that:
A) AI can combine natural language processing
with strategic decision-making
B) Only symbolic AI is effective in language-
11. 1 CO4 L1 PO1
based tasks
C) Game-playing AI cannot be generalized
D) Rule-based approaches are always superior
The main breakthrough of AlphaGo was:
A) Using deep reinforcement learning to surpass
human Go players
12. 1 CO4 L1 PO1
B) Relying on brute-force search
C) Using only human expert moves for training
D) Applying tabular Q-learning
AlphaGo’s training combined:
A) Supervised learning from human games and
reinforcement learning through self-play
13. 1 CO4 L1 PO1
B) Only random move selection
C) Pure tree search without learning
D) A predefined rule-based system
One of the key techniques used by AlphaGo
was:
A) Monte Carlo Tree Search (MCTS)
14. B) Linear regression 1 CO4 L1 PO1
C) Genetic algorithms
D) Simple depth-first search
AlphaGo Zero improved upon AlphaGo by:
A) Learning entirely from self-play without
human data
15. B) Increasing reliance on expert demonstrations 1 CO4 L1 PO1
C) Avoiding neural networks
D) Reducing computational resources
