VU Past Papers CS607 – Artificial Intelligence Solved Subjective Questions

Q1: Differentiate between Mutation and Crossover (2 Marks)

Answer:

• Mutation: Each “individual” (solution) has one parent.
• Crossover (Inheritance): Each “individual” (solution) has two parents.

Q2: CLIPS command to remove only facts (2 Marks)

Answer:

• The retract command is used to remove facts.
  • Example: (retract 1) removes fact 1
  • (retract 1 3) removes fact 1 and 3

Q3: “Bike is heavy” – Uncertain fact or not? (3 Marks)

Answer:

• This statement is a Fuzzy fact because it is ambiguous.
• Fuzzy facts use certainty factor values to define the degree of truth.
• Example: “The book is heavy/light” – subjective description, requires fuzzy representation.

Q4: Importance of Knowledge Base in Expert System (3 Marks)

Answer:

• Contains domain knowledge:
  • Problem facts and rules
  • Concepts
  • Relationships
• Power of ES lies in richness of knowledge.
• Knowledge engineer’s role: encode expert knowledge using knowledge representation techniques.

Q5: Conflict Resolution Strategies (5 Marks)

Answer:

1. Fire first rule in sequence: Rules ordered by sequence; fire first matching rule.
2. Assign rule priorities: Explicit priorities to resolve conflicts.
3. Prefer more specific rules: Rules with more premises preferred.
4. Prefer recently added premises: Timestamp-based prioritization.
5. Parallel strategy: Branch execution into multiple threads; maintain multiple viewpoints.

Q6: “Riding a Horse is same as Riding a Donkey” – Type of reasoning (5 Marks)

Answer:

• Belongs to Knowledge Representation and Reasoning (KR & R).
• KR & reasoning are interdependent:
  • Same information can be represented differently depending on purpose.
  • Example: Half of something: 0.5x or a shaded diagram – both convey the same information.
• Representation must be useful for reasoning in AI systems.

Q7: GA using mutation procedure (Example)

Answer:

• Given 32-bit word: first 16 bits 0, last 16 bits 1.
• Apply mutation by flipping bits randomly according to mutation probability.

Q8: Step-by-step Backward Chaining (5 Marks)

Answer:

1. Start with the goal.
2. Check if goal is in working memory (WM).
3. Search for goal in THEN part of rules (goal rule).
4. Check if goal rule’s premises are in WM.
5. Premises not listed → become sub-goals.
6. Recursive process until a primitive (cannot be concluded) is found.
7. Ask user for primitive info; backtrack to prove sub-goals and goal.

Q9: How Knowledge Representation & Reasoning are coupled (3 Marks)

Answer:

• KR & reasoning are closely coupled, but representation alone is not meaningful.
• Example: Ratios in algebra vs. hand-drawn symbols – both convey same info, but algebraic is compact.
• AI designer must consider interdependence in problem solving.

Q10: Structure of Expert System & analogy (3 Marks)

Answer:

• Expert = domain specialist (e.g., doctor)
• Components:
  • Focused expertise
  • Specialized knowledge (LTM)
  • Case facts (STM)
  • Reasoning to generate new knowledge
  • Solving problems using this knowledge

Q12: Shallow vs. Structural Knowledge (2 Marks)

Answer:

• Shallow (Heuristic) Knowledge: Rule-of-thumb, empirical, e.g., seeing shops → near market.
• Structural Knowledge: Describes structures and relationships, e.g., how car parts fit together.

Q13: Best memory type vs Knowledge Base (5 Marks)

Answer:

• Pictorial memory: best for recognition & structural info.
• Knowledge base: stores rules, concepts, relationships.
• Pictures are easy for humans but harder for computers.

Q14: System to model humans (2 Marks)

Answer:

• Expert system

Q15: Cost comparison – Expert System vs Human Expert (2 Marks)

Answer:

• Expert system is better; low cost.

Q16: Meta Knowledge vs Heuristic Knowledge (3 Marks)

Answer:

• Meta Knowledge: Knowledge about knowledge (e.g., blood pressure more important than eye color).
• Heuristic Knowledge: Rule-of-thumb; shallow, empirical, not deterministic.

Q17: Conventional System vs Expert System (3 Marks)

Answer:

• Expert System separates knowledge & control, unlike conventional programs.
• Components: Knowledge base, Working Memory, Inference Engine
• Changes to knowledge do not affect control and vice versa.

Q18: Monotonic vs Non-Monotonic Reasoning

Answer:

• Monotonic: Facts remain true even if conditions change.
• Non-Monotonic: Facts can change; uses truth maintenance system.
• Example: Wind blows → curtains sway; wind stops → curtains no longer sway.

Q19: AI Languages (2 Marks)

Answer:

• Lisp, Microworlds

Q20: Forward Chaining (2 Marks)

Answer:

• Starts from known facts, infers new facts using rules until goal reached.
• Steps:
  1. Add facts to WM
  2. Match premises of rules with WM
  3. If match → assert conclusion to WM
  4. Repeat until no new facts

Q21: Learning Ability – Human Expert vs Expert System

Answer:

• Human Expert has better learning ability.

Q22: Appropriate Domains for Expert System (5 Marks)

Answer:

• Problems not solvable by conventional programs.
• Domains well-bounded; expert cooperation required.
• Especially useful for ill-structured, heuristic, uncertain domains.

Q23: Issues in Forward Chaining (2 Marks)

Answer:

• Undirected search
• Conflict resolution (strategies to select which rule to fire)

Q24: CNF Conversion (Example)

• (A v B) → (C → D) → Convert using logical equivalences.

Q25: Adversarial Search – Evaluation Function (3 Marks)

Answer:

• Evaluates board positions → produces score/number.
• Positive = advantage to maximizing player, Negative = advantage to minimizing player.

Q26: Perception & Knowledge Representation Coupling (2 Marks)

Answer:

• AI perception gathers info from environment (visual/audio).
• Must form meaningful internal representation.
• Coupled with learning component to detect trends from data.