(5th sem) AIML Chapter 2: Questions & Answers
Chapter 2: Knowledge Representation and Concept Learning -> Generated and Prepared By Thiruselvan (ThiruXD)
A. Multiple Choice Questions (with Answers)
1. Knowledge Representation is the method of:
a) Storing data in files b) Storing information so that a computer can reason with it
c) Compiling programs d) Sorting numbers
Answer: b) Storing information so that a computer can reason with it
2. Which is not a requirement of good Knowledge Representation?
a) Representational Adequacy b) Inferential Efficiency
c) High storage cost d) Acquisitional Efficiency
Answer: c) High storage cost
3. “Earth is round” is an example of:
a) Procedural Knowledge b) Declarative Knowledge
c) Heuristic Knowledge d) Meta Knowledge
Answer: b) Declarative Knowledge
4. Predicate Logic is also known as:
a) Propositional Logic b) First-Order Logic
c) Fuzzy Logic d) Temporal Logic
Answer: b) First-Order Logic
5. The symbol ∀ stands for:
a) There exists b) For all c) Not equal d) Implies
Answer: b) For all
6. The symbol ∃ stands for:
a) For all b) There exists c) And d) Or
Answer: b) There exists
7. In a Rule-Based System, the component that applies rules is called:
a) Knowledge Base b) Inference Engine c) Working Memory d) Database
Answer: b) Inference Engine
8. Forward Chaining starts from:
a) Goal b) Known facts c) Hypothesis d) Random rule
Answer: b) Known facts
9. Backward Chaining starts from:
a) Facts b) Goal c) Sensors d) Random state
Answer: b) Goal
10. Concept Learning is the process of learning a general concept from:
a) Only positive examples b) Positive and negative examples
c) Only rules d) Only numbers
Answer: b) Positive and negative examples
11. The set of all hypotheses consistent with the training examples is called:
a) Hypothesis Space b) Version Space c) Search Space d) State Space
Answer: b) Version Space
12. Find-S algorithm finds the:
a) Most general hypothesis b) Most specific hypothesis
c) All possible hypotheses d) Random hypothesis
Answer: b) Most specific hypothesis
13. Find-S algorithm ignores:
a) Positive examples b) Negative examples c) All examples d) Attributes
Answer: b) Negative examples
14. In Candidate Elimination, the S-boundary represents:
a) Most general hypotheses b) Most specific hypotheses
c) All hypotheses d) Negative examples
Answer: b) Most specific hypotheses
15. In Candidate Elimination, the G-boundary represents:
a) Most specific hypotheses b) Most general hypotheses
c) Only positive examples d) Training data
Answer: b) Most general hypotheses
16. Inductive Bias is required because:
a) It makes learning slower b) Without it, generalization is not possible
c) It increases memory d) It removes all hypotheses
Answer: b) Without it, generalization is not possible
17. Preference Bias means:
a) Restricting the hypothesis space b) Preferring some hypotheses over others
c) Ignoring all examples d) Using only negative examples
Answer: b) Preferring some hypotheses over others
18. Which algorithm maintains the entire Version Space?
a) Find-S b) Candidate Elimination c) BFS d) Hill Climbing
Answer: b) Candidate Elimination
19. Unification is used in:
a) Propositional Logic only b) Predicate Logic
c) Clustering d) Decision Trees
Answer: b) Predicate Logic
20. Which type of knowledge represents “rules of thumb”?
a) Declarative b) Procedural c) Heuristic d) Meta
Answer: c) Heuristic
B. Theory Questions with Answers
1. What is Knowledge Representation? Why is it important in AI?
Answer:
Knowledge Representation is the method of storing facts, rules, relationships and information in a form that a computer can understand and use for reasoning and decision-making.
It is important because without proper representation, an AI system cannot reason, draw conclusions, answer questions or solve problems intelligently.
2. List and explain the four requirements of a good Knowledge Representation system.
Answer:
- Representational Adequacy – Ability to represent all kinds of required knowledge.
- Inferential Adequacy – Ability to draw new conclusions from existing knowledge.
- Inferential Efficiency – Ability to draw conclusions quickly and efficiently.
- Acquisitional Efficiency – Ease of adding new knowledge to the system.
3. Differentiate between Declarative, Procedural, Heuristic and Meta Knowledge.
Answer:
| Type | Meaning | Example |
|---|---|---|
| Declarative | Facts and statements | “Water freezes at 0°C” |
| Procedural | How to perform a task | Steps to solve a problem |
| Heuristic | Rules of thumb / expert experience | “If fever + cough → likely flu” |
| Meta Knowledge | Knowledge about knowledge | “Rule A is more reliable than Rule B” |
4. Explain the components of Predicate Logic with examples.
Answer:
- Constants – Specific objects (Ram, Delhi, 5)
- Variables – Placeholders (x, y)
- Predicates – Properties or relations (Student(x), Loves(x,y))
- Functions – Mapping between objects (FatherOf(x), Age(x))
- Quantifiers – ∀ (for all), ∃ (there exists)
5. What are Forward Chaining and Backward Chaining? Give one example of each.
Answer:
-
Forward Chaining (Data-driven): Starts from known facts and applies rules to reach a conclusion.
Example: Patient has fever and cough → System concludes “Flu”.
-
Backward Chaining (Goal-driven): Starts from a goal and works backwards to find supporting facts.
Example: Goal = Check if patient has Flu → System checks whether fever and cough are present.
6. Explain the Concept Learning Task.
Answer:
Concept Learning is the process of learning a general concept (category) from a set of training examples.
Main elements:
- Training examples (positive and negative)
- Target concept (the true concept to be learned)
- Hypothesis space (all possible concepts the learner can consider)
7. Explain the working of Find-S Algorithm.
Answer:
- Initialize hypothesis to the most specific hypothesis (all attributes null).
- For each positive training example:
- Generalize the hypothesis just enough to cover the example.
- Ignore all negative examples.
- The final hypothesis is the most specific hypothesis consistent with all positive examples.
8. What is Version Space? How is it represented in Candidate Elimination Algorithm?
Answer:
Version Space is the set of all hypotheses that are consistent with the training examples seen so far.
In Candidate Elimination, it is represented by two boundaries:
- S-boundary → Set of most specific hypotheses
- G-boundary → Set of most general hypotheses
All hypotheses between S and G form the Version Space.
9. Explain the working of Candidate Elimination Algorithm.
Answer:
- Start with S = most specific hypothesis, G = most general hypothesis.
- For a positive example: Generalize S-boundary so that it covers the example. Remove any hypothesis in G that is inconsistent.
- For a negative example: Specialize G-boundary so that it excludes the example. Remove any hypothesis in S that is inconsistent.
- Continue until all examples are processed. The remaining S and G represent the Version Space.
10. What is Inductive Bias? Why is it necessary?
Answer:
Inductive Bias is the set of additional assumptions a learning algorithm makes so that it can generalize beyond the given training examples.
It is necessary because without bias, the system cannot decide which hypothesis is better when many hypotheses fit the training data. Bias enables useful generalization and efficient learning.
C. Analytical / Application Questions with Answers
1. Represent the following statements in Predicate Logic:
a) All students are intelligent.
b) There exists a student who likes AI.
c) Ram is a student and Ram likes AI.
Answer:
a) ∀x Student(x) → Intelligent(x)
b) ∃x Student(x) ∧ Likes(x, AI)
c) Student(Ram) ∧ Likes(Ram, AI)
2. A medical expert system has the following rule:
IF fever AND cough THEN flu
Patient has fever and cough. Which inference method will be used and what will be the result?
Answer:
Forward Chaining will be used.
Starting from facts (fever, cough), the rule is applied and the system concludes that the patient has flu.
3. Why does Find-S algorithm sometimes give an incomplete solution?
Answer:
Find-S considers only positive examples and finds only the most specific hypothesis. It ignores negative examples and does not maintain other possible consistent hypotheses. Therefore, it may miss better or more general hypotheses that also fit the data.
4. Compare Find-S and Candidate Elimination Algorithms.
| Feature | Find-S | Candidate Elimination |
|---|---|---|
| Hypothesis found | Only most specific | All consistent hypotheses (Version Space) |
| Negative examples | Ignored | Used |
| Output | Single hypothesis | S and G boundaries |
| Completeness | Low | High |
| Computational cost | Low | Higher |
5. A learning system is trained with examples of “birds”. It correctly classifies sparrows and eagles as birds but also classifies airplanes as birds. What type of problem is this and how can Inductive Bias help?
Answer:
This is a generalization error (overly general hypothesis).
Inductive Bias can help by preferring simpler or more restricted hypotheses (e.g., “has feathers” or “can lay eggs”) so that airplanes are correctly excluded.
6. Why is Version Space useful in concept learning?
Answer:
Version Space keeps all hypotheses that are still consistent with the training data. It allows the system to:
- Avoid committing too early to one hypothesis
- Handle future examples better
- Identify when more training data is needed (when S and G become empty or too wide)
7. Give one real-life example where Backward Chaining is more suitable than Forward Chaining.
Answer:
Medical diagnosis when the doctor has a suspected disease (goal) and wants to check whether the patient’s symptoms support that diagnosis. Backward chaining starts from the suspected disease and looks for confirming symptoms.
Quick Revision Checklist for Chapter 2
- 4 Requirements of KR
- Types of Knowledge (Declarative, Procedural, Heuristic, Meta)
- Predicate Logic components + Quantifiers (∀, ∃)
- Forward vs Backward Chaining
- Concept Learning + Version Space
- Find-S (most specific, ignores negatives)
- Candidate Elimination (S & G boundaries)
- Inductive Bias (Restriction + Preference)
(5th sem) AIML Chapter 2: Complete Concepts Guide
Chapter 2: Knowledge Representation and Concept Learning -> Generated and Prepared By Thiruselvan (ThiruXD)
(5th sem) AIML Chapter 3: Complete Concepts Guide
Chapter 3: Decision Tree Learning and Artificial Neural Networks -> Generated and Prepared By Thiruselvan (ThiruXD)