wayground logo

Free Printable Worksheets

Font size

S
M
L
XL
Worksheets

Worksheet Questions Extraction

Total questions: 84

Worksheet time: 42mins

Name
Class
Date
1.

A learning task is defined by ⟨T, P, E⟩. If a system improves its performance at playing chess with experience, then P refers to:

a)

Winning strategy

b)

Percentage of games won

c)

Number of board states

d)

Training dataset size

2.

Which of the following is NOT a valid perspective of machine learning?

a)

Statistical

b)

Algorithmic

c)

Biological

d)

Deterministic logic

3.

A concept is represented as a Boolean function over attribute space. This representation is most commonly used in:

a)

Reinforcement learning

b)

Concept learning

c)

Clustering

d)

Dimensionality reduction

4.

In concept learning, noise in training data primarily affects:

a)

Hypothesis representation

b)

Learning rate

c)

Consistency of hypothesis

d)

Target function

5.

If the hypothesis space contains the target concept, the learner is said to be:

a)

Complete

b)

Sound

c)

Realizable

d)

Optimal

6.

The version space represents:

a)

All hypotheses consistent with training data

b)

All possible hypotheses

c)

Only the most general hypothesis

d)

Only the most specific hypothesis

7.

The S-boundary in Candidate Elimination represents:

a)

Maximally general hypotheses

b)

Minimally general hypotheses

c)

Maximally specific hypotheses

d)

Minimally specific hypotheses

8.

When a positive example is misclassified by S, the algorithm will:

a)

Remove S

b)

Generalize S minimally

c)

Specialize G

d)

Discard the example

9.

When a negative example is covered by G, the algorithm will:

a)

Generalize G

b)

Remove S

c)

Specialize G minimally

d)

Expand version space

10.

Candidate Elimination fails when:

a)

Hypothesis space is finite

b)

Training data contains noise

c)

Data is linearly separable

d)

Version space is non-empty

11.

Inductive bias refers to:

a)

Error in data

b)

Prior assumptions made by the learner

c)

Sampling bias

d)

Training bias

12.

Without inductive bias, a learner can:

a)

Generalize perfectly

b)

Learn any function

c)

Only memorize training data

d)

Always overfit

13.

Occam’s Razor is an example of:

a)

Statistical bias

b)

Computational bias

c)

Inductive bias

d)

Representation bias

14.

A learner preferring linear hypotheses over nonlinear ones is an example of:

a)

Search bias

b)

Representation bias

c)

Sample bias

d)

Noise bias

15.

Which learning algorithm has no explicit inductive bias?

a)

Decision Trees

b)

KNN

c)

Naïve Bayes

d)

Neural Networks

16.

Decision tree learning performs a:

a)

Depth-first search

b)

Greedy search

c)

Exhaustive search

d)

Random search

17.

Information Gain is based on:

a)

Variance

b)

Entropy

c)

Mean squared error

d)

Probability density

18.

A highly unbalanced dataset may cause decision trees to:

a)

Underfit

b)

Ignore minority class

c)

Increase depth unnecessarily

d)

Fail to converge

19.

Which measure reduces bias toward multi-valued attributes?

a)

Information Gain

b)

Gini Index

c)

Gain Ratio

d)

Chi-square

20.

Pruning in decision trees helps to:

a)

Increase training accuracy

b)

Reduce variance

21.

Which representation allows disjunctions but not conjunctions?

a)

Decision trees

b)

Propositional logic

c)

Linear classifiers

d)

Conjunctive hypotheses

22.

The expressive power of a hypothesis space determines:

a)

Learning speed

b)

Generalization ability

c)

Whether the target can be represented

d)

Training time

23.

A more expressive hypothesis space increases the risk of:

a)

Underfitting

b)

Overfitting

c)

Bias

d)

Data leakage

24.

Which hypothesis representation is most interpretable?

a)

Neural networks

b)

Decision trees

c)

SVM

d)

KNN

25.

Conjunctive hypothesis spaces are limited because they cannot represent:

a)

AND relations

b)

OR relations

c)

Linear boundaries

d)

Boolean logic

26.

A heuristic is used to:

a)

Guarantee optimality

b)

Reduce search space

c)

Increase hypothesis space

d)

Eliminate noise

27.

Heuristic search in learning primarily trades off:

a)

Bias and variance

b)

Optimality and efficiency

c)

Accuracy and recall

d)

Precision and speed

28.

Which search strategy is used in decision tree learning?

a)

Backtracking search

b)

Greedy heuristic search

c)

Uniform-cost search

d)

Hill climbing

29.

A heuristic that always chooses the locally optimal split may lead to:

a)

Global optimum

b)

Local optimum

c)

Overfitting only

d)

Infinite loop

30.

Search in hypothesis space is exponential mainly due to:

a)

Noise

b)

Dimensionality

c)

Sample size

d)

Overfitting

31.

If both S and G converge to a single hypothesis, the learner has:

a)

Failed

b)

Generalized

c)

Learned the target concept

d)

Overfitted

32.

A large version space indicates:

a)

High confidence

b)

Insufficient data

c)

Perfect learning

d)

Low hypothesis complexity

33.

Adding irrelevant attributes to a decision tree dataset usually:

a)

Improves accuracy

b)

Reduces tree depth

c)

Increases overfitting risk

d)

Has no effect

34.

Which learning setting is most affected by inductive bias?

a)

Supervised

b)

Unsupervised

c)

Reinforcement

d)

All learning settings

35.

A learner that memorizes all examples suffers from:

a)

High bias

b)

High variance

c)

Underfitting

d)

Low complexity

36.

Candidate Elimination assumes the training data is:

a)

Probabilistic

b)

Noisy

c)

Noise-free

d)

Continuous

37.

Which of the following is a search bias?

a)

Limiting hypothesis space

b)

Preferring shorter trees

c)

Using entropy

d)

Removing noise

38.

A decision tree with depth equal to number of attributes implies:

a)

Underfitting

b)

Balanced learning

c)

Possible overfitting

d)

Optimal learning

39.

If entropy before and after split remains same, information gain is:

a)

1

b)

−1

c)

0

d)

Maximum

40.

Learning problems with continuous attributes often require:

a)

Discretization

b)

Elimination

c)

Encoding

d)

Normalization only

41.

Which factor primarily controls decision tree complexity?

a)

Learning rate

b)

Depth

c)

Feature scaling

d)

Dataset size

42.

A hypothesis consistent with all examples may still fail due to:

a)

Noise

b)

Overgeneralization

c)

Poor generalization

d)

Low bias

43.

Greedy learning algorithms are preferred because they:

a)

Guarantee optimality

b)

Reduce computational cost

c)

Avoid bias

d)

Remove noise

44.

Which component defines what is learnable?

a)

Training set

b)

Hypothesis space

c)

Learning rate

d)

Loss function

45.

Increasing training examples generally causes version space to:

a)

Expand

b)

Remain same

c)

Shrink

d)

Randomize

46.

Decision trees naturally handle:

a)

Missing values

b)

Only numeric data

c)

Only Boolean data

d)

Only balanced datasets

47.

A heuristic that is admissible ensures:

a)

Faster learning

b)

Optimal solution

c)

No overfitting

d)

Noise tolerance

48.

Which is a limitation of decision tree learning?

a)

Interpretability

b)

Handling non-linear boundaries

c)

High variance

d)

Greedy search

49.

Bias-variance tradeoff mainly affects:

a)

Training accuracy

b)

Model selection

c)

Feature scaling

d)

Data cleaning

50.

A learner that always predicts the majority class demonstrates:

a)

High variance

b)

High bias

c)

Optimal learning

d)

Noise fitting

51.

Learning Problems & Concept Learning: A learning problem where the output is a continuous real value is best categorized as:

a)

Classification

b)

Clustering

c)

Regression

d)

Reinforcement

52.

Learning Problems & Concept Learning: If the training accuracy is high but test accuracy is low, the model is likely:

a)

Underfitting

b)

Overfitting

c)

Well-generalized

d)

Noise-free

53.

Learning Problems & Concept Learning: In concept learning, the target function represents:

a)

Learner hypothesis

b)

Optimal hypothesis

c)

True mapping from instances to labels

d)

Training data distribution

54.

Learning Problems & Concept Learning: Which factor determines whether a hypothesis is consistent?

a)

Hypothesis size

b)

Dataset size

c)

Agreement with all training examples

d)

Prediction confidence

55.

Learning becomes impossible if:

a)

Hypothesis space is infinite

b)

No inductive bias is present

c)

Training data is large

d)

Learning rate is small

56.

Version Space Theory: The version space shrinks when:

a)

More hypotheses are added

b)

More training examples are observed

c)

Hypothesis space expands

d)

Noise is added

57.

Version Space Theory: If the S-boundary becomes empty, it indicates:

a)

Successful learning

b)

No hypothesis fits positive examples

c)

Overgeneralization

d)

Complete version space

58.

Version Space Theory: If G-boundary becomes empty during learning, it implies:

a)

Overfitting

b)

No hypothesis can explain negative examples

c)

Learning failure

d)

Noise-free data

59.

Candidate Elimination maintains:

a)

Only S

b)

Only G

c)

Both S and G

d)

Only most recent hypothesis

60.

Which operation is applied to S when encountering a negative example?

a)

Generalization

b)

Specialization

c)

No change

d)

Deletion

61.

Inductive Bias & Generalization: Bias introduced by restricting hypothesis space is known as:

a)

Search bias

b)

Statistical bias

c)

Representation bias

d)

Sampling bias

62.

Inductive Bias & Generalization: A learner that prefers smaller trees has bias toward:

a)

Accuracy

b)

Simplicity

c)

Completeness

d)

Variance

63.

Inductive Bias & Generalization: Which of the following best explains generalization?

a)

The ability of a model to perform well on unseen data by applying learned patterns

b)

Memorizing the training set to achieve perfect fit on observed examples

c)

Fitting highly complex models to capture every fluctuation in the training data

d)

Relying solely on large datasets without considering inductive bias

64.

Bias-variance tradeoff implies:

a)

Lower bias always improves accuracy

b)

Lower variance always improves accuracy

c)

Optimal balance is required

d)

Bias and variance are independent

65.

Increasing model complexity generally:

a)

Increases bias

b)

Decreases variance

c)

Decreases bias and increases variance

d)

Improves generalization always

66.

Which criterion selects the attribute that best separates classes?

a)

Entropy

b)

Information Gain

c)

Probability

d)

Variance

67.

Decision tree learning is an example of:

a)

Lazy learning

b)

Eager learning

c)

Reinforcement learning

d)

Unsupervised learning

68.

A tree with zero depth corresponds to:

a)

Perfect classifier

b)

Random classifier

c)

Single node tree

d)

Overfitted model

69.

Which stopping condition helps avoid overfitting?

a)

All attributes used

b)

Node purity threshold

c)

Maximum depth restriction

d)

All of the above

70.

Which scenario leads to the deepest decision tree?

a)

Few attributes, balanced classes

b)

Many irrelevant attributes

c)

Linearly separable data

d)

Small dataset

71.

A hypothesis space that cannot represent the target function leads to:

a)

Overfitting

b)

Underfitting

c)

High variance

d)

Noise sensitivity

72.

Which representation can model XOR easily?

a)

Linear classifier

b)

Decision tree

c)

Single-layer perceptron

d)

Naïve Bayes

73.

Increasing hypothesis expressiveness affects learning by:

a)

Reducing training error

b)

Increasing search complexity

c)

Increasing overfitting risk

d)

All of the above

74.

Which representation supports recursive partitioning?

a)

Neural networks

b)

Decision trees

c)

Rule-based systems

d)

Linear regression

75.

A hypothesis that predicts the same output for all inputs has:

a)

Zero bias

b)

Zero variance

c)

High bias

d)

High complexity

76.

Heuristic search is preferred because exhaustive search is:

a)

Impossible

b)

Inaccurate

c)

Computationally expensive

d)

Unreliable

77.

Greedy algorithms may fail due to:

a)

High bias

b)

Getting trapped in local optima

c)

Overfitting

d)

Noise

78.

Search space in learning refers to:

a)

Input space

b)

Feature space

c)

Hypothesis space

d)

Output space

79.

A heuristic that underestimates cost is called:

a)

Consistent

b)

Inadmissible

c)

Admissible

d)

Complete

80.

Which factor increases heuristic effectiveness?

a)

Randomness

b)

Domain knowledge

c)

Noise

d)

High dimensionality

81.

If adding training data does not reduce error, the likely cause is:

a)

Overfitting

b)

High bias

c)

Data leakage

d)

Noise

82.

A learning algorithm that updates hypothesis after every example is:

a)

Batch learning

b)

Online learning

c)

Unsupervised learning

d)

Lazy learning

83.

Decision tree pruning mainly addresses:

a)

Bias

b)

Variance

c)

Noise

d)

Data imbalance

84.

If hypothesis space is too small, the learner suffers from:

a)

Overfitting

b)

High variance

c)

Underfitting

d)

Noise sensitivity