Machine Learning

MID-2 Objective
1. What is a Decision Tree in machine learning?
a) A graphical representation of possible solutions to a decision based on conditions
b) A type of clustering algorithm
c) A mathematical representation of neural networks
d) A linear regression model
Answer: a) A graphical representation of possible solutions to a decision based on
conditions

2. Which of the following is used as a splitting criterion in Decision Trees?

a) Mean squared error
b) Entropy and Information Gain
c) K-Means
d) Naive Bayes
Answer: b) Entropy and Information Gain

3. What is the main advantage of using a Decision Tree?

a) It cannot overfit the data
b) It requires a lot of data preprocessing
c) It is easy to interpret and understand
d) It always produces a linear decision boundary
Answer: c) It is easy to interpret and understand

4. What does "pruning" in Decision Trees mean?

a) Adding more branches to the tree
b) Reducing the size of the tree by removing unnecessary branches
c) Growing the tree as large as possible
d) Performing feature scaling
Answer: b) Reducing the size of the tree by removing unnecessary branches

5. Gini Index is used to measure:

a) The entropy of a dataset
b) The homogeneity of a dataset
c) The accuracy of the decision tree
d) The regression error
Answer: b) The homogeneity of a dataset

6. What type of problem can a Decision Tree solve?

a) Only Classification
b) Only Regression
c) Both Classification and Regression
d) Clustering
Answer: c) Both Classification and Regression

7. Which algorithm is used to build a Decision Tree?

a) Backpropagation Algorithm
b) ID3, C4.5, and CART
c) KNN Algorithm
d) Apriori Algorithm
Answer: b) ID3, C4.5, and CART

8. Which of the following causes overfitting in a Decision Tree?

a) Too many levels (deep tree)
b) Shallow trees with fewer splits
c) Pruning of nodes
d) Using Information Gain for splitting
Answer: a) Too many levels (deep tree)

9. Entropy in Decision Trees is measured in terms of:

a) Gini Impurity
b) Information Gain
c) Probability and log base 2
d) Mean squared error
Answer: c) Probability and log base 2

10. What is a leaf node in a Decision Tree?

a) The starting node of the tree
b) A decision-making node with children
c) The final node that represents the output
d) A node used to prune the tree
Answer: c) The final node that represents the output

11. In CART (Classification and Regression Trees), which splitting measure is used for
regression tasks?
a) Entropy
b) Gini Index
c) Mean Squared Error
d) Chi-Square
Answer: c) Mean Squared Error
12. Which of the following is NOT a benefit of Decision Trees?
a) Easy to interpret and visualize
b) Works with both numerical and categorical data
c) Computationally inexpensive for large datasets
d) Can handle missing values
Answer: c) Computationally inexpensive for large datasets

13. What happens when a Decision Tree is overfitted?

a) It performs well on the training data but poorly on test data
b) It performs well on both training and test data
c) It uses very few nodes
d) It fails to split the data
Answer: a) It performs well on the training data but poorly on test data

14. What is "Information Gain" in Decision Trees?

a) The difference in Gini Index before and after splitting
b) The reduction in entropy after splitting
c) The increase in prediction error
d) The total number of nodes in a tree
Answer: b) The reduction in entropy after splitting

15. What does "CART" stand for?

a) Classification And Regression Tree
b) Cluster Analysis and Regression Tree
c) Classification Algorithm for Rapid Testing
d) Conditional Analysis Regression Tree
Answer: a) Classification And Regression Tree

16. Which of the following can help reduce overfitting in Decision Trees?
a) Increasing tree depth
b) Pruning the tree
c) Using more features
d) Reducing the size of training data
Answer: b) Pruning the tree

17. A Decision Tree with a maximum depth of 1 is called:

a) A Root Tree
b) A Stump
c) A Pruned Tree
d) A Binary Tree
Answer: b) A Stump

18. What is "Random Forest"?

a) A single deep Decision Tree
b) A collection of Decision Trees trained using bagging
c) A clustering algorithm
d) A linear model
Answer: b) A collection of Decision Trees trained using bagging

19. Which of the following methods helps Decision Trees generalize better?
a) Increasing tree depth
b) Removing all features
c) Cross-validation
d) Using a small training dataset
Answer: c) Cross-validation

20. In Decision Trees, what is the significance of "root node"?

a) It is the starting point of the tree where the first split happens
b) It is the most frequent output of the tree
c) It has no significant role
d) It stores the labels of the dataset
Answer: a) It is the starting point of the tree where the first split happens

21. What is the main inspiration behind Genetic Algorithms?

a) Human brain's learning process
b) Evolution and natural selection
c) Statistical models
d) Mathematical optimization techniques
Answer: b) Evolution and natural selection

22. Which of the following is NOT a component of a Genetic Algorithm?

a) Selection
b) Crossover
c) Mutation
d) Clustering
Answer: d) Clustering

23. What does the term "fitness function" refer to in Genetic Algorithms?
a) A function used to evaluate the performance of individuals
b) A function to perform mutation operations
c) A function to eliminate weak individuals
d) A function used to split datasets
Answer: a) A function used to evaluate the performance of individuals

24. What is the role of crossover in Genetic Algorithms?

a) To introduce random changes in the offspring
b) To evaluate the fitness of individuals
c) To combine the genetic material of two parents to produce offspring
d) To eliminate the weakest individuals
Answer: c) To combine the genetic material of two parents to produce offspring

25. What type of problems are Genetic Algorithms best suited for?
a) Linear programming problems
b) Problems with a large search space and no known optimal solution
c) Sorting and searching problems
d) Problems with exact mathematical solutions
Answer: b) Problems with a large search space and no known optimal solution

26. What does mutation operation do in Genetic Algorithms?

a) Evaluates the fitness of individuals
b) Introduces random changes in individuals to maintain diversity
c) Selects the fittest individuals
d) Creates copies of the individuals
Answer: b) Introduces random changes in individuals to maintain diversity

27. Which selection technique gives a higher chance of selection to individuals with
higher fitness?
a) Roulette Wheel Selection
b) Random Selection
c) Uniform Selection
d) Exhaustive Search
Answer: a) Roulette Wheel Selection

28. In Genetic Algorithms, a population is composed of:

a) Only a single individual
b) A set of candidate solutions
c) Fitness values
d) Randomly generated numbers
Answer: b) A set of candidate solutions
29. What is "elitism" in Genetic Algorithms?
a) Replacing the entire population with new offspring
b) Preserving the best individuals of the current generation into the next generation
c) Ensuring all individuals are equally likely to survive
d) Introducing random changes to every individual
Answer: b) Preserving the best individuals of the current generation into the next
generation

30. What is the significance of the initial population in Genetic Algorithms?

a) It must contain only the optimal solution
b) It determines the diversity of solutions and search space exploration
c) It is not important as the algorithm generates new solutions
d) It contains the fitness function
Answer: b) It determines the diversity of solutions and search space exploration

31. Which of the following is an advantage of Genetic Algorithms?

a) They always find the global optimum solution
b) They can handle complex search spaces and optimization problems
c) They require very little computational power
d) They do not need a fitness function
Answer: b) They can handle complex search spaces and optimization problems

32. The termination condition of a Genetic Algorithm is usually defined by:

a) Reaching a maximum number of generations
b) Finding a solution with acceptable fitness
c) Reaching a stagnation point where no improvement occurs
d) Any of the above
Answer: d) Any of the above
33. In Genetic Algorithms, what does a "chromosome" represent?
a) The entire population
b) A candidate solution
c) A random mutation
d) The fitness function
Answer: b) A candidate solution

34. Which type of crossover swaps segments of genetic material between parents?
a) Point Mutation
b) Single-Point Crossover
c) Uniform Mutation
d) Random Replacement
Answer: b) Single-Point Crossover

35. What is the primary objective of mutation in Genetic Algorithms?

a) To create identical offspring
b) To introduce diversity and avoid local optima
c) To improve fitness values directly
d) To perform fitness evaluation
Answer: b) To introduce diversity and avoid local optima

36. Which of the following is NOT a type of mutation in Genetic Algorithms?

a) Bit-flip mutation
b) Gaussian mutation
c) Heuristic mutation
d) Polynomial mutation
Answer: c) Heuristic mutation

37. Genetic Algorithms belong to which category of algorithms?

a) Deterministic Algorithms
b) Gradient-based Algorithms
c) Evolutionary Algorithms
d) Greedy Algorithms
Answer: c) Evolutionary Algorithms
38. Which of the following is true about Genetic Algorithms?
a) They are always deterministic
b) They explore and exploit the search space using probabilistic operations
c) They are used only for classification problems
d) They do not require any parameters
Answer: b) They explore and exploit the search space using probabilistic operations

39. What is the purpose of a "generation" in Genetic Algorithms?

a) To evaluate individuals one at a time
b) To simulate a cycle of selection, crossover, and mutation
c) To define the stopping criteria for the algorithm
d) To calculate the fitness function
Answer: b) To simulate a cycle of selection, crossover, and mutation

40. Genetic Algorithms are generally used in:

a) Linear optimization
b) Heuristic optimization problems
c) Exact algorithms
d) Sorting algorithms
Answer: b) Heuristic optimization problems
41. A Markov Decision Process is defined by which key components?
a) States, Actions, Rewards, Transition Probabilities, Policy
b) States, Actions, Q-values, Hyperparameters
c) Rewards, Loss, Activation Functions
d) Actions, Gradients, Learning Rate
Answer: a) States, Actions, Rewards, Transition Probabilities, Policy

42. In MDP, the probability of transitioning from one state to another depends on
_______________.
a) The entire history of past states
b) Only the current state and action
c) The rewards collected so far
d) The future states
Answer: b) Only the current state and action

43. The Bellman Equation in MDP is used to compute the _______________.

a) Immediate reward for an action
b) Value of a state under a policy
c) Optimal learning rate
d) Distance between states
Answer: b) Value of a state under a policy
44. A policy in an MDP is a function that maps _______________.
a) States to rewards
b) States to actions
c) Actions to rewards
d) Rewards to states
Answer: b) States to actions

45. In MDP, which algorithm is used to find the optimal policy?

a) Q-Learning
b) Principal Component Analysis
c) Backpropagation
d) Genetic Algorithms
Answer: a) Q-Learning

46. Which property defines a Markov Process?

a) Independence from rewards
b) The Markov Property (memorylessness)
c) State-action divergence
d) Random state initialization
Answer: b) The Markov Property (memorylessness)

47. What is the role of "discount factor" in MDP?

a) It prioritizes immediate rewards over future rewards.
b) It removes uncertainty in transitions.
c) It prevents overfitting in policies.
d) It computes the probability of each action.
Answer: a) It prioritizes immediate rewards over future rewards.

48. In an MDP, if the reward structure is deterministic, what does it imply?

a) Actions lead to fixed rewards without randomness.
b) Rewards change based on policy.
c) States transition randomly.
d) States cannot have rewards.
Answer: a) Actions lead to fixed rewards without randomness.

49. What is the purpose of the value function in MDP?

a) To determine the immediate reward
b) To measure the total expected reward from a state
c) To evaluate the policy's gradient
d) To minimize the transition cost
Answer: b) To measure the total expected reward from a state

50. Which of the following algorithms solves MDPs by iteratively improving the
policy?
a) Policy Iteration
b) K-Means Clustering
c) Gradient Descent
d) Linear Regression
Answer: a) Policy Iteration

51. Isomap is a _______________ dimensionality reduction technique.

a) Linear
b) Nonlinear
c) Supervised
d) Hybrid
Answer: b) Nonlinear
52. Isomap is an extension of which dimensionality reduction method?
a) t-SNE
b) PCA
c) LLE (Locally Linear Embedding)
d) Q-Learning
Answer: b) PCA

1. Genetic Algorithms are inspired by the principles of _______________.

Answer: Evolution and Natural Selection
2. __________________ algorithm is used to build a Decision Tree.
3. In ID3, which splitting measure is used for regression tasks ________________
4. What does CART stand for? ________________________
5. In a Genetic Algorithm, a chromosome represents a _______________.
Answer: Candidate Solution
6. The function that evaluates how "good" a solution is called the _______________.
Answer: Fitness Function
7. _______________ is the operation that combines genetic material from two
parents to create offspring.
Answer: Crossover
8. _______________ is used to introduce diversity in the population by randomly
altering genes.
Answer: Mutation
9. In Genetic Algorithms, the _______________ operator selects individuals for the
next generation based on their fitness.
Answer: Selection
10. _______________ is a technique where the best individuals are preserved to
ensure good solutions are not lost.
Answer: Elitism
11. The initial population in Genetic Algorithms is usually generated
_______________.
Answer: Randomly
12. In the Roulette Wheel Selection, individuals with higher _______________ have a
higher probability of being selected.
Answer: Fitness
13. The termination condition for a Genetic Algorithm could be reaching a
maximum _______________ or achieving an acceptable solution.
Answer: Number of Generations
14. The process of _______________ avoids local optima by introducing new genetic
material.
Answer: Mutation
15. A Genetic Algorithm evolves solutions over successive _______________.
Answer: Generations
16. The _______________ function measures how well an individual solution performs
relative to the objective.
Answer: Fitness
17. Crossover can be of different types, such as single-point, _______________, or
uniform crossover.
Answer: Multi-point
18. _______________ helps to maintain diversity in the population and prevents
premature convergence.
Answer: Mutation
19. GAs are categorized as _______________ algorithms because they rely on
probabilistic search.
Answer: Evolutionary
20. A _______________ is the smallest unit of genetic material in Genetic Algorithms.
Answer: Gene
21. In Binary Encoding, a chromosome is represented as a string of _______________.
Answer: 0s and 1s
22. _______________ is the process of reducing the size of the search space by
retaining only the best individuals.
Answer: Selection
23. In Genetic Algorithms, _______________ refers to splitting chromosomes at
specific points to exchange segments.
Answer: Crossover

24. The Markov Decision Process consists of a set of _______________, a set of actions,
a reward function, and transition probabilities.
Answer: States
25. In MDP, the Markov Property states that the next state depends only on the
_______________ and the current action.
Answer: Current state

26. The function that determines the expected reward of a state is called the
_______________ function.
Answer: Value

27. A policy in MDP maps each state to an _______________.

Answer: Action

28. The _______________ factor determines the importance of future rewards relative
to immediate rewards.
Answer: Discount

29. The Bellman Equation describes the relationship between a state and its
_______________ states in terms of value.
Answer: Successor

30. The transition probability in an MDP represents the probability of moving to a

new state given the _______________ and the current state.
Answer: Action

31. The goal of solving an MDP is to find the _______________ that maximizes the total
reward over time.
Answer: Optimal policy

32. A finite MDP has a finite number of _______________ and actions.

Answer: States

33. The reward function assigns a _______________ to each state or state-action pair.
Answer: Numerical value

34. Isomap is a _______________ dimensionality reduction technique.

Answer: Nonlinear

35. Isomap is effective for datasets with a _______________ structure embedded in a

high-dimensional space.
Answer: Nonlinear

36. The primary limitation of Isomap is its inability to handle _______________ data
manifolds.
Answer: Disconnected

37. Isomap is an extension of _______________, which focuses on preserving linear

distances.
Answer: Principal Component Analysis (PCA)

38. The computational complexity of Isomap depends on the number of

_______________ in the dataset.
Answer: Samples