UNIT-1

FILL IN THE BLANKS

1.Machine learning is divided into three main types: supervised learning, ___, and _____
learning.

Answer: unsupervised learning , reinforcement learning

2. In supervised learning, the model is trained using ____ data to make predictions.

Answer: labelled data

3. A ____ in artificial neural networks mimics the function of a biological neuron.

Answer: neuron

4. The process of ____ in a learning system involves updating the model's parameters to
improve its performance.

Answer: training

5. One of the key issues in machine learning is the balance between bias and ____.

Answer: variance

6. Data set means_____.

Answer: collection of data

7. The goal of concept learning is to infer a ____ based on positive and negative
examples.

Answer: hypothesis

8. Concept learning can be viewed as a search through a ____ space of possible

hypotheses.

Answer: Hypothesis

9. A maximally ____ hypothesis is the one that is specific enough to cover only positive
examples and exclude all negative ones.

Answer: specific

10. who introduced the hebb network_____.

Answer: Donald hebb

11. The ____ space in concept learning represents all the hypotheses consistent with
the training examples.
Answer: version

12. A ____ discriminant is used to classify data by separating it into different classes
using a linear function.

Answer: linear

13. The perceptron algorithm is a simple linear ____ used for binary classification tasks.

Answer: Classifier

14. Data is considered ____ separable if it can be divided by a single straight line or
hyperplane.

Answer: linearly

15. ____ regression is used to predict a continuous target variable by fitting a straight
line to the data.

Answer: Linear

16. The ____ method is used in linear regression to minimize the sum of squared errors
between observed and predicted values.

Answer: Least squares

17. In linear discriminants, data points are classified by projecting them onto a ____.

Answer: Hyperplane

18. A perceptron updates its ____ based on the error between its predicted output and
the actual output.

Answer: Weights

19. ____ learning occurs without labeled data, where the model tries to learn the
structure or patterns within the input data.

Answer: Unsupervised Learning

20. The ____ of a machine learning system refers to its ability to perform well on unseen
or new data.

Answer: Generalization

21. The trade-off between ____and _______ is a common issue in machine learning
model design.

Answer: complexity , generalization

22. In supervised learning, each training example consists of an input-output pair,
where the output is known as a ____ label.

Answer: Target

23. A hypothesis is said to be ____ if it correctly classifies all the training examples.

Answer: Consistent

24. The ____ elimination algorithm refines the version space by progressively eliminating
inconsistent hypotheses.

Answer: Candidate

25. ____ refers to the process of iteratively improving a model's performance by

adjusting its parameters based on feedback from the training data.

Answer: Learning
 MULTIPLE CHOICE QUESTIONS

1. Which of the following is not a type of machine learning?

a) Supervised learning

b) Unsupervised learning

c) Reinforcement learning

d) Reactive learning

Answer: d) Reactive learning

2. In supervised learning, the output variable is also known as the:

a) Feature

b) Label

c) Hidden node

d) Bias

Answer: b) Label

3. The primary inspiration for artificial neural networks comes from:

a) Genetic algorithms

b) Evolutionary computation

c) The structure of the human brain

d) Decision trees

Answer: c) The structure of the human brain

4. In a neural network, what do the weights between neurons represent?

a) The strength of the connection between neurons

b) The number of neurons in the layer

c) The total number of hidden layers

d) The output value of the network

Answer: a) The strength of the connection between neurons

5. Which of the following is a key component of a machine learning system?

a) Clustering algorithm
b) Data acquisition and preprocessing

c) Game theory analysis

d) Heuristic search

Answer: b) Data acquisition and preprocessing

6. Which issue arises when a model fits the training data too well but performs poorly on
unseen data?

a) Overfitting

b) Underfitting

c) Generalization

d) Linear separability

Answer: a) Overfitting

7. Which of the following is a challenge in machine learning?

a) Minimizing the use of labeled data

b) Ensuring that models overfit

c) Balancing bias and variance

d) Ignoring test data during training

Answer: c) Balancing bias and variance

8. In concept learning, the task is typically to infer:

a) A data structure

b) A target concept

c) A clustering algorithm

d) An unsupervised pattern

Answer: b) A target concept

9. Concept learning can be framed as a search through which of the following?

a) Search tree

b) Hypothesis space

c) Decision matrix

d) Linear classifier
Answer: b) Hypothesis space

10. The most specific hypothesis in a concept learning task is the one that:

a) Covers all training examples

b) Covers all positive examples and excludes as many negative examples as possible

c) Is identical to the most general hypothesis

d) Contains all the features of the dataset

Answer: b) Covers all positive examples and excludes as many negative examples as
possible

11. The Candidate Elimination Algorithm works by maintaining which two boundaries?

a) Training and test boundaries

b) Feature and target boundaries

c) Specific and general hypotheses

d) Input and output boundaries

Answer: c) Specific and general hypotheses

12. In the version space, which hypotheses are considered?

a) Only the most general hypotheses

b) Only the most specific hypotheses

c) All hypotheses consistent with the training data

d) Hypotheses inconsistent with training data

Answer: c) All hypotheses consistent with the training data

13. Linear discriminants are used to:

a) Perform clustering

b) Classify data by separating it using a linear decision boundary

c) Generate new features from existing data

d) Predict continuous output values

Answer: b) Classify data by separating it using a linear decision boundary

14. Which of the following is a limitation of the perceptron?

a) It can only be used for regression tasks

b) It cannot handle non-linearly separable data

c) It requires labelled data for training

d) It can only solve multi-class classification problems

Answer: b) It cannot handle non-linearly separable data

15. Linear separability implies that:

a) Data points can be separated by a straight line or hyperplane

b) All data points have the same label

c) Data can be clustered without supervision

d) The decision boundary must be non-linear

Answer: a) Data points can be separated by a straight line or hyperplane

16. Which algorithm can be used to classify data that is linearly separable?

a) K-means clustering

b) Perceptron

c) Naive Bayes

d) Principal Component Analysis

Answer: b) Perceptron

17. Linear regression is primarily used to predict:

a) Categorical values

b) Continuous values

c) Classes of data points

d) Clusters in unsupervised learning

Answer: b) Continuous values

18. In linear regression, the goal is to minimize the:

a) Number of features in the dataset

b) Squared differences between observed and predicted values

c) Number of iterations during training

d) Distance between hyperplanes

Answer: b) Squared differences between observed and predicted values

19. Which of the following best describes the process of training in supervised learning?

a) Learning without the use of labels

b) Using labelled data to adjust the model’s parameters

c) Building a model without any human intervention

d) Grouping data points into clusters

Answer: b) Using labelled data to adjust the model’s parameters

20. A common method used in linear regression to optimize the model is:

a) Gradient Descent

b) Decision Trees

c) K-means Clustering

d) Perceptron

Answer: a) Gradient Descent

21. What does the perceptron learning rule update based on each mistake in
classification?

a) Bias term

b) Weights

c) Learning rate

d) Feature values

Answer: b) Weights

22. The goal of version spaces in concept learning is to:

a) Identify the smallest set of features

b) Identify all hypotheses consistent with the training data

c) Determine which data points are outliers

d) Maximize the decision boundary

Answer: b) Identify all hypotheses consistent with the training data

23. The candidate elimination algorithm is used to:

a) Reduce the size of the training dataset

b) Eliminate inconsistent hypotheses in version space

c) Select the most complex hypothesis

d) Predict continuous output variables

Answer: b) Eliminate inconsistent hypotheses in version space

24. Which of the following is not typically a part of designing a learning system?

a) Data preprocessing

b) Selecting features

c) Ignoring the performance evaluation phase

d) Choosing an appropriate learning algorithm

Answer: c) Ignoring the performance evaluation phase

25. In machine learning, generalization refers to:

a) The ability of a model to fit the training data perfectly

b) The model's performance on unseen data

c) Reducing the number of features in a dataset

d) The process of overfitting a model

Answer: b) The model's performance on unseen data