Q.1 Define Machine Learning.

Explain how Supervised learning is different from

Unsupervised learning.
Machine Learning (ML) is a subset of artificial intelligence (AI) that involves training
computers to learn from data, without being explicitly programmed. In other words, machine
learning algorithms are designed to automatically identify patterns and relationships in data,
and use these insights to make predictions or decisions.

There are several different types of machine learning, but two of the most common are
supervised learning and unsupervised learning.

Supervised learning is a type of machine learning in which the algorithm is trained on a

labeled dataset, meaning that each data point has a corresponding label or output value. The
goal of supervised learning is to build a model that can accurately predict the output value for
new, unseen data points.

For example, a supervised learning algorithm might be trained on a dataset of housing prices,
with input features such as square footage, number of bedrooms, and location, and output
labels representing the actual sale prices. Once trained, the model can be used to predict the
sale price for new houses based on their input features.

Unsupervised learning, on the other hand, is a type of machine learning in which the
algorithm is trained on an unlabeled dataset, meaning that there are no corresponding output
labels. Instead, the algorithm is designed to identify patterns or structure in the data on its
own.

For example, an unsupervised learning algorithm might be trained on a dataset of customer

purchase histories, with input features such as product types and purchase dates. The
algorithm might then use clustering or dimensionality reduction techniques to identify groups
of customers with similar purchasing habits, or to identify the most important features for
predicting customer behavior.

In summary, supervised learning requires labeled data to train models to make accurate
predictions, while unsupervised learning doesn't require labeled data and instead focuses on
identifying patterns or structure in the data on its own.

Q.2 Explain the steps of developing Machine Learning applications.

Developing a machine learning application involves several steps, which are generally as
follows:

1. Define the Problem: The first step is to clearly define the problem you want to solve
with machine learning. This involves understanding the business problem or use case,
defining the input data and desired output, and selecting the appropriate machine
learning techniques to achieve the desired results.
2. Collect and Prepare Data: The next step is to collect and prepare the data required for
the machine learning algorithm. This involves gathering and cleaning the data,
ensuring that it is in the correct format, and selecting the relevant features to use in the
model.
 3. Train the Model: Once the data is prepared, the next step is to train the machine
learning model using an appropriate algorithm. This involves selecting the appropriate
machine learning algorithm, dividing the data into training and validation sets, and
fine-tuning the algorithm parameters to optimize performance.
4. Evaluate the Model: After training the model, it is important to evaluate its
performance using appropriate metrics and validation techniques. This helps to ensure
that the model is accurate and can generalize well to new data.
5. Deploy the Model: Once the model has been trained and evaluated, it can be deployed
into production. This involves integrating the model into a software system or
application, and ensuring that it can handle real-world inputs and produce accurate
results.
6. Monitor and Improve the Model: Finally, it is important to monitor the performance
of the model in production and continuously improve it over time. This may involve
updating the model with new data or adjusting the algorithm parameters to improve
accuracy and efficiency.

Overall, developing a machine learning application is an iterative process that involves

several steps, from defining the problem and collecting data to training the model, evaluating
its performance, and deploying it into production. With careful planning and execution,
machine learning can be a powerful tool for solving complex business problems and driving
innovation.

Q.3 Explain training, testing and validation datasets, cross validation, overfitting &
underfitting of model.
Training, testing, and validation datasets are all important components of the machine
learning workflow, which are used to evaluate and optimize the performance of a model.

1. Training dataset: This is the portion of the data used to train the machine learning
model. It typically includes labeled examples of input data and their corresponding
output values. During training, the model is adjusted to minimize the difference
between its predicted output and the actual output.
2. Testing dataset: This is a separate portion of the data that is used to evaluate the
performance of the trained model. It is used to measure the accuracy of the model's
predictions on new, unseen data.
3. Validation dataset: This is a portion of the data that is used to tune the
hyperparameters of the model. Hyperparameters are the configuration settings for the
model that cannot be learned from the training data. The validation dataset is used to
optimize these settings to improve the model's performance on new data.

Cross-validation is a technique used to evaluate the performance of a machine learning model

by splitting the available data into multiple subsets, or "folds", and using each fold as both a
training and testing dataset. This helps to ensure that the model is not overfitting to a specific
subset of the data, and provides a more robust estimate of its performance on new data.

Overfitting occurs when a machine learning model is too complex and fits the training data
too closely, resulting in poor generalization to new data. This often occurs when a model has
too many parameters relative to the size of the training dataset, and can be addressed by using
techniques such as regularization or reducing the complexity of the model.
Underfitting occurs when a machine learning model is too simple and cannot capture the
underlying patterns in the data. This often occurs when the model is not trained for long
enough or has too few parameters to capture the complexity of the data. Underfitting can be
addressed by increasing the complexity of the model or training it for longer periods.

Q.4 Explain different Performance Metrics with proper illustrations.

Performance metrics are used to evaluate the effectiveness of a machine learning model in
solving a particular task. Here are some common performance metrics and how they are used:

1. Accuracy: This is the most commonly used performance metric and measures the
percentage of correct predictions made by the model. It is defined as:
accuracy = (number of correct predictions) / (total number of predictions)
For example, if a model predicts 80 out of 100 test examples correctly, its accuracy is
80%.
2. Precision and Recall: Precision and recall are used to evaluate the performance of a
model on imbalanced datasets, where one class may have many more examples than
the other.
Precision measures the percentage of positive predictions that are correct. It is defined
as:
precision = (true positives) / (true positives + false positives)
Recall measures the percentage of actual positives that are correctly predicted. It is
defined as:
recall = (true positives) / (true positives + false negatives)
For example, in a medical diagnosis task, high precision means that few patients are
incorrectly diagnosed with a disease, while high recall means that few patients with
the disease are missed.
3. F1 Score: The F1 score is a combination of precision and recall that provides a
balanced evaluation of the model's performance. It is defined as the harmonic mean of
precision and recall, and is given by:
F1 score = 2 * (precision * recall) / (precision + recall)
The F1 score ranges from 0 to 1, with higher values indicating better performance.
4. ROC Curve and AUC: The Receiver Operating Characteristic (ROC) curve is used to
evaluate the performance of a binary classifier by plotting the true positive rate (TPR)
against the false positive rate (FPR) at different decision thresholds. The area under
the ROC curve (AUC) is a measure of the overall performance of the classifier, with
higher values indicating better performance.
The following figure shows an example ROC curve for a binary classifier:

The dotted line represents a random classifier, while the curve for the actual classifier
is shown in blue. The closer the curve is to the upper-left corner, the better the
performance of the classifier.

Overall, performance metrics are an important tool for evaluating the effectiveness of
machine learning models and selecting the best one for a particular task. Different metrics
may be more appropriate depending on the specific requirements of the task at hand.
Q.5 Write a short note on- a) Issues in Machine Learning. b) Machine Learning Applications.
c) Diagonalization of Matrix
a) Issues in Machine Learning:

1. Bias and Fairness: Machine learning models can exhibit bias and discriminate against
certain groups of people or types of data. This can result in unfair outcomes and
negative impacts on individuals or society as a whole.
2. Overfitting and Underfitting: Machine learning models can suffer from overfitting or
underfitting, which can lead to poor generalization and performance on new data.
3. Data Quality and Quantity: Machine learning models are only as good as the data they
are trained on. Poor quality or insufficient data can result in inaccurate or biased
models.
4. Explainability and Transparency: Many machine learning models are complex and
difficult to interpret, making it difficult to understand how they make predictions and
to detect errors or biases.

b) Machine Learning Applications:

Machine learning has many practical applications across a wide range of fields, including:

1. Natural Language Processing: Machine learning is used to power applications such as

chatbots, language translation, and sentiment analysis.
2. Image and Video Recognition: Machine learning is used to identify objects, people,
and activities in images and videos, and is used in applications such as self-driving
cars and security cameras.
3. Healthcare: Machine learning is used to analyze medical images, predict disease
outcomes, and develop personalized treatment plans.
4. Finance: Machine learning is used to predict stock prices, detect fraud, and make loan
decisions.

c) Diagonalization of Matrix:

Diagonalization is a process of transforming a matrix into a diagonal matrix. A diagonal

matrix is a square matrix where all non-diagonal elements are zero. To diagonalize a matrix,
one needs to find a set of eigenvectors and eigenvalues of the matrix.

If A is a square matrix, and λ is an eigenvalue of A, then there exists a nonzero vector x such
that Ax = λx. The set of all eigenvectors of A forms a basis for the vector space, and the
corresponding eigenvalues form the diagonal elements of the diagonal matrix.

Diagonalization is useful in many areas of mathematics and physics, including solving

systems of linear differential equations, finding the normal modes of vibration in a
mechanical system, and determining the principal axes of an object. It also has applications in
machine learning, such as in principal component analysis (PCA), which is a technique for
reducing the dimensionality of high-dimensional datasets.

Q.6 Explain Symmetric Positive Definite Matrices with example.

A symmetric positive definite (SPD) matrix is a square matrix A that satisfies the following
two conditions:
1. A is symmetric, i.e., A = A^T, where A^T denotes the transpose of A.
2. For any nonzero vector x, x^T A x > 0, where x^T denotes the transpose of x.

In other words, an SPD matrix is a symmetric matrix whose eigenvalues are all positive. This
implies that an SPD matrix has many desirable properties, such as being invertible and
having a unique Cholesky decomposition, which is a factorization of A as A = LL^T, where
L is a lower triangular matrix with positive diagonal entries.

Here's an example of an SPD matrix:

A = [3 1; 1 4]

To show that A is symmetric, we compute its transpose:

A^T = [3 1; 1 4]^T = [3 1; 1 4]

Since A = A^T, A is symmetric.

To show that A is positive definite, we need to check that x^T A x > 0 for any nonzero vector
x. Let's take x = [1; 2]:

x^T A x = [1 2] [3 1; 1 4] [1; 2] = [1 2] [5; 9] = 23

Since x^T A x is positive, A is positive definite.

SPD matrices have many applications in mathematics, physics, and engineering, including in
optimization problems, linear systems of equations, and finite element analysis. They are also
commonly used in machine learning algorithms, such as the Gaussian process and kernel
methods.

Q.7 Explain the terms: Norms, Inner products, Length of Vector, Determinant and trace.
Norms, inner products, length of vector, determinant, and trace are all mathematical concepts
that are frequently used in linear algebra and have applications in various fields, including
machine learning.

1. Norms: A norm is a mathematical function that assigns a scalar value to a vector,

representing its "length" or "size". The most common norm used is the Euclidean
norm or L2 norm, which is defined as ||x|| = sqrt(x1^2 + x2^2 + ... + xn^2), where x is
a vector of length n.
2. Inner Products: An inner product is a mathematical operation that takes two vectors
and returns a scalar value. The most common inner product used is the dot product,
which is defined as x.y = x1y1 + x2y2 + ... + xn*yn, where x and y are vectors of
length n.
3. Length of Vector: The length of a vector is defined as the norm of the vector, which
represents its "size" or "magnitude". The length of a vector is always non-negative.
 4. Determinant: The determinant is a scalar value associated with a square matrix. It is
used to determine various properties of the matrix, such as invertibility and
eigenvalues. The determinant is denoted as det(A) and can be computed using various
methods, such as cofactor expansion and Gaussian elimination.
5. Trace: The trace is a scalar value associated with a square matrix. It is defined as the
sum of the diagonal elements of the matrix. The trace is denoted as tr(A) and has
various applications, such as in computing the eigenvalues and diagonalizing the
matrix.

These concepts are used extensively in linear algebra and have many practical applications in
various fields, including machine learning. For example, norms are used to measure the
"distance" between vectors, inner products are used to measure the similarity between
vectors, and determinants and traces are used in computing various properties of matrices,
such as their eigenvalues and eigenvectors.

Q.8 Explain Singular value Decomposition (SVD) with example and give its applications.
Singular Value Decomposition (SVD) is a widely used matrix factorization technique in
linear algebra and machine learning. It decomposes a given matrix A into three matrices as A
= UΣV^T, where U and V are orthogonal matrices and Σ is a diagonal matrix with non-
negative entries called singular values.

Here's an example of SVD for a 3x2 matrix A:

A = [1 2; 2 3; 3 4]

We can compute the SVD of A as follows:

1. Compute A^T A:

A^T A = [1 2; 2 3; 3 4]^T [1 2; 2 3; 3 4] = [14 20; 20 29]

2. Compute the eigenvalues and eigenvectors of A^T A:

The eigenvalues of A^T A are λ1 = 38.61 and λ2 = 4.39. The corresponding eigenvectors are
v1 = [0.7432; 0.6696] and v2 = [-0.6696; 0.7432].

3. Compute the singular values and the columns of U:

The singular values of A are the square roots of the eigenvalues of A^T A, i.e., σ1 = 6.21 and
σ2 = 2.10. The columns of U are the normalized eigenvectors of A^T A, i.e., u1 = [0.6293;
0.7776] and u2 = [0.7776; -0.6293].

4. Compute the columns of V:

The columns of V are given by V = A U Σ^-1, where Σ^-1 is the inverse of Σ. Thus, we have
V = [0.4036; 0.9147] and V = [-0.9147; 0.4036].
Therefore, we have A = UΣV^T = [0.6293 0.7776; 0.7776 -0.6293; 0 0] [6.21 0; 0 2.10]
[0.4036 -0.9147; 0.9147 0.4036]^T.

SVD has many applications in various fields, including machine learning, signal processing,
image compression, and recommendation systems. For example, SVD is used in
recommendation systems to factorize a user-item rating matrix into two lower-dimensional
matrices representing user and item features, respectively. SVD can also be used for image
compression by reducing the dimensionality of the image matrix while preserving its
important features. Additionally, SVD is used in data analysis to identify the most important
features in a dataset and reduce its dimensionality.

Q.9 Explain Linear Models and Linear Regression in detail.

Linear models are a class of statistical models that assume a linear relationship between the
input variables and the output variable. Linear models are used to make predictions, estimate
the strength of the relationship between the variables, and identify the most important
features in a dataset. One of the most common types of linear models is linear regression.

Linear regression is a technique for modeling the relationship between a dependent variable
Y and one or more independent variables X. In linear regression, we assume that the
relationship between Y and X is linear, i.e., Y can be expressed as a linear function of X:

Y = β0 + β1X1 + β2X2 + ... + βnXn + ε

where β0 is the intercept, β1, β2, ..., βn are the coefficients of the independent variables X1,
X2, ..., Xn, and ε is the error term.

The goal of linear regression is to estimate the values of the coefficients β0, β1, β2, ..., βn that
minimize the sum of the squared errors between the predicted values and the actual values of
the dependent variable Y. This is known as the method of least squares.

There are two main types of linear regression: simple linear regression and multiple linear
regression. Simple linear regression involves only one independent variable, while multiple
linear regression involves two or more independent variables.

To estimate the coefficients in linear regression, we use the training data to calculate the least
squares estimates of the coefficients. Once we have estimated the coefficients, we can use the
model to make predictions on new data.

Linear regression has several assumptions, including:

1. Linearity: The relationship between the independent and dependent variables is linear.
2. Independence: The observations in the data set are independent of each other.
3. Normality: The residuals (i.e., the differences between the predicted and actual values
of the dependent variable) are normally distributed.
4. Homoscedasticity: The variance of the residuals is constant across all levels of the
independent variable.
5. No multicollinearity: The independent variables are not highly correlated with each
other.
Linear regression is widely used in various fields, including finance, economics, biology, and
social sciences. Linear regression is also used in machine learning for tasks such as
regression analysis, feature selection, and model interpretation.

Q.10 The data for the midterm and final exam grades obtained for students in Machine
learning subject are as given in table below. Use the method of least squares using regression
to predict the final exam grade of a student who received 94 in the Midterm exam: Midterm
Marks(X) Final Marks(Y) 72 84 50 53 81 77 74 78 94 90 86 75 59 49 83 79 86 77 33 52
To predict the final exam grade of a student who received 94 in the midterm exam, we can
use linear regression. The first step is to calculate the regression equation, which is given by:

Y = β0 + β1X

where Y is the final exam grade, X is the midterm exam grade, β0 is the intercept, and β1 is
the slope.

To calculate the slope β1 and intercept β0, we need to first calculate the means of X and Y,
which are given by:

Mean(X) = (72 + 50 + 81 + 74 + 94 + 86 + 59 + 83 + 86 + 33)/10 = 68.8

Mean(Y) = (84 + 53 + 77 + 78 + 90 + 75 + 49 + 79 + 77 + 52)/10 = 70.4

Next, we need to calculate the sum of the products of the deviations of X and Y from their
respective means, which is given by:

Σ((X - Mean(X))(Y - Mean(Y))) = (72 - 68.8)(84 - 70.4) + (50 - 68.8)(53 - 70.4) + (81 -
68.8)(77 - 70.4) + (74 - 68.8)(78 - 70.4) + (94 - 68.8)(90 - 70.4) + (86 - 68.8)(75 - 70.4) + (59
- 68.8)(49 - 70.4) + (83 - 68.8)(79 - 70.4) + (86 - 68.8)(77 - 70.4) + (33 - 68.8)(52 - 70.4) =
11723.2

We also need to calculate the sum of the squares of the deviations of X from its mean, which
is given by:

Σ((X - Mean(X))^2) = (72 - 68.8)^2 + (50 - 68.8)^2 + (81 - 68.8)^2 + (74 - 68.8)^2 + (94 -
68.8)^2 + (86 - 68.8)^2 + (59 - 68.8)^2 + (83 - 68.8)^2 + (86 - 68.8)^2 + (33 - 68.8)^2 =
14092.8

Using these values, we can calculate the slope β1 as:

β1 = Σ((X - Mean(X))(Y - Mean(Y))) / Σ((X - Mean(X))^2) = 11723.2 / 14092.8 = 0.831

Next, we can calculate the intercept β0 as:

β0 = Mean(Y) - β1 * Mean(X) = 70.4 - 0.831 * 68.8 = 12.526

Therefore, the regression equation for the given data is:

Y = 12.526 + 0.831X

To predict the final exam grade for a student who received 94 in the midterm exam, we can
substitute X = 94 in the equation:

Y = 12.526 + 0.831 * 94 = 90.744

Therefore, the predicted final exam grade for the student who received 94 in the midterm
exam is 90.744.

Note: It is important to check the assumptions of linear regression before using the regression
equation for predictions. These assumptions include linearity, independence, normality, hom
Find the Eigen Values & Eigen Vectors of the following Matrix. 1 2 3 A= 0 -2
6 0 0 -3
To find the eigenvalues and eigenvectors of the given matrix A, we need to
solve the characteristic equation:

det(A - λI) = 0

where I is the identity matrix of the same size as A, and λ is the eigenvalue
we are trying to find.

Substituting the values of A and I, we get:

|1-λ 2 3 | | 0 -2-λ 6 | | 0 0 -3-λ| = 0

Expanding the determinant along the first row, we get:

(1-λ)[(-2-λ)(-3-λ) - 0] - 2[0 - 0] + 3[0 - 0] = 0

Simplifying, we get:

(λ-1)(λ+2)(λ+3) = 0

Therefore, the eigenvalues of A are λ1 = 1, λ2 = -2, and λ3 = -3.

To find the eigenvectors corresponding to each eigenvalue, we need to

solve the system of equations:

(A - λI)x = 0

where x is the eigenvector we are trying to find.

For λ1 = 1, we get:

(1 - 1)x1 + 2x2 + 3x3 = 0 0x1 - (2 + 1)x2 + 6x3 = 0 0x1 + 0x2 - (3 + 1)x3 = 0

Simplifying, we get:

2x2 + 3x3 = 0 -3x2 + 6x3 = 0 -4x3 = 0

Solving for x3, we get x3 = 0. Substituting this value in the second equation,
we get x2 = 0. Substituting these values in the first equation, we get x1 = -
3/2.

Therefore, the eigenvector corresponding to λ1 = 1 is:

v1 = [-3/2, 0, 0]

For λ2 = -2, we get:

(1 + 2)x1 + 2x2 + 3x3 = 0 0x1 - (2 - 2)x2 + 6x3 = 0 0x1 + 0x2 - (3 - 2)x3 = 0

Simplifying, we get:

3x1 + 2x2 + 3x3 = 0 0x2 + 6x3 = 0 0x3 = 0

Solving for x3, we get x3 = 0. Substituting this value in the second equation,
we get x2 = 0. Substituting these values in the first equation, we get x1 = -
2/3.

Therefore, the eigenvector corresponding to λ2 = -2 is:

v2 = [-2/3, 0, 0]

For λ3 = -3, we get:

(1 + 3)x1 + 2x2 + 3x3 = 0 0x1 - (2 + 3)x2 + 6x3 = 0 0x1 + 0x2 - (3 + 3)x3 =

0

Simplifying, we get:

4x1 + 2x2 + 3x3 = 0 -5x2 + 6x3 = 0 -6x3 = 0

Solving for x3