Amity Institute of Information Technology

CSIT737 - Machine Learning using Python

Module 1 - Basics
Introduction to Machine Learning, Different forms of Learning;
Basics of Probability Theory, Linear Algebra and Optimization,
Overview of Target Function representations, Types of ML
techniques, Hypothesis selection through cross validation.

Dr. Partha Sarathi Chakraborty

Amity University Uttar Pradesh Noida

1
 Amity Institute of Information Technology

Current Scenario

2
 Amity Institute of Information Technology

Development Timeline

3
 Amity Institute of Information Technology

Introduction to Machine Learning

What is
Machine See this is a keyboard.
Learning?

See one more keyboard.

Now, Identify What is this object?

Before
answering
it, look into
Human can learn from past experience few items.
and make decision of its own
4
 Amity Institute of Information Technology

Hi Mogli!
What is this
Object?

As he is human being. He
can observe and learn.
5
 Amity Institute of Information Technology

Machine follow instruction. It performs operations

such as arithmetic operation.

Training
through Data

dog dog cat cat

6
 Amity Institute of Information Technology

Dataset
This what we called as Data
or Training dataset
So, we first need to provide
training dataset to the
machine
7
 Amity Institute of Information Technology

Dataset

Common definition: Machine Learning is the study of computer algorithms that improve
automatically through experience 8
 Amity Institute of Information Technology

Common definition: Machine Learning is the study of computer algorithms that improve
automatically through experience

9
 Amity Institute of Information Technology

What is Machine Learning?

Machine learning (ML)† is a branch of artificial intelligence
(AI) focused on enabling computers and machines to imitate What is
the way that humans learn, to perform tasks autonomously, Machine
and to improve their performance and accuracy through Learning?
experience and exposure to more data.

†Source: [Link]
10
 Amity Institute of Information Technology

Improve on task T, with respect to performance

metric P, based on experience E

11
 Amity Institute of Information Technology

Traditional vs ML Approach

12
 Amity Institute of Information Technology

Machine Learning

13
 Amity Institute of Information Technology

Relate It!
No, more like gardening

• Seeds = Algorithms
• Nutrients = Data
• Gardener = You
• Plants = Model (Program)

14
 Amity Institute of Information Technology

Why Machine Learning?

• Explosive growth of data.
• Infeasibility of manual programming for
complex tasks.
• Applications in (not limited):
– Speech & image recognition
– Fraud detection
– Recommendation systems
– Autonomous vehicles

15
 Amity Institute of Information Technology

Why Machine Learning?

16
 Amity Institute of Information Technology

When?

17
 Amity Institute of Information Technology

Who?

18
 Amity Institute of Information Technology

Learning from Examples

19
 Amity Institute of Information Technology

Machine Learning Workflow

20
 Amity Institute of Information Technology

Machine Learning Workflow

21
 Amity Institute of Information Technology

Different forms of Learning

• Supervised (inductive) learning
– Training data includes desired outputs
• Unsupervised learning
– Training data does not include desired outputs
• Semi-supervised learning
– Training data includes a few desired outputs
• Reinforcement learning
– Rewards from sequence of actions
• Evolutionary Learning
– Biological evolution can be seen as learning process.

22
Amity Institute of Information Technology

23
 Amity Institute of Information Technology

Supervised Learning
• Decision tree induction
• Evaluation of classifiers
• Naïve Bayesian classification
• Naïve Bayes for text classification
• Support vector machines
• Linear regression and gradient descent
• Neural networks
• K-nearest neighbor
• Ensemble methods
24
 Amity Institute of Information Technology

An application of supervised learning

• Endless applications of supervised learning.

• An emergency room in a hospital measures 17
variables (e.g., blood pressure, heart rate, etc) of newly
admitted patients.
• A decision is needed: whether to put a new patient in
an intensive-care unit (ICU).
– Due to the high cost of ICU, those patients who may survive
less than a month are given higher priority.
• Problem: to predict high-risk patients and discriminate
them from low-risk patients.

25
 Amity Institute of Information Technology

Another application
• A credit card company receives thousands of
applications for new cards. Each application
contains information about an applicant,
– age
– annual salary
– outstanding debts
– credit rating
– etc.
• Problem: Decide whether an application should
approved, i.e., classify applications into two
categories, approved and not approved.
26
 Amity Institute of Information Technology

Supervised machine learning

• We humans learn from past experiences.
• A computer does not “experience.”
– A computer system learns from data, which represents “past
experiences” in an application domain.
• Our focus: learn a target function that can be used to
predict the values (labels) of a discrete class attribute,
e.g.,
– high-risk or low risk and approved or not-approved.
• The task is commonly called: supervised learning,
classification, or inductive learning.

27
 Amity Institute of Information Technology

The Data and the Goal

• Data: A set of data records (also called
examples, instances, or cases) described by
– k data attributes: A1, A2, … Ak.
– One class attribute: a set of pre-defined class labels
– In other words, each record/example is labelled with
a class label.
• Goal: To learn a classification model from the
data that can be used to predict the classes of
new (future or test) instances/cases.
28
 Amity Institute of Information Technology

An example: data (loan application)

Approved or not

29
 Amity Institute of Information Technology

An example: the learning task

Sub-tasks:
• Learn a classification model from the data
• Use the model to classify future loan
applications into
– Yes (approved) and
– No (not approved)
• What is the class for following applicant/case?

30
 Amity Institute of Information Technology

Supervised learning process: Two steps

• Learning or training: Learn a model using the
training data (with labels)
• Testing: Test the model using unseen test data
(without labels) to assess the model accuracy
Number of correct classifications
Accuracy = ,
Total number of test cases

31
 Amity Institute of Information Technology

Supervised Learning
• Supervised learning: classification is
supervised learning from examples.
– Supervision: The data (observations,
measurements, etc.) are labeled with pre-defined
classes, which is
– like a “teacher” gives us the classes (supervision).

32
 Amity Institute of Information Technology

Unsupervised learning
• Unsupervised learning
– Class labels of the data are not given or unknown
– Goal: Given a set of data, the task is to establish the
existence of classes or clusters in the data
• More Simply,
– There is no supervisor and only input data is available.
– The aim is now to find regularities, irregularities,
relationships, similarities and associations in the input.

33
 Amity Institute of Information Technology

Unsupervised learning
• Purpose:
– Discover similarities, anomalies, or clusters in data.

• Common Techniques:
– Clustering (e.g., K-Means, DBSCAN)
– Dimensionality Reduction (e.g., PCA, t-SNE)
– Association Rules (e.g., Market Basket Analysis)
– Pattern Mining
• It is adopted as amore general term than frequent pattern mining
or association mining.
– Outlier Detection
• It is the process of finding data examples with behaviours that are
very different from the expectation (outliers or anomalies).
34
 Amity Institute of Information Technology

Unsupervised learning
• Applications of Unsupervised Learning
– Domain Use Case Example
– E-commerce Customer segmentation
– Cybersecurity Anomaly detection
– Healthcare Patient risk profile grouping
– Marketing Product recommendations
– Finance Fraud detection, transaction grouping

• Advantages:
– No need for labeled data
– Can handle large-scale, unlabeled datasets
– Reveals unexpected patterns

35
 Amity Institute of Information Technology

Clustering
• Clustering is a technique for finding similarity
groups in data, called clusters. I.e.,
– it groups data instances that are similar to (near) each
other in one cluster and data instances that are very
different (far away) from each other into different clusters.
• Clustering is often called an unsupervised
learning task as no class values denoting an a
priori grouping of the data instances are given,
which is the case in supervised learning.

36
 Amity Institute of Information Technology

Example
• Problem:
– Group customers based on purchasing behavior.
• Data (Unlabeled):
– Age, Annual Income, Spending Score
• Steps:
– Input the data (no labels).
– Choose number of clusters (e.g., 3).
– K-Means algorithm groups customers into clusters.
• Result:
– Cluster 1: Low spenders
– Cluster 2: Moderate spenders
– Cluster 3: High spenders
• Helps business target specific groups for marketing. 37
 Amity Institute of Information Technology

Supervised learning vs. Unsupervised learning

• Supervised learning: discover patterns in the

data that relate data attributes with a target
(class) attribute.
– These patterns are then utilized to predict the values
of the target attribute in future data instances.
• Unsupervised learning: The data have no target
attribute.
– We want to explore the data to find some intrinsic
structures in them.

38
 Amity Institute of Information Technology

Semi-supervised learning
• A machine learning approach that uses a small amount of
labeled data and a large amount of unlabeled data for
training.
• It combines the strengths of supervised and unsupervised
learning.

• Key Points:
– Cost-effective: Labeling data is expensive; unlabeled data is
plentiful.
– Learns from both: Labeled data gives accuracy, unlabeled data
improves generalization.
39
 Amity Institute of Information Technology

Real-World Applications
• Speech Recognition (only a few labeled
transcriptions)
• Image Classification (millions of
unlabeled photos)
• Web Content Classification
• Medical Imaging (labels by experts are
expensive)
40
 Amity Institute of Information Technology

Example – Image Classification

• Problem:
– Classify images of cats and dogs. You have:
– 500 labeled images (250 cats, 250 dogs)
– 10,000 unlabeled images
• Steps:
– Train initial model using the 500 labeled images.
– Use model to predict labels for 10,000 unlabeled images.
– Select confident predictions and add them to labeled set (pseudo-
labeling).
– Retrain model on larger dataset.
• Result:
– Model achieves higher accuracy than training on only 500 images.
41
 Amity Institute of Information Technology

Taxonomy of the different ML Methods

Source Credit: An Introduction to Machine Learning, Fig. 1(a) doi:10.1002/cpt.1796 42

 Amity Institute of Information Technology

ML Methods: Interpretability vs. Performance

Source Credit: An Introduction to Machine Learning, Fig. 1(b) doi:10.1002/cpt.1796 43

 Amity Institute of Information Technology

Evolution of Machine Learning

44
 Amity Institute of Information Technology

Evolution of Types of
Machine Learning

45
 Amity Institute of Information Technology

Linear Algebra

Source Credit: Machine Learning, [Link]

[Link] 46
 Amity Institute of Information Technology

What is Linear Algebra?

47
 Amity Institute of Information Technology

Why do we need to know it?

48
 Amity Institute of Information Technology

Scalar

49
 Amity Institute of Information Technology

Vectors

50
 Amity Institute of Information Technology

Geometry of Vectors

51
 Amity Institute of Information Technology

Geometry of Vectors

52
 Amity Institute of Information Technology

Dot Product and Angles

53
 Amity Institute of Information Technology

Norm of a Vector

54
 Amity Institute of Information Technology

Hyperplanes

55
 Amity Institute of Information Technology

Hyperplanes

56
 Amity Institute of Information Technology

Hyperplanes

57
 Amity Institute of Information Technology

Matrices

58
 Amity Institute of Information Technology

Matrices

59
 Amity Institute of Information Technology

Matrices

60
 Amity Institute of Information Technology

Matrices

61
 Amity Institute of Information Technology

Matrices

62
 Amity Institute of Information Technology

Matrix-Vector Products

63
 Amity Institute of Information Technology

Matrix-Matrix Products

64
 Amity Institute of Information Technology

Inverse of a Matrix

65
 Amity Institute of Information Technology

Pseudo-Inverse of a Matrix

66
 Amity Institute of Information Technology

Tensors

67
 Amity Institute of Information Technology

Eigen Decomposition

68
 Amity Institute of Information Technology

Eigen Decomposition

69
 Amity Institute of Information Technology

Eigen Decomposition

70
 Amity Institute of Information Technology

Differential Calculus

71
 Amity Institute of Information Technology

Differential Calculus

72
 Amity Institute of Information Technology

Optimization

73
 Amity Institute of Information Technology

Optimization

[Link] 74
 Amity Institute of Information Technology

Optimization
• Combinatorial optimization
– E.g.: Greedy search
• Convex optimization
– E.g.: Gradient descent
• Constrained optimization
– E.g.: Linear programming

75
 Amity Institute of Information Technology

Probability Theory

Source Credit: Machine Learning, [Link]

[Link] 76
 Amity Institute of Information Technology

Basics of Probability Theory

Slide Source Credit: Jeff Howbert — Machine Learning Math Essentials 77

 Amity Institute of Information Technology

Basics of Probability Theory

Slide Source Credit: Jeff Howbert — Machine Learning Math Essentials 78

 Amity Institute of Information Technology

Basics of Probability Theory: Random variables

Slide Source Credit: Jeff Howbert — Machine Learning Math Essentials 79

 Amity Institute of Information Technology

Basics of Probability Theory: Axioms of probability

Slide Source Credit: Jeff Howbert — Machine Learning Math Essentials 80

 Amity Institute of Information Technology

Basics of Probability Theory: Discrete Variables

Slide Source Credit: Jeff Howbert — Machine Learning Math Essentials 81

 Amity Institute of Information Technology

Basics of Probability Theory:

Multivariate Random Variables

Slide Source Credit: Jeff Howbert — Machine Learning Math Essentials 82

 Amity Institute of Information Technology

Basics of Probability Theory: Joint Probability Distribution

Slide Source Credit: Jeff Howbert — Machine Learning Math Essentials 83

 Amity Institute of Information Technology

Basics of Probability Theory: Marginal Probability Distribution

Slide Source Credit: Jeff Howbert — Machine Learning Math Essentials 84

 Amity Institute of Information Technology

Basics of Probability Theory:

Conditional Probability Distribution

Slide Source Credit: Jeff Howbert — Machine Learning Math Essentials 85

 Amity Institute of Information Technology

Bayes’ Theorem

Slide Source Credit: Jeff Howbert — Machine Learning Math Essentials 86

 Amity Institute of Information Technology

Independence

Slide Source Credit: Jeff Howbert — Machine Learning Math Essentials 87

 Amity Institute of Information Technology

Expected Value

Slide Source Credit: Jeff Howbert — Machine Learning Math Essentials 88

 Amity Institute of Information Technology

Variance

Slide Source Credit: Jeff Howbert — Machine Learning Math Essentials 89

 Amity Institute of Information Technology

Covariance

Slide Source Credit: Jeff Howbert — Machine Learning Math Essentials 90

 Amity Institute of Information Technology

Correlation

Slide Source Credit: Jeff Howbert — Machine Learning Math Essentials 91

 Amity Institute of Information Technology

Probability Distributions

Slide Source Credit: Jeff Howbert — Machine Learning Math Essentials 92

 Amity Institute of Information Technology

Probability Distributions

Slide Source Credit: Jeff Howbert — Machine Learning Math Essentials 93

 Amity Institute of Information Technology

Probability Distributions

Slide Source Credit: Jeff Howbert — Machine Learning Math Essentials 94

 Amity Institute of Information Technology

Cross-entropy

Slide Source Credit: Jeff Howbert — Machine Learning Math Essentials 95

 Amity Institute of Information Technology

Overview of Target Function Representations

• In supervised machine learning, the target function (also called

hypothesis function, mapping function, or ground truth function)
refers to the ideal function that maps inputs (features) to outputs
(labels). Since this function is generally unknown, learning
algorithms approximate it using data.
• Different ML approaches use different representations for this
function. Choosing the right representation is crucial because it
affects accuracy, interpretability, efficiency, and generalization.

96
 Amity Institute of Information Technology

Target Function Representations

97
 Amity Institute of Information Technology

Types of ML Techniques
• Based on task: • Based on approach:
– Classification – Symbolic (Logic-
– Regression based)
– Clustering – Sub-symbolic (Neural
– Anomaly Detection Networks, Deep
Learning)
– Recommendation
– Evolutionary (Genetic
Algorithms)

98
 Amity Institute of Information Technology

Hypothesis selection through cross validation

• What is Hypothesis Selection?
• Hypothesis space H: Set of all candidate models (e.g.,
linear models with different coefficients, decision trees of
different depths, neural networks with different
architectures).
• Goal: Select the hypothesis h∗∈H that generalizes best
to unseen data.
• Problem: Training error alone is not reliable, since a
complex model may fit training data perfectly but fail on
new data.

99
 Amity Institute of Information Technology

Cross-Validation for Hypothesis Selection

k-Fold Cross-Validation Leave-One-Out Cross-

Validation (LOOCV)

Stratified Cross-
Validation
Ensures that each fold preserves the
class distribution (important for
imbalanced datasets).

100
 Amity Institute of Information Technology

How Cross-Validation Helps in Hypothesis Selection?

• Avoids overfitting: Penalizes overly complex

hypotheses that don’t generalize well.
• Fair comparison: Each hypothesis is evaluated on the
same validation scheme.
• Model tuning: CV helps select not just the hypothesis
type but also hyperparameters (e.g., depth of decision
tree, regularization strength, number of hidden layers).

101
 Amity Institute of Information Technology

Example Workflow

102
 Amity Institute of Information Technology

Bias-Variance Trade-off in Hypothesis Selection

• Low bias, high variance (complex models): May fit training well
but CV error high.
• High bias, low variance (simple models): CV error high due to
underfitting.
• Best hypothesis: Minimizes expected generalization error, as
estimated by cross-validation.

Summary

103
 Amity Institute of Information Technology

Some Key Machine Learning Definitions

• Model: A machine learning model can be a mathematical representation of a real-
world process. To generate a machine learning model you will need to provide
training data to a machine learning algorithm to learn from.
• Algorithm: Machine Learning algorithm is the hypothesis set that is taken at the
beginning before the training starts with real-world data. When we say Linear
Regression algorithm, it means a set of functions that define similar
characteristics as defined by Linear Regression and from those set of functions
we will choose one function that fits the most by the training data.
• Training: While training for machine learning, you pass an algorithm with training
data. The learning algorithm finds patterns in the training data such that the input
parameters correspond to the target. The output of the training process is a
machine learning model which you can then use to make predictions. This
process is also called “learning”.
• Regression: Regression techniques are used when the output is real-valued
based on continuous variables. For example, any time series data. This technique
involves fitting a line.
• Classification: In classification, you will need to categorize data into predefined
classes. For example, an email can either be ‘spam’ or ‘not spam’.
104
 Amity Institute of Information Technology

Some Key Machine Learning Definitions

• Target: The target is whatever the output of the input variables. It could be the
individual classes that the input variables maybe mapped to in case of a classification
problem or the output value range in a regression problem. If the training set is
considered then the target is the training output values that will be considered.
• Feature: Features are individual independent variables that act as the input in your
system. Prediction models use features to make predictions. New features can also
be obtained from old features using a method known as ‘feature engineering’. More
simply, you can consider one column of your data set to be one feature. Sometimes
these are also called attributes. And the number of features are called dimensions.
• Label: Labels are the final output. You can also consider the output classes to be the
labels. When data scientists speak of labeled data, they mean groups of samples that
have been tagged to one or more labels.
• Overfitting: An important consideration in machine learning is how well the
approximation of the target function that has been trained using training data,
generalizes to new data. Generalization works best if the signal or the sample that is
used as the training data has a high signal to noise ratio. If that is not the case,
generalization would be poor and we will not get good predictions. A model is
overfitting if it fits the training data too well and there is a poor generalization of new
data.
105
 Amity Institute of Information Technology

Some Key Machine Learning Definitions

• Regularization: Regularization is the method to estimate a

preferred complexity of the machine learning model so that the
model generalizes and the over-fit/under-fit problem is avoided.
This is done by adding a penalty on the different parameters of the
model thereby reducing the freedom of the model.
• Parameter and Hyper-Parameter: Parameters are configuration
variables that can be thought to be internal to the model as they
can be estimated from the training data. Algorithms have
mechanisms to optimize parameters. On the other hand,
hyperparameters cannot be estimated from the training data.
Hyperparameters of a model are set and tuned depending on a
combination of some heuristics and the experience and domain
knowledge of the data scientist.
106
 Amity Institute of Information Technology

Suggested Video and Free Courses

• MIT Open Course Ware
– Probabilistic Systems Analysis and Applied Probability
[Link]
applied-probability-fall-2013/

107