Introduction to

Machine Learning
Overview
• Overview of Human Learning and Machine
Learning

• Types of Machine Learning

• Applications of Machine Learning

• Tools and Technology for Machine Learning

 Reference Books
1.Machine Learning, Saikat Dull, S. Chjandramouli,
Das, Pearson
2.Machine Learning with Python for Everyone, Mark
Fenner, Pearson
3.Machine Learning, Anuradha Srinivasaraghavan, Vincy
Joseph, Wiley
4.Machine Learning with Python, U Dinesh Kumar
Manaranjan Pradhan, Wiley 5. Python Machine Learning,
Sebastian Raschka, Vahid Mirjalili, Packt Publishing
Online Popular Courses:

• Machine Learning by Stanford University (Coursera)

• Machine Learning Crash Course by Google AI

• Machine Learning by Georgia Tech (EdX)

• Deep Learning, YouTube: Alexander Amini (MIT)

History of ML
(Credits: Alan Turing)
 Overview of Human Learning
• Learning: Process of gaining information through observations.
• Types of Learning:
 Learning directly under expert guidance
- Parents teach to infants
- Learning from teachers at school
- Professional mentors

 Learning guided by knowledge gained from experts

- Learning happens with the knowledge which has been imparted
by teacher or mentor at some point of time in some other
form/context.
- Baby can group together all objects of same colour
 Self-learning
- Learning when we are left to learn on our own.
Before defining Machine Learning..

• Do machines really learn?

• If so, how do they learn?

• What are the important features that are required to

well-define a learning problem?
• Defining Machine Learning

“A computer program is said to learn from

experience E with respect to some class of tasks
T and performance measure P, if its performance
at tasks in T, as measured by P, improves with
experience E.”
- Prof. Tom M. Mitchell,
Carnegie Mellon University
How do machines learn?
1. Data Input: Past data or information is utilized
as a basis for future decision-making.

2. Abstraction: The input data is represented in a

broader way through the underlying
algorithm.

3. Generalization: The abstracted representation

is generalized to form a framework for making
decisions.
Abstraction
• Helps in deriving a conceptual map based on the input
data known as machine learning paradigm.
(Summarized knowledge representation of the raw data)

• Data can be arranged in form of.. (The choice is ours!)

o Computational blocks like if/else rules
o Mathematical equations
o Specific data structures like trees or graphs
o Logical groupings of similar observations
• Choice can be based on,
• The type of problem to be solved
• Nature of the input data
• Domain of the problem

• The next task is to fit the model based on the

input data (known as training).
Generalization:
• Key part of ML is to tune up the abstracted
knowledge to a form which can be used to take
future decisions. It can be achieved by
Generalization.

• Quite difficult to achieve because the model is

trained based on a finite set of data, which may
possess a limited set of characteristics.
• When we apply the model to take decision on test
data, we may encounter two problems:

1. The trained model is aligned with the training

data too much.

2. The test data possess certain characteristics

apparently unknown to the training data.
 Types of Machine Learning

1. Supervised Learning (Predictive Learning)

2. Unsupervised Learning (Descriptive Learning)

3. Reinforcement Learning
1. Supervised Learning:
• A machine predicts the class of unknown
objects based on prior class-related
information of similar objects.

• Data: (x,y) x is input data, y is label

• Goal: Learn function to map, x to y

2. Unsupervised Learning:

• A machine finds patterns in unknown objects by

grouping similar objects together.

• Data: x

• Goal: Learn the hidden pattern of data

Example of Unsupervised learning
 Types of ML

Machine Learning

Supervised Learning Unsupervised Learning Reinforcement

learning

Classification Regression Clustering Association

Analysis
 Supervised Learning
• Major Motivation: to learn from past information.

• Labelled Training Data: contains past information comes as an input.

• Based on the training data, the machine builds a predictive model that can
be used on test data to assign a label for each record in the test data.
• Examples:
1. Predicting the results of a game
2. Predicting whether a tumour is malignant or benign
3. Predicting the price of domains like real estate, stocks, etc.
4. Classifying texts such as classifying a set of emails as spam or
non-spam

Are Problem (2) and (3) same in nature?

Answer: No
• Classification Problem: When we are trying to predict a
categorical or nominal variable, the problem is known as a
classification problem.

• Regression Problem: Whereas when we are trying to predict a

real-valued variable, the problem falls under the category of
regression.
Classification:

Algorithms: Naïve Bayes, Decision tree, and k-Nearest Neighbour

Examples of Classification
• Image classification
• Prediction of disease
• Win–loss prediction of games
• Prediction of natural calamity like earthquake, flood,
etc.
• Recognition of handwriting
Regression Problem
• Objective is to predict numerical features like real
estate or stock price, temperature, marks in an
examination, sales revenue, etc.

• The underlying predictor variable and the target

variable are continuous in nature.
• Simple linear regression: Only one predictor variable

• Multiple linear regression: Multiple predictor variables

Applications of Regression

• Sales prediction for managers

• Price prediction in real estate
• Weather forecast
• Skill demand forecast in job market
 Unsupervised Learning
• No labelled training data to learn from and no prediction to
be made.

• Objective: To take a dataset as input and try to find natural

groupings or patterns within the data elements or records.

• The process of unsupervised learning is referred as pattern

discovery or knowledge discovery.
Unsupervised Learning
• Clustering is the main type of unsupervised learning
(intends to group or organize similar objects together).
 Reinforcement learning
• Feedback based ML technique.
• Consider an example of baby learning to walk first time.

• A child is an agent and the place with hurdles on which the child
is trying to walk resembles the environment.
• It tries to improve its performance of doing the task.
• When a sub-task is accomplished successfully, a reward is given.
a sub-task is not executed correctly= no reward is given.
Reinforcement Learning
 Example:
• Self-driving car

• The tasks that have to be taken care of are start/stop,

accelerate/decelerate, turn to left / right, etc.

• Critical information which it needs to consider are speed

and speed limit in different road segments, traffic
conditions, road conditions, weather conditions, etc.
Comparison of Supervised, Unsupervised, and
Reinforcement Learning
Supervised Unsupervised Reinforcement
Learning Learning
- Used when you know - Used when there is no - Used when there is no
how to classify a given idea about class or idea about the class or
data. particular data. particular data. The
model has to do
classification. It will be
rewarded if it is correct.
- Labelled training data is - Any unknown and - The model learns and
needed. Model is built unlabeled dataset is given updates itself through
based on training data. to the model as input and reward/punishment.
records.
 Supervised Unsupervised Reinforcement
Learning Learning
- The model performance - Difficult to measure - Model is evaluated by
can be evaluated based of whether the model did means of the reward
how many something useful or function after it had
misclassification have interesting. Homogeneity sometime to learn.
been done based on a of records grouped
comparison b/w together is the only
predicted and actual measure.
values.
- Types: Classification, - Types: Clustering, - No types
Regression Association
 Supervised Unsupervised Reinforcement
Learning Learning
- Simple to understand - More difficult to - Most complex to
and implement. understand and understand and
implement. implement.
- Standard Algorithms: - Standard Algorithms: Standard Algorithms: Q-
Naïve Bayes, KNN, K-means, PCA, Self- learning, Sarsa
Decision tree, Linear organizing map, Apriori
regression, Logistic algorithms, DBSCAN
regression, SVM
Applications: Applications: Market Applications: Self-
Handwriting recognition, basket analysis, Customer driving cars, Intelligent
Stock market prediction segmentation robots, AlphaGo Zero
 Applications of ML
• Image Recognition: Used to identify objects,
persons, places, digital images, etc.

• Speech Recognition: Used by various applications

of speech recognition. E.g.: Google
assistant, Siri, Cortana, and Alexa

• Traffic Prediction: Predicts the traffic conditions.

 Applications of ML
• Product recommendations: For product
recommendation to the user.

• Self driving cars: To train the car models to detect

people and objects while driving, e.g. Tesla.

• Email Spam and Malware Filtering: Filters

automatically as important, normal, and spam.
E.g.: Content Filter, Header filter, General
blacklists filter, Rules-based filters, Permission
filters.
 Applications of ML
• Virtual Personal Assistant: Help us in finding the
information using our voice instruction. E.g.
Google assistant, Alexa, Cortana, Siri.

• Online Fraud Detection: Making online

transaction safe and secure by detecting fraud
transaction.
- Feed Forward Neural network

• Stock Market Trading: Machine learning's long

short term memory neural network is used for the
prediction of stock market trends.
 Applications of ML
• Medical Diagnosis: Medical technology is
growing very fast and able to build 3D models that
can predict the exact position of lesions in the
brain.

• Automatic Language Translation: Helps us by

converting the text into our known languages. E.g.:
Google's GNMT (Google Neural Machine
Translation)