Machine Learning

Techniques

KCS 055
 A physical system using
What is power to apply forces and
Machines? control movement to
perform an action.
 What is Learning?

To acquire knowledge or skills through

study, experience, or being taught.

 Types of Human Learning

• Learning under expert guidance.

• Learning guided by knowledge
gained from experts.
• Learning by self.
 What is Machine Learning?
• The construction and study of
systems that can learn from data.
• Machine learning is a branch of
an Artificial Intelligence.
 What is
Artificial
Intelligence?

• The intelligence of
machines or
software.
• Machine imitates like
human beings.
• Ability of machines
to think and learn.
AI Tools
Data science is the
study of data to
extract meaningful
insights for business
 Definition of Machine Learning
• In 1959, Arthur Samuel defined machine learning as a “Field of Study
that gives computers the ability to learn without being explicitly
programmed”.
• The goal of machine learning is to build computer systems that can
adapt and learn from their experience.” -Tom Dieterich
• Tom M. Mitchell provided a widely quoted, more formal definition:
“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, improved with
experience E.”
 So, what is Machine Learning?

• Automating automation.
• Getting computers to program
themselves
• Writing software is the
bottleneck.
Let the data to do the work instead!
Well Posed Learning Problem
• Task (T)
• Experience (E)
• Performance (P)
“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, improved with
experience E.”
Checkers Game Learning Problem
• Task (T) – Playing
Checker
• Experience (E) – Playing
practice games
• Performance (P) –
Percentage of game won
against opponent.
Handwriting Recognition Learning
Problem
• Task (T) – Recognizing and
classifying Handwritten
words from the given images
• Experience (E) – A database
of images of handwritten
words
• Performance (P) –
Percentage of words
correctly classified
 Robot Car driving Learning
Problem
• Task (T) – Driving on 4-lane
Highway using vision
sesnors.
• Experience (E) – Sequence
of images and steering
commands recorded while
observing human drivers.
• Performance (P) – Average
distance travelled before
an error.
 Traditional Programming

Data
Computer Output
Program

Machine Learning

Data
Computer Program
Output
 Why ML?
• Develop systems that can automatically adapt
and customize themselves to individual users.
– Personalized news and mail filter.
• Discover new knowledge from larger databases.
– Market basket analysis
• Ability to mimic human and replace certain
monotonous tasks- which require some
intelligence.
– Like recognizing handwritten characters.
– Face recognition
ML Applications
History of
ML
 How do Machine Learn?

Input Data Abstraction Generalization

 How do Machine Learn?

Input Data Abstraction Generalization

• Input Data – Past data or information that is used to train the

machines known as training data.
• Abstraction- Input data represented in a broader way through
the underlying algorithm.
• Generalization – Abstracted representation is generalized to
form a framework for making decisions.
 Problems not to
be solved using
Machine
• Inadequate Training Data. Learning
• Poor quality of data.
• Frequent human intervention is needed.
– Air traffic control
• For very simple task which can be
implemented using traditional programming.
– For example: formula-driven or rule-
based applications such as price
calculator engine.
Biggest Issue in Machine
Learning
Breach of one’s Privacy
Other Issues in Machine
Learning
Lack of data for training the model
Other Issues in Machine
Learning
Biased Data
 Semi-
Supervised
Reinforcement Learning
Supervised Learning
Clustering vs Association
4 types of Machine Learning
Algorithm
Supervised Learning

Unsupervised Learning

Semi-Supervised Learning

Reinforcement Learning
Criteria Supervised ML Unsupervised ML Reinforcement ML

Definition Machine learns by using Machine is trained using Agent interacts with the
labelled data unlabelled data without environment by
any guidance performing action. Learns
by errors and rewards.
Types of Data Labelled Data Unlabelled Data No-Predefined Data
Types of Problems Regression and Association and Rewards and error based
Classification Clustering
Supervision External Supervision No Supervision No Supervision
Algorithms Linear regression, Logistic K-Means Clustering, PCA, Monte-Carlo, Q-learning,
Regression, Naïve Bayes, Neural Networks SARSA
Decision Tree, KNN
Aim Calculate Outcomes Discover Underlying Learn a series of action
patterns
Approach Maps labelled inputs to Understand patterns & Follow the trial and error
the known outputs discover output method
Application Risk Evaluation, Forecast Recommendation System, Self Driving Cars, Gaming,
Sales Anomaly Detection Health Care
 Regression
Supervised
ML
Classification
 Machine
Learning
Algorithms
Did you know?
Machine Learning saves life – It can spot 52% of cancer cells a year before the
patients are diagnosed.
Uses – Facial Recognition
Uses – Voice Recognition
 Artificial
Bayesian Neural
Networks Network
s

Decision Genetic
Tree Algorithm

Various Machine
Learning Naive
Algorithms KNN
Bayes

Cluster Support
Vector
ing Machine
Bayesian Belief Networks
Naive Bayes Classifier
Bayes’ Theorem
Artificial Neural Networks
 Feed - Forward

Backward Propogation
K-Nearest Neighbour
Decision Tree Algorithm
Random
Forest
Classifier
Genetic Algorithm
Clustering
Independent Variable and
Dependent Variable
Dataset

Target Variable –
Healthy

Feature Variable –
weight. Height,
drinks, alcohol
 area 7420 Feature Variable

Target Variable price 13,300,000

bedrooms 4
mainroad Yes
guestroom No
basement No
hotwaterheating No
airconditioning Yes
parking 2
furnishingstatus Furnished
 Reference Books

Tom M. Mitchell, Ethem Alpaydin, ―Introduction to Stephen Marsland, Bishop, C., Pattern
―Machine Learning, Machine Learning (Adaptive ―Machine Learning: An Recognition and Machine
McGraw-Hill Education Computation and Machine Algorithmic Perspective, Learning. Berlin: Springer-
(India) Private Limited, Learning), The MIT Press 2004. CRC Press, 2009. Verlag.
2013.
 Text Books

Saikat Dutt, Subramanian Andreas C. Müller and Sarah John Paul Mueller and Luca
Chandramaouli, Amit Guido - Introduction to Massaron - Machine Learning for
Kumar Das – Machine Machine Learning with Dummies
Learning, Pearson Python