W H AT I S

MACH I NE
LEARN I NG
SAMEER MALIK
 1 2 3 4 5 6 7

Data Data Choose a Training the Evaluate the Parameter Make

Collection: Preparation: Model/Algorith Model: Using Model: Tuning: Fine- Predictions:
Gathering Cleaning, m: Selecting the prepared Assessing the tuning the Using the
relevant data transforming, the data to train model's model trained and
from various and organizing appropriate the selected performance parameters to optimized
sources to be the data to machine model to learn and accuracy optimize its model to make
used in the make it learning model patterns and using performance predictions on
machine suitable for or algorithm relationships. validation and new, unseen
learning model based on the techniques. generalization. data.
process. training. nature of the
problem.

HOW DOES MACHINE

LEARNING WORKS?
HOW DOES MACHINE
LEARNING WORKS?
MACHINE LEARNING TYPES:

Supervised

Unsupervised

Reinforcement
 SUPERVISED
MACHINE
LEARNING
• Supervised machine
learning is a type of
machine learning where
the algorithm learns from
labeled training data, with
each input data point
paired with the
corresponding correct
output. The goal is for the
algorithm to learn the
mapping between input
and output variables,
making predictions on
new data based on the
learned patterns.
S U P E RV I S E D : E X A M P L E
 Support Vector
Linear Logistic
Machines
Regression Regression
(SVM)

Neural
Decision Trees Random Forest
Networks

k-Nearest
Naive Bayes Neighbours (k-
NN)

SUPERVISED ML: ALGORITHMS

 Recommendation
Image Sentiment Fraud Detection:
Speech Systems:
Recognition: Analysis: Analysing Identifying
Recognition: Suggesting
Identifying objects, and categorizing fraudulent activities
Converting spoken products, movies, or
people, or patterns opinions expressed in financial
language into text. content based on
in images. in text data. transactions.
user preferences.

Predictive Autonomous
Medical Diagnosis: Credit Scoring: Weather
Maintenance: Driving: Enabling
Assisting in the Assessing the Forecasting:
Predicting when vehicles to navigate
diagnosis of creditworthiness of Predicting weather
equipment or and make decisions
diseases based on individuals based on conditions based on
machinery is likely based on sensor
patient data. various factors. historical data.
to fail. data.

SUPERVISED ML: APPLICATIONS

 U N - S U P E RV I S E D
M AC H I N E L E A R N I N G
• Unsupervised learning is a
type of machine learning
where the model learns
patterns and relationships
in the data without being
explicitly trained on
labeled data. In
unsupervised learning, the
algorithm explores the
data to find hidden
structures or patterns
without any guidance or
supervision. Clustering and
dimensionality reduction
are common tasks in
unsupervised learning.
U N - S U P E RV I S E D : E X A M P L E
UN-SUPERVISED ML:
ALGORITHMS

K-MEANS CLUSTERING HIERARCHICAL ASSOCIATION

CLUSTERING ALGORITHM
Customer segmentation: Identifying different
segments of customers based on their purchasing
behaviour.

Anomaly detection: Detecting unusual patterns or

outliers in data that may indicate fraud, errors, or
other anomalies.
UN-
Image and video recognition: Grouping similar images
or videos together based on visual similarities.
SUPERVISE
Market basket analysis: Identifying patterns in
D ML:
customer purchase behaviour to recommend related
products. APPLICATIO
Dimensionality reduction: Reducing the number of
features in a dataset while preserving important
NS
information.

Natural language processing: Clustering documents or

words based on semantic similarities for tasks like
document classification or topic modelling.
REINFORCEMENT
MACHINE LEARNING

Reinforcement learning is a type of machine

learning where an agent learns to make
decisions by interacting with an environment.
The agent receives feedback in the form of
rewards or penalties based on its actions, and
its goal is to learn a policy that maximizes its
cumulative reward over time.
Reinforcement learning is often used in
scenarios where explicit training data is not
available, and the agent must learn through
trial and error.
It is commonly applied in robotics, game
playing, autonomous driving, and other
sequential decision-making tasks.
REINFORCEMENT
MACHINE LEARNING: WORKING
PRINCIPLE
REINFORCEMN
ET TRANING
WITH DOG
 Robotics: Teaching robots Autonomous driving:
Game playing: Training
to perform tasks such as Developing self-driving
agents to play games like
grasping objects, cars that learn to navigate
chess, Go, or video games
navigating environments, traffic, follow rules, and
by learning optimal
or manipulating objects make decisions on the
strategies through
using reinforcement road through
reinforcement.
learning. reinforcement learning.

Recommendation systems: Finance: Optimizing Healthcare: Personalizing

Personalizing trading strategies, treatment plans,
recommendations for portfolio management, or predicting patient
users based on their risk assessment in outcomes, or optimizing
interactions with a financial markets using healthcare workflows
platform or content using reinforcement learning through reinforcement
reinforcement learning. algorithms. learning techniques.

REINFORCEMENT LEARNING:
EXAMPLES
Thank
s
 AI VSMACHINE
LEARNINGVSDEEP LEARNING

SAMEER MALIK
HISTORY OF AI
WHAT IS AI?

Artificial intelligence (AI) is a branch of computer

science that focuses on creating systems or
machines that can perform tasks that typically
require human intelligence. These tasks may
include learning, reasoning, problem-solving,
perception, language understanding, and
decision-making.
AI systems are designed to simulate human
cognitive functions and adapt to new situations by
learning from data. AI encompasses various
subfields such as machine learning, natural
language processing, computer vision, robotics,
and expert systems, among others.
The goal of AI is to create intelligent machines
that can mimic human behaviour and perform
tasks autonomously.
EXAMPLES OF AI
WHY MACHINE LEARNING
WAS INTRODUCED?
Machine learning was 1. Handling complex and
introduced to address the unstructured data: Traditional
2. Automation and efficiency: 3. Personalization and
limitations of traditional rule- programming methods struggle
Machine learning enables recommendation: Machine
based programming and to to handle large volumes of
automation of tasks that are learning algorithms can analyze
enable computers to learn from complex and unstructured data,
difficult or time-consuming for user behavior and preferences
data and improve their such as images, text, and
humans to perform manually. to provide personalized
performance over time without sensor data. Machine learning
By learning from data, recommendations in areas such
being explicitly programmed for algorithms can learn patterns
machines can make decisions as e-commerce, content
every possible scenario. Some and relationships from this data
faster and more efficiently. delivery, and online services.
key reasons for the introduction to make predictions or
of machine learning include: decisions.

5. Prediction and forecasting: Overall, the introduction of

4. Adaptability and scalability: Machine learning algorithms machine learning has
Machine learning models can can analyze historical data to revolutionized the field of
adapt to new data and changing make predictions about future artificial intelligence and
environments, making them trends, outcomes, or events in enabled computers to learn
scalable and flexible for a wide various domains such as from data and improve their
range of applications. finance, healthcare, and performance in a wide range of
weather forecasting. applications.
EXAMPLE OF MACHINE
LEARNING
WHAT IS DEEP LEARNING?

Deep learning is a subset of machine learning that uses artificial neural networks with multiple
layers (deep neural networks) to learn complex patterns and representations from data. These
deep neural networks are inspired by the structure and function of the human brain, with
interconnected layers of nodes that process and transform input data to produce output
predictions.

Deep learning algorithms excel at tasks such as image and speech recognition, natural language
processing, and other tasks that require understanding and processing large amounts of data.

Deep learning has shown remarkable success in various fields, including computer vision, speech
recognition, recommendation systems, and autonomous driving, among others.
 Natural language
Image classification: processing: Using
Using convolutional recurrent neural
neural networks to networks or
classify images into transformers to
different categories, generate text or
such as identifying understand and
animals in pictures. respond to human
language.

Autonomous driving:
Using deep learning
Healthcare: Using
deep learning for
medical image
EXAMPLE
S OF
analysis, such as
algorithms to process
detecting tumours in
sensor data and make
MRI scans or
decisions for self-
predicting patient
driving cars.

DEEP
outcomes based on
clinical data.

Recommendation Speech recognition:

LEARNIN
systems: Using deep Using deep learning
learning to analyse models to transcribe
user behaviour and spoken language into
preferences to text, enabling voice

G
recommend products, assistants and
movies, or music. dictation software.

Game playing: Using

deep reinforcement
learning to train
agents to play
complex games like
chess or Go at a
superhuman level.
DEEP LEARNING IS MACHINE
LEARNING
ML VS DL:
FACTORS
M AC H I N E L E A R N I N G
ALGORITHMS

• 1- Classification
• 2- Anomaly
Detection
• 3- Clustering
• 4- Linear
Regression
ALGORITHMS:
TYPE, EXAMPLE,
USECASE