E

LEARNI
NG
GROUP MEMBERS :
M. Arman Haider Zaidi ( Team Lead ) (
02-131242-116 )
M. Affan bin Aamir ( 02-
131242-064 )
Syed Shayan Agha ( 02-
131242-093 )
 Table of
What is
Contents Working of
01 02
Machine Machine
. Learning . Learning
Machine Types of
03 Learning
04 Machine
. Algorithms . Learning
Applications Algorithms
Challenges of
05 06 Machine
of Machine
. Learning . Learning
Future Trends 08
07 Conclusion
of Machine
. Learning
.
WHAT IS MACHINE LEARNING?

Machine Learning is a way

for computers to learn and
make decisions without
being explicitly
programmed. Instead of
following fixed instructions,
the computer learns from
data and improves its
performance over time
WORKING OF MACHINE LEARNING
ALGORITH
M TRAINING
A Choose the method
for learning methods
T Use historical data to
teach model

MODEL TESTING
M The trained
representation of
T Validate the model’s
predictions on new
learned patterns data
WHAT ARE MACHINE
LEARNING
ALGORITHMS?
Machine learning algorithms are
computational models that allow
computers to understand patterns and
forecast or make judgments based on
data without explicit programming.
These algorithms form the foundation
of modern artificial intelligence and
are used in various applications,
including image and speech
recognition, natural language
processing, recommendation systems,
TYPES OF MACHINE LEARNING
ALGORITHMSSupervised
Learning

Unsuperised
Learning
Types of ML
Algorithms
Reinforcement
Learning
 SUPERVISED LEARNING

REGRESSIO Supervised learning

involves training a
CLASSIFICAT
N model on labeled IONS
data, where the
Regression involves desired output is Classification, on the
predicting a continuous known. The model other hand, involves
numerical value based learns to map inputs predicting a discrete
on input features, SUPERVISE
to outputs based on
the provided
label or category,
where the output is a where the output is a
real number. D examples. specific class or group.
LEARNING
 METHODS FOR CLASSIFICATION AND
REGRRESSION
NUERAL
NETWORKS

1
RANDOM DECISION
FOREST 4 2 TREES

3
GRADIENT
BOOSTING
 UNSUPERVISED LEARNING
Clustering is the process of CLUSTERI
grouping similar data points into
clusters based on their features, NG
without predefined labels. The goal
is to discover inherent structures in
the data, allowing for insights and
patterns to emerge naturally.
DIMENSION Dimensionality reduction refers to
the process of reducing the number
ALITY of features in a dataset while
preserving its essential structure
REDUCTION and information. This technique
helps improve computational
efficiency and mitigate issues like
overfitting.
Association is the process of
discovering relationships or patterns
ASSOCIA
between variables in large datasets.
It is commonly used in market
TION
basket analysis to identify items that
frequently co-occur, helping to
uncover hidden associations
between different
 REINFORCEMENT
0 LEARNING
0 0
1
MODEL- 2 3
FREE MODEL- VALUE-
METHOD BASED BASED
S MOTHODS METHODS
Model-free methods in machine Model-based methods in Value-based methods in
learning do not rely on machine learning build a reinforcement learning
predefined models or model of the data or estimate the value of states
assumptions about data. environment to make or actions to guide decision-
Instead, they learn directly predictions and decisions, making. The agent selects
from data through trial-and- often using probabilistic or actions to maximize
error or pattern observation, statistical assumptions. In expected future rewards,
often used in reinforcement reinforcement learning, with Q-learning being a
learning to optimize actions they involve modeling the common example for
without explicit environment to plan and learning optimal policies.
environmental models. optimize actions.
 APPLICATIONS OF MACHINE
LEARNING

Fraud STOCK SELF

Detection MARKET RECOMMENDE DRIVING
TRADING R SYSTEMS CARS
ML can help financial
It analyzes the It analyzes users Machine learning is
institutions detect
trend of stock preference and fundamental in
potentially fraudulent
market and predict search history enabling self-
transactions and save
future trend and based on that it driving cars to
consumers from false
market value as recommend recognize their
charges by flagging
well. content and environment,
those that seem
services to them. make decisions,
suspicious or out
and navigate
of the ordinary.
safely on the road.
 APPLICATIONS OF MACHINE
LEARNING

Detect ENERGY CHATBOT

SPEECH medical CONSUMPTION S
RECOGNITION conditions: OPTIMIZATION
Speech Recognition Energy consumption Using natural language
Researchers use
based smart systems optimization uses machine processing, these
machine learning to
like Alexa and Siri learning to analyze data chatbots are capable of
identify cancerous
have certainly come from sensors and weather, responding to
growth in medical
across and used to predicting energy demand. consumers' unique
scans, others are
communicate with It adjusts systems like queries and directing
using it to create
them. In the heating and lighting in them to the appropriate
software that can
backend, these real-time to reduce waste, resources so that
help health care
systems are based lower costs, and maintain customer support
professionals make
basically on Speech comfort, benefiting smart specialists can assist
more
Recognition systems. homes, those with the trickiest
accurate diagnoses.
 CHALLENGES IN MACHINE
DATA QUALITY LEARNING
AND 01
AVAILABILITY
Machine learning models need large
amounts of high-quality data to perform
BAIS AND
well. Data can be poor quality if it is FAIRNESS IN
incomplete, outdated, or biased. This can 02
lead to inaccurate predictions and poor ALGORITHM
system performance. Data bias can lead to unfair outcomes and
perpetuate existing societal inequalities. To
ensure machine learning models are fair
and equitable, it's important to identify and
mitigate bias in the data
 SCALABILI
TY 03
Machine learning models need large
amounts of high-quality data to perform
well. Data can be poor quality if it is
SLOW
incomplete, outdated, or biased. This can IMPLEMENTATI
lead to inaccurate predictions and poor
system performance.
04 ON
It can take a long time for
machine learning
algorithms to start
producing efficient results.
The user have to give
multiple prompts to get the
correct information out of
it.
 ETHICAL CONSIDERATIONS

PRIVAC  Personal data might be

Y misused.

CONCE  Balancing innovation and

privacy is tough.
RNS

SECURI  Models can be hacked or

TY manipulated.
RISKS  Protecting systems is critical.
FUTURE TRENDS IN MACHINE LEARNING
ARTIFICIAL
D DEEP A GENERAL
LEARNING INTELLIGENCE
L FEDERAT GI
ED
F R RESPONSIVE
LEARNIN
AI
L G AI
 DEEP LEARNING FEDRAL LEARNING
We use deep learning in speech, video and Fedral training is a machine learning
audio recognition. It uses artificial neural paradigm that enables training a machine
networks to learn features from large amount learning algorithm, for instance deep neural
of data. networks, on multiple local datasets
some pre-trained deep learning models contained in local nodes without explicitly
includes: exchanging data samples.
• CHAT-GPT( Generative Pre-trained some features includes:
transformer) • predictive health analytics
• YOLO(You Only look Once) • fraud detection
• DALL-E • collaborative mapping
• DeepSpeach
RESPONSIVE AI ARTIFICIAL GENERAL INTELLIGENCE
Implementation of Responsive AI means
creating AI systems that can quickly Artificial General Intelligence (AGI) refers
understand and react to changes in the to a type of AI that possesses the ability
environment, user needs, or real-time data. to understand, learn, and perform any
These systems adjust their actions to give intellectual task that a human can do.
better and more personalized responses. example:
example: • Robots
• Google Assistant
• Tesla’s Autopilot
 CONCLUSION

Machine learning has impacted our

lives very effectively. From image
recognition to natural language
processing, machine learning has
demonstrated its potential to
transform the world. We can
expect to see even more
innovative application of machine
learning in our daily lives.