MACHINE LEARNING

FUNDAMENTAL
SESSION 02

SUBJECT MATTER EXPERT

Henry Lucky, S.Kom., M.Kom.
LEARNING OUTCOMES
SESSION 2

At the end of this session, students will be able to:

• LO 1: Describe what is AI, its applications, use cases, and how it is
transforming organizations
• LO 2: Explain terms like machine learning, deep learning, neural
networks, natural language processing, computer vision, speech
recognition, etc.
OUTLINE
SESSION 2

• Introduction to Machine Learning

• Supervised Learning
• Unsupervised Learning
• Reinforcement Learning
• Deep Learning
INTRODUCTION TO MACHINE LEARNING
WHAT IS MACHINE LEARNING?

• Machine learning is the cornerstone of AI, considered fundamental

by many experts.
• AI applications are expected to possess machine learning capabilities
to be deemed truly intelligent.
• Simply being able to perform tasks like chess-playing doesn't signify
intelligence; machine learning is essential for advanced abilities.
• In essence, a chess-playing system lacking machine learning is akin
to a basic computer program, lacking true intelligence.
• So, what is machine learning?
INTRODUCTION TO MACHINE LEARNING
WHAT IS MACHINE LEARNING?

What is Machine Learning?

Source: https://youtu.be/ukzFI9rgwfU
INTRODUCTION TO MACHINE LEARNING
THE VON NEUMANN MACHINE

• A computer architecture proposed by polymath in 1945 based on Sir

Alan Turing’s work on Turing Machine.
• The basic components of vNM which consists of:
 Central processing unit (CPU) – system brain consists of system
control unit, basic arithmetic, and logical unit
 Memory unit - data and information storage
 Input and output units – data input (e.g. keyboard) and output
(e.g. screen)
INTRODUCTION TO MACHINE LEARNING
THE VON NEUMANN MACHINE
INTRODUCTION TO MACHINE LEARNING
THE VON NEUMANN MACHINE

• All computer systems that we use today; their basic architectures are
all inherited from the blueprint of von Neumann Machine.
• Are they intelligent machines? Why or why not?
• Let’s look at two cases of computer systems in our daily activities.
INTRODUCTION TO MACHINE LEARNING
CASE 1: NUMERICAL WEATHER PREDICTION

• Weather forecasting relies on Numerical

Weather Prediction (NWP) systems, which
utilize complex mathematical models and
global weather observations.
• NWP accurately predicts weather components
like air pressure, temperature, humidity, and
wind speed for up to 7 days into the future.

• However, NWP lacks intelligence and cannot learn or improve

autonomously.
• The system doesn’t have the ability to adapt or learn from new
information.
INTRODUCTION TO MACHINE LEARNING
CASE 2: ALPHAGO

• AlphaGo, developed by DeepMind Technologies and acquired by

Google in 2015, is an AI-based computer program designed to play the
board game Go.
• AlphaGo's ability to learn and improve its gameplay distinguishes it
from standard von Neumann Machines.
• Similar to playing against an experienced chess master, AlphaGo
continuously adapts its strategies based on past experiences and
opponent moves.
• This is what we call machine learning – intelligence.
INTRODUCTION TO MACHINE LEARNING
APPLICATIONS

• Healthcare:
• Diagnostics: Predictive models for disease detection (e.g., cancer,
diabetes).
• Personalized Medicine: Treatment plans based on patient data.
• Medical Imaging: Automated analysis of X-rays, MRIs.
• Finance:
• Fraud Detection: Identifying fraudulent transactions.
• Algorithmic Trading: Automated, high-frequency trading
strategies.
• Credit Scoring: Assessing creditworthiness using ML models.
INTRODUCTION TO MACHINE LEARNING
APPLICATIONS

• Retail:
• Recommendation Systems: Personalized product recommendations
(e.g., Amazon, Netflix).
• Inventory Management: Predictive analytics for stock optimization.
• Customer Segmentation: Targeted marketing based on customer
behaviour.
• Other Advanced Applications:
• Autonomous vehicle, Natural Language Processing, Computer
Vision
• Smart Cities, Agriculture, Education
• etc.
INTRODUCTION TO MACHINE LEARNING
PILLARS OF MACHINE LEARNING

REINFORCEMENT

Source: https://towardsdatascience.com/coding-deep-learning-
for-beginners-types-of-machine-learning-b9e651e1ed9d
SUPERVISED LEARNING
DEFINITION

• Supervised Learning involves

learning from labeled data,
where the algorithm is trained
on input-output pairs.
• Common supervised learning
tasks include classification
(predicting categories) and
regression (predicting
continuous values).

Source: https://alanjeffares.wordpress.com/2018/07/24/test/
UNSUPERVISED LEARNING
DEFINITION

• Unlike supervised Learning, it

doesn't rely on labeled data for
training.
• Instead, it explores patterns and
structures within unlabeled Source: https://alanjeffares.wordpress.com/2018/07/24/test/
data.
• It aim to uncover hidden patterns, groupings, or relationships in data.
• Common unsupervised learning tasks include clustering,
dimensionality reduction, and anomaly detection.
REINFORCEMENT LEARNING
DEFINITION

• In reinforcement learning, an agent learns to make decisions by

interacting with an environment.
• Unlike other types of learning, reinforcement learning doesn't need
labeled or unlabeled data - it learns from trial and error.
• The goal for the agent is to learn the best actions to take in different
situations to maximize rewards.
• It's like teaching a pet tricks: they learn what actions lead to treats and
repeat those actions.
• Even ChatGPT is built using some kind of reinforcement learning!
REINFORCEMENT LEARNING
EXAMPLES: A.I. LEARNS TO PLAY FLAPPY BIRD
REINFORCEMENT LEARNING
EXAMPLES: MULTI-AGENT HIDE AND SEEK
DEEP LEARNING
AI VS ML VS DL
DEEP LEARNING
NEURON

• A human brain is a complex

organization of a cell called as
neuron.

• A neuron can process information by

running this process:
Receive signals Tweak the Send new signals
from other signals in its cell to other neurons
neurons via body, via synaptic
dendrites, terminals
DEEP LEARNING
ARTIFICIAL NEURAL NETWORK

• ANNs are a computational model inspired by the structure and function

of biological neural networks in the human brain.
• Here is the comparation between biological neuron and artificial neuron:
DEEP LEARNING
ARTIFICIAL NEURAL NETWORK

• In ANN, the neurons are

interconnected and
organized in layers.
• Information flows through
the network from input
layers to output layers.
• Typical ANN has 3 layers:
Input, Hidden and Output
layers.
DEEP LEARNING
DEFINITION

• Deep learning revolves

around ANN with multiple
hidden layers (hence "deep"),
• It can learn complex patterns
in data.
• It can automatically learn
features from raw data,
eliminating the need for
manual feature extraction.
CASE STUDY
3 MACHINE LEARNING METHODS

• The 3 learning methods in machine learning are always integrated and

mixed with our daily learning activities, including:
o Attending lectures;
o Reading books and papers;
o Completing assignments and exercises;
o Participating in group discussions;
o Completing term tests and quizzes;
o Study and revision;

• Identify the learning methods involved in each activity above and give the
reasoning!
REFERENCES
SESSION 2

• Raymond S. T. Lee. 2020. Artificial Intelligence in Daily Life. Springer.

China. ISBN 9789811576959
• A visual and interactive guide to the basics of neural network
http://jalammar.github.io/visual-interactive-guide-basics-neural-netwo
rks/

• Supervised-Unsupervised-Reinforcement learning concept

https://youtu.be/mMc_PlemSnU