EEEE4133

Artificial Intelligence and

Intelligent Systems
Ir. Ts. Dr Daniel Pu
Department of Electrical and Electronic Engineering

Version 2.0
Introduction to Artificial
Intelligence (AI)
Part 1
AI has already achieved a great deal
 Why Does AI Matter

• AI is estimated to create $13 trillion in

economic value worldwide by 2030,
according to a McKinsey forecast.
• That’s because AI is transforming
engineering in nearly every industry and
application area.
• AI is also used in models that predict machine failure,
indicating when they will require maintenance; health and
sensor analytics.
• Applications are patient monitoring systems; and robotic
systems that learn and improve directly from experience.
Governments are investing in AI

• Malaysia Digital Economy Corporation Sdn Bhd (MDEC) is

expected to complete the development of the National
Artificial Intelligence (AI) Framework by year-end to drive
the country’s AI ecosystem. [NST, 2 Aprl 2019]

• $950 million to “make the UK a global leader in this

technology that will change all our lives" [AI Sector Deal
2018]

• “AI holds the potential to be a major driver of economic

growth and social progress" [White House report, 2016]

• Released domestic strategic plan to become world leader

in AI by 2030 [2017]

• Whoever becomes the leader in this sphere [AI] will become

the ruler of the world" [Putin, 2017]
 WHAT IS INTELLIGENCE?

• These properties include the ability to plan,

solve problems, and in general, reason.
• A simpler definition could be that intelligence
is the ability to make the right decision
given a set of inputs and a variety of possible
actions
 WHAT IS ARTIFICIAL INTELLIGENCE (AI)?

• Artificial intelligence can be described as any

task performed by a program or a machine
that, if a human carried out the same activity,
we would say the human had to apply
intelligence to accomplish the task.

• AI systems will typically demonstrate at least

some of the following behaviours associated
with human intelligence: planning,
learning, reasoning, problem solving,
knowledge representation, perception,
motion, and manipulation.
 Key Components to an AI Workflow

• Success with AI requires more than training an AI model,

especially in AI-driven systems that make decisions and
take action.
• A solid AI workflow involves preparing the data, creating a
model, designing the system on which the model will run,
and deploying to hardware or enterprise systems.
AI uses an algorithm/ fits a model to make inferences from data

Most likely
animal: dog

You might
like this film
 Data Preparation

• Taking raw data and making it useful for an

accurate, efficient, and meaningful model is a
critical step. In fact, it represents most of your AI
effort.

• Data preparation requires domain expertise, such as

experience in speech and audio signals, navigation
and sensor fusion, image and video processing, and
radar and lidar.
Supervised learning
 AI Modeling

Key factors for success in modeling AI systems are to:

• Start with a complete set of algorithms and prebuilt models

for machine learning, deep learning, reinforcement learning,
and other AI techniques
• Use apps for productive design and analysis
• Work in an open ecosystem where AI tools like MATLAB®,
PyTorch, and TensorFlow™ can be used together
• Manage compute complexity with GPU acceleration and
scaling to parallel and cloud servers and on-premise data
centers
 AI problem type Q1: Is it regression or classication?

• Is it classification ? • Is it regression ?

• Are outputs • Are outputs continuous?

categorical/discrete?
 AI problem type Q2: Is it supervised or unsupervised?

• Consider an example
• These flowers are all irises but different species.
• Difficult to tell apart
We can take measurements of each flower

How do we predict species

based on these measurements?
 AI problem type Q2: Is it supervised or unsupervised?

Is it supervised ? Is it unsupervised ?

e.g. Find number of

e.g. Predict species of
distinct species of iris
iris given measurements
based on measurements
 AI problem type Q3: Are there labels or rewards?

Labels: `dog' or `not dog‘ Not labelled, only a reward

after many moves: e.g. `win
or lose'

Techniques like neural

networks such as deep learning Use techniques like Reinforcement
neural network would be learning
suitable
 System Design

• AI models exist within a complete system.

• In automated driving systems, AI for perception
must integrate with algorithms for localization and
path planning and controls for braking,
acceleration, and turning.
 Deployment

• AI models need to be deployed to CPUs, GPUs,

and/or FPGAs in your final product, whether part
of an embedded or edge device, enterprise system,
or cloud.

• AI models running on the embedded or edge device

provide the quick results needed in the field, while
AI models running in enterprise systems and the
cloud provide results from data collected across
many devices.
Development &
Applications
 Driverless cars could be legal on British roads by early 2021

• The British government has launched a consultation that

could pave the way to driverless cars being introduced on
British roads as early as next year.

• The authority is seeking views on whether vehicles

equipped with Automated Lane Keeping System (ALKS)
should be legally defined as an automated vehicle

https://eandt.theiet.org/content/articles/2020/08/driverless-cars-could-become-
legal-on-british-roads-next-spring/
 Brain’s memory inspires neural networks to be less ‘forgetful’

• Researchers say they have successfully addressed

what they call a ‘major, long-standing obstacle to
increasing AI capabilities’ by drawing inspiration
from a human brain memory mechanism known
as ‘replay’.

• “catastrophic forgetting” – upon learning new

lessons, the networks forget what they had learned
before.
• One solution would be to store previously encountered
examples and revisit them when learning something new.
Although such ‘replay’ or ‘rehearsal’ solves catastrophic
forgetting, constantly retraining on all previously learned
tasks is highly inefficient and the amount of data that would
have to be stored becomes unmanageable quickly.

• https://eandt.theiet.org/content/articles/2020/09/brain-s-memory-inspires-
neural-networks-to-be-less-forgetful/
 Startup and Academics Find Path to Powerful Analog AI

• A form of AI has been researched that could

drastically lower the energy required to do typical
AI things like recognize words and images.

• This analog form of machine learning does one of

the key mathematical operations of neural networks
using the physics of a circuit instead of digital logic.
• Analog circuits could save power in neural networks in
part because they can efficiently perform a key
calculation, called multiply and accumulate.

• https://spectrum.ieee.org/tech-talk/robotics/artificial-
intelligence/startup-and-academics-find-path-to-
powerful-analog-ai
 Students’ works on AI

Highway surveillance system

Surveillance Car Park Management System
Module outline
 EEEE4133
Artificial Intelligence and Intelligent Systems

Total Credits: 20.00

Level: 4
Target Students: MEng, MSc and PhD students of Electrical and
Electronic Engineering

Summary Of Content:
This module introduces selected topics from the field of artificial
intelligence with particular focus on the decision trees, support vector
machines, unsupervised learning, neural networks and
reinforcement learning. Students able to grab the knowledge of the
fundamentals of artificial intelligence technologies and the interface to
electronic engineering applications.
 Learning Outcome

By the end of this module, students should be able to:

• LO1: Describe the key principles of artificial intelligence.

• LO2: Discuss the applications and operating principles of

artificial intelligence systems.

• LO3: Describe practical implementation issues of artificial

intelligence in software.

• LO4: Select and critically appraise different artificial

intelligence components.
 Skills you will need

• Coding (in MATLAB)

• First lab is a MATLAB revision session
• Real-world AI engineers use Python (or R)
• Python has powerful toolkits: Tensorow, pytorch, scikitlearn
• Skills are transferrable
• Maths (linear algebra, vectors, matrices)
• Probability (Bayes rule, conditional probabilities)
• A good `cheat sheet' on the kind of maths you will need can be found
http://mlg.eng.cam.ac.uk/zoubin/course04/cribsheet.pdf .
• If you like coding and maths then (hopefully) you will like this course
 What is MATLAB?

• MATrix LABoratory
• Optimised for working with matrices and vectors
• A high-level language
• Don't have to worry about memory or compiling,
like C/C++
• Fully-featured development environment, debugger.
• Available for downloading via University’s email
account.
 Download MATLAB

1. Register to this link

https://au.mathworks.com/mwaccount/register?uri=https%253A%252F%
252Fau.mathworks.com%252Fproducts%252Fget-matlab.html using email
UNM

2. Download https://au.mathworks.com/downloads/web_downloads/.
 Why use MATLAB?

Very easy to write code/debug:

1. Lots of easy-to-use built in tool-boxes containing standard
algorithms from many fields, such as, statistics, image
processing, signal processing, neural networks, wavelets,
communications systems etc.
2. Most operators and functions work on entire arrays/matrices.
3. Popular for experimental/rapid-prototype number crunching.
4. Widely used as a visualization and teaching tool.
5. Widely used by engineers in industry.
 Top 10 programming languages

https://spectrum.ieee.org/computing/software/the-top-programming-
languages-2019
 Why use MATLAB?

• Start with a complete set of algorithms and prebuilt

models for machine learning, deep learning,
reinforcement learning, and other AI techniques.
• Use apps for productive design and analysis.
• Work in an open ecosystem where AI tools like
MATLAB®, PyTorch, and TensorFlow™ can be used
together.
• Manage compute complexity with GPU acceleration
and scaling to parallel and cloud servers and on-
premise data centers.
• MATLAB online demo available
https://matlab.mathworks.com/
 How can you learn MATLAB?

• This will be covered in the first lab

• Basically you will be using MATLAB academy:

http://matlabacademy.mathworks.com
 Why not use MATLAB?

• Not free or open-source

• Not good for large programs.
• Limited support for data structures, object-
oriented programming.
• Interpreted language = slow.
• Free alternatives like Python are very, very good
nowadays.
• Becoming standard for AI/machine learning.
 Lecture topics

Weeks Lecture topics

23 Introduction to AI and Course Outline, Assessment
24 Conditional Probability, Entropy and Decision tree
25 Support vector machines
26 Unsupervised learning
27 Gradient descent and neural networks
28 Neural networks
29 Neural networks
30 Reinforcement learning
31 Reinforcement learning
32 Bayesian inference and networks
33 Bayesian inference and networks
34 Limitations and ethics of AI/Revision
 Assessment

• Coursework 1 (Individual)
• 25% of grade
• Deliverables: Report (docx/pdf) + MATLAB
notebook .m file.
• Coursework 2 (Individual)
• 25% of grade
• Deliverables: Report (docx/pdf) + MATLAB
notebook .m file.
• Exam
• 50% of grade
• Four subjective questions, some questions are
open-ended
Q&A
 References

1. Ville, Barry de, and Padraic Neville (2003), "Chapter 1 - Decision Trees—
What Are They?". Decision Trees for Analytics: Using SAS Enterprise
Miner. SAS Institute. © 2013.
2. M. Tim Jones, (2008), Artificial Intelligence: A Systems Approach by
Infinity Science Press.
3. https://www.zdnet.com/article/what-is-ai-everything-you-need-to-know-
about-artificial-intelligence/
4. https://www.mathworks.com/discovery/artificial-intelligence.html
5. https://eandt.theiet.org/content/articles/2020/09/brain-s-memory-
inspires-neural-networks-to-be-less-forgetful/
6. https://spectrum.ieee.org/tech-talk/robotics/artificial-
intelligence/startup-and-academics-find-path-to-powerful-analog-ai