10/4/2019

EC-350 AI and Decision Support Systems

Week 1
Introduction

Dr. Arslan Shaukat

Acknowledgement: Lecture slides material from

Stuart Russell

Introduction
 Education
– BE and MS Computer System Engineering
• College of EME, NUST

– PhD Computer Science (Machine Learning)

• The University of Manchester, U. K.
 Experience
– More than 9 years

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 2

1
 10/4/2019

Areas of Interest
 Machine Learning & Pattern Recognition

 Digital Image & Speech Processing

 Facial and Speech Emotion Recognition

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 3

Introduction – (Yours)
– Aims and goals for life?

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 4

2
 10/4/2019

Introduction – (Yours)
 Name
 Previous institution/City
 Area/Subject of interest
 Where you see yourself in future

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 5

Course Information
 Course Title: AI and Decision Support Systems
 Course Code: EC-350
 Credit Hours: 3-1
 Semester: Fall 2019
 Email: [email protected]
 Class: Wed & Thu

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 6

3
 10/4/2019

Books
 Textbook:
– S. Russell and P. Norvig, Artificial Intelligence: A Modern
Approach, 3rd Edition 2010
 Reference:
– Artificial Intelligence: Structures and Strategies for Complex
Problem Solving, 6th ed. G. Luger, Addison Wesley, 2009
– Artificial Intelligence: A Systems Approach. M. Tim Jones,
Infinity Science Press, 2008
– Pattern Classification (2nd edition 2006), by Richard O. Duda,
Peter E. Hart and David G. Stork, Wiley Inter-science.
– Pattern Recognition and Machine Learning by Christopher
Bishop, Springer 2006.

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 7

Course Contents
 Introduction to Artificial Intelligence
 Agents, PEAS model, Rationality, Nature & properties of
environment
 Structures of agents
 Problem solving by searching, Uninformed search strategies,
Breadth first search (BFS)
 Depth first search (DFS) , Depth limited search, Iterative
deepening DFS
 Informed search strategies, Greedy best first search, A* search
 Genetic algorithms
 Games, Minimax algorithm, Alpha beta pruning

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 8

4
 10/4/2019

Course Contents
 Introduction to Machine Learning and basic types of classifiers,
Performance parameters for evaluation
 K-Nearest Neighbor (KNN), ROC Analysis
 Bayesian Decision Theory
 Naïve Bayes Classifier
 Neural networks and single layer Perceptron
 Introduction to decision trees

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 9

Marks Distribution
 Written Exams
– 2 Sessional Exams 25%
– Final Exam 35-40%
 Quizzes 10%
 Assignments 5%
 Lab 15%
 Lab project 5-10%

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 10

5
 10/4/2019

Policies
 No extensions in assignment deadlines.
 Quizzes will be unannounced.
 Never cheat.
– “Better fail NOW or else will fail somewhere LATER in life”
 Plagiarism will also have strict penalties.

Adapted from What is Plagiarism PowerPoint

http://mciu.org/~spjvweb/plagiarism.ppt

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 11

Course Learning Outcomes

Level of
Course Learning Outcomes (CLOs) Learning PLO
1. Solving problems using various uninformed and
C3 2
informed search strategies.
2. Apply local search algorithms like Genetic Algorithm
(GA) on optimization problems and perform Minimax C3 2
search on games.
3. Design machine learning systems, demonstrating
understanding of machine learning concepts including C6 3
feature extraction and classification.
4. Implement projects in the lab work that use Python
and MATLAB for execution of the theoretical knowledge
P2 12
gained during class lectures, requiring some
independent reading, programming and learning.

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 12

6
 10/4/2019

Artificial Intelligence (AI)

What is Intelligence?
 A machine is intelligent if it has:
– Ability to learn and adapt from environment
– Ability to acquire knowledge
– Ability to memorize and proceed
– Ability to plan and schedule
– Ability to recognize voice, patterns, faces
– Ability to understand and perceive
– Ability to solve complex problems

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 14

7
 10/4/2019

What is Artificial Intelligence

 If a machine could do all or most of the mentioned tasks,
we can call that machine an intelligent machine
 The ‘intelligence’ in the machine is NOT natural but
artificial
 Name was coined in 1956.

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 15

What is Artificial Intelligence

 Concerned with building intelligence in artificial – man
made devices
 Making machines to behave like humans – if we consider
humans to be intelligent
 Making machines to behave in most rational manner
– Thinking/reasoning intelligently
– Acting/behaving intelligently

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 16

8
 10/4/2019

Approaches to AI
 Think
– Systems that “think” like humans
– Systems that “think” rationally
 Act
– Systems that “act” like humans
– Systems that “act” rationally

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 17

To Think
 Make machines to think intelligently [like human beings]
as if they have real conscious minds
 Deals with “How brain works”
 Actually tries to recreate the functions of the inside of the
brain
 Advocates Machine can replace a human being
 Many researchers believe it is NOT possible to attain

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 18

9
 10/4/2019

To Act
 Machines can be made to act as if they are intelligent
 Treats brain as a black box which creates output after it
receives input
 Is NOT concerned with inner functionality of brain
 Most of the research in done in this field

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 19

Acting Humanly – Turing Test

 Alan Turing (1950) “Computing Machinery and
intelligence” at The University of Manchester
 Can Machine behave intelligently?
 Operational test for intelligent behavior

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 20

10
 10/4/2019

Turing Test
 Suggested major components of AI
– natural language processing
• to enable it to communicate successfully in English
– knowledge representation
• to store what it knows or hears
– automated reasoning
• to use the stored information to answer questions and to draw new
conclusions
– machine learning
• to adapt to new circumstances and to detect and extrapolate patterns.
– computer vision
• to perceive objects
– robotics
• to manipulate objects and move about.
04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 21

Thinking Humanly – Cognitive Science

 A system is intelligent if it thinks like a human.
 If a system uses the same reasoning processes as a
human, then it is intelligent
 Requires scientific theories of internal working of brain
 Three ways:
– Try to catch our own thoughts as they go by
– Through psychological experiments
– Observing the brain in action
 Cognitive Science brings together computer models from
AI and experimental techniques from psychology
 To construct theories of the workings of human mind
04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 22

11
 10/4/2019

Thinking Rationally – Laws of Thought

 A system is intelligent if it thinks rationally.
 An intelligent system is one that follows sound reasoning
processes that always lead to correct outcomes.
 Logicians in the 19th century developed a precise
notation for statements about all kinds of objects in the
world and the relations among them.
 This leads to the study of logic and formal reasoning
 However, logic has its problems.
– Formalizing common-sense knowledge
– Computational issues

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 23

Acting Rationally
 A system is intelligent if it acts rationally.
 A rational agent is one that acts so as to achieve the best
possible (expected) outcome, given its knowledge and
ability.
 We can build agents that do certain tasks intelligently
without having human-level intelligence.
 All the skills needed for Turing Test also allow an agent
to act rationally.

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 24

12
 10/4/2019

Foundations of AI
 Many disciplines provided the foundation for Artificial
Intelligence. Few to mention are as follows:
 Philosophy (428 BC – Present)
– Made AI conceivable by considering the ideas that the mind is
in some ways like a machine
 Mathematics (800 – Present)
– Set the groundwork for understanding computation and
reasoning about algorithms.
 Economics (1776 – Present)
– Formalized the problem of making decisions that maximize the
expected outcome to the decision maker.

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 25

Foundations of AI
 Neuroscience (1861 – Present)
– Discovered some facts about how the brain works and the ways in which it is
similar to and different from computers.
 Psychology (1879 – Present)
– Studied about humans and animals thinking and actions
 Computer Engineering (1940 – Present)
– Provided the ever-more-powerful machines that make AI applications
possible.
 Control theory & Cybernetics (1948 – Present)
– Deals with designing devices that act optimally on the basis of feedback
from the environment.
 Linguistics (1957 – Present)
– Showed how does language relate to thought.

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 26

13
 10/4/2019

Few Applications using AI

 Game playing
 Robotics and robot vehicles
 Tutoring systems
 Medicine
 Biometrics
 Speech Recognition
 Autonomous planning and
Scheduling
 Spam fighting

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 27

Few applications using AI

 Deep Blue
– http://www-
03.ibm.com/ibm/history/ibm100/us/en/icons/deepblue/
 Aaron the Robot as an Artist
– http://aaronshome.com/aaron/index.html
– http://www.scinetphotos.com/aaron.html
 Speech recognition
– http://www.nuance.com/
 Microsoft
– https://www.microsoft.com/en-us/research/research-
area/artificial-intelligence/

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 28

14
 10/4/2019

Few applications using AI

 Google
– https://ai.google/
 Honda Humanoid Robot
– https://global.honda/innovation/robotics/ASIMO.html
 Mars rover curiosity
– http://mars.jpl.nasa.gov/msl/mission/overview/
 Association for the Advancement of AI
– http://www.aaai.org
 AI Book
– http://aima.cs.berkeley.edu/

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 29

Intelligent Agents

15
 10/4/2019

Intelligent Agent
 An agent is anything that can be viewed as perceiving its
environment through sensors and acting upon that
environment through actuators.
 Agent Examples
– A human agent has eyes, ears, and other organs for sensors and
hands, legs, mouth and other body parts for actuators
– A robotic agent substitutes cameras and infrared range finders
for the sensors and various motors for the actuators.
– A software agent receives keystrokes, network packets, file
contents as sensory input and acts upon the environment by
displaying on screen, sending network packets and writing files.

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 31

Percept & Percept Sequence

 Percept refers to agent’s perceptual input at any given
instance
 Percept sequence is the complete history of everything
agent has ever perceived
– An agent’s choice of action at any given instance can depend
on the entire percept sequence observed to-date

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 32

16
 10/4/2019

Agent Function
 Mathematically, an agent’s behavior is described by the
agent function that maps any given percept sequence to
an action
– For most agent this would become a very large table. May be of
infinite size
– Have to put a bound on the length of the percept sequence we
want to consider
– Internally, agent function for an artificial agent shall be an
agent program
– Agent Function is a mathematical description
– Agent Program is a concrete implementation

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 33

An Intelligent Agent
Sensors
percepts

Environment
Agent

actions

Actuators

The agent function maps from percept histories to actions:

f : P*  A
P* = Percept History
A = Actions

The agent program runs on the physical architecture to produce f

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 34

17
 10/4/2019

Vacuum-Cleaner Agent

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 35

Vacuum-Cleaner Agent

 Locations: square A, square B

 Percepts: location and contents, e.g., [A, Dirty]
 Actions: Left, Right, Suck, NoOp

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 36

18
 10/4/2019

Vacuum-Cleaner Agent

Percept Sequence Action

[A, Clean] Right
[A, Dirty] Suck
[B, Clean] Left
[B, Dirty] Suck

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 37

Good Behavior - Rationality

 Rational Agent is the one which does the “right thing”
based on what it perceives and the actions it performs.
 Every entry in the table for the agent function is filled out
correctly
 The right action is the one that will cause the agent to be
most successful
 How can we say an agent is successful?
 We need some criterion to measure its success

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 38

19
 10/4/2019

Performance Measure
 An objective criterion for success of an agent's behavior
 e.g., performance measure of a vacuum-cleaner agent
could be amount of dirt cleaned up, amount of time
taken, amount of electricity consumed, amount of noise
generated, etc
 No single universal criterion for every agent
 As a general rule, it is better to design a performance
measure according to what one actually wants in the
environment rather than according to how one thinks the
agent should behave
 The selection of performance measure is not always easy

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 39

Rationality
 What is rational at given time depends upon four things
– The performance measure that defines criterion for success
– The agent’s prior knowledge of the environment
– The actions that agent can perform
– The agent’s percept sequence to-date

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 40

20
 10/4/2019

Rational Agent
 For each possible percept sequence, a rational agent
should select an action that is expected to maximize its
performance measure, given the evidence provided by
the percept sequence and whatever built-in knowledge the
agent has.
 Vacuum-cleaner agent?

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 41

Vacuum-Cleaner Agent
 The performance measure awards one point for each
clean square at each time step, over a "lifetime" of 1000
time steps.
 The "geography" of the environment is known a priori
but the dirt distribution and the initial location of the
agent are not.
 The only available actions are Left, Right, and Suck.
 The agent correctly perceives its location and whether
that location contains dirt.

04/10/2019 EC-350 AI and DSS Dr Arslan Shaukat EME (NUST) 42

21