8/28/2003

Soft Computing

Neuro-Fuzzy and Soft Computing

chapter 1
J.-S.R. Jang
Bill Cheetham
Kai Goebel

Soft Computing

What is covered in this class?

We will teach techniques useful in creating
intelligent software systems that can deal with
the uncertainty and imprecision of real world
problems
Some components of Intelligent systems are
human-like - they possess human-like
expertise within a specific domain,
adaptable - they adapt themselves and learn to
do better in a changing environment, and
explanations - they explain how they make
decisions or take actions
2

Page 1
1

8/28/2003

Soft Computing

How will we teach the techniques?

We will present
multiple techniques from Soft Computing +,
when each technique is applicable
examples of industrial applications

If the only tool you have is a hammer, then

every problem looks like a nail
- anonymous

Soft Computing

Soft Computing
Soft computing is an emerging
approach to computing which
parallels the remarkable ability of
the human mind to reason and
learn in an environment of
uncertainty and imprecision
- Lotfi Zadeh

Page 2
2

8/28/2003

Soft Computing

Why Soft Computing?

Farming
corn & cows

Industrial
Revolution

Manufacturing
chairs & cars

Information
Revolution

Service

The information revolution

going on is allowing us to
automate information
processing tasks which require
intelligence much like the
industrial revolution automated
manufacturing tasks
Soft Computing techniques
have already been applied
successfully.

content and code

5

Soft Computing

What is Soft Computing?

Soft Computing is a field that currently includes
Fuzzy Logic
Neural Networks
Probabilistic Reasoning(Genetic Algorithms, BBN), and
Other related methodologies
Case-Based Reasoning

Soft Computing combines knowledge,

techniques, and methodologies from the sources
above to create intelligent systems
6

Page 3
3

8/28/2003

Soft Computing

Fuzzy Logic - Kai

Sets with fuzzy boundaries
A = Set of tall people
Crisp set A

Fuzzy set A

1.0

1.0
.9

Membership

.5

function

170

170

Heights
(cm)

180

Heights
(cm)

Soft Computing

Fuzzy Set Theory - Kai

Fuzzy set theory provides a systematic calculus
to deal with imprecise or incomplete information
Fuzzy if-then rules are used in fuzzy inference
systems
If <1> is tall and <1> is athletic then <1> is good
basketball player.
B B T-norm

A A

Page 4
4

8/28/2003

Soft Computing

Neural Networks - Kai

Pattern matching technique where inputs are
matched with a specific output pattern.
Network architecture
Modeled after the neurons in the brain.
Weights on the links
Learns by modifying the weights
x1

y1

x2

y2

Soft Computing

Genetic Algorithms - Bill

Use Idea of Evolution to Guide Search

Human
Evolution

Find Max of
a Function
10

Page 5
5

8/28/2003

Soft Computing

Genetic Algorithms - Bill

An optimization technique
10010110
01100010
10100100
10011001
01111101
...
...
...
...

Elitism

Selection Crossover Mutation

Current
generation

10010110
01100010
10100100
10011101
01111001
...
...
...
...

Next
generation

11

Soft Computing

Case-Based Reasoning - Bill

A methodology of
solving new problems
by adapting the
solutions of previous
similar problems

12

Models the way

experts reason using
their experience

Page 6
6

8/28/2003

Soft Computing

What Is CBR? - Quiz

What is 12 x 12?

144
What is 12 x 13?
12 x 12 + 12
156

13

Soft Computing

Other Techniques Bill & Kai

Bayesian belief networks

Earthquake

Burglary
0.9

represent and reason with

probabilistic knowledge

0.3
Alarm
0.7

0.9
JohnCalls

MaryCalls
x<a

Decision Trees

n
y<c

y<b

classification using tree

structure

y
z=f1

Least-squares estimator
statistical regression

z=f3

z=f4

+3
+2
y +1
0

y=x-3
0

14

n
z=f2

Hybrid approaches
use multiple techniques

Page 7
7

8/28/2003

Soft Computing

Soft Computing is a Hybrid Method

Neural
Character
Recognizer

x1
x2

y1
y2

dog
dag

Animal?

Knowledge
base

dog
15

Soft Computing

How does SC Relate to Other Fields

What is AI?
AI is the study of agents that exist in an environment
and perceive and act. (S. Russell and P. Norvig)

AI is the art of making computers do smart things. (Waldrop)

AI is a programming style, where programs operate on data
according to rules in order to accomplish goals. (W. A. Taylor)
AI is the activity of providing such machines as computers
with the ability to display behavior that would be regarded as
intelligent if it were observed in humans. (R. McLoed)
16

Page 8
8

8/28/2003

Soft Computing

How does SC Relate to Other Fields

What is AI? (Jang)
The long term goal of AI research is the
creation and understanding of machine
intelligence

Broad
Definition

Narrow
Definition

Conventional AI research focuses on an

attempt to mimic human intelligent behavior
by expressing it in symbolic rules

17

Soft Computing

How does SC Relate to Other Fields

What is an Expert System (ES)?
Questions
Responses

Inference
Engine

User

Knowledge
Acquisition

KB
Rules if a then b
Facts a is true

Knowledge
Engineer
18

Page 9
9

8/28/2003

Soft Computing

Types of Programming
Advantages

Disadvantages

Functional
Programming

Precise
Deterministic

Reasoning
Learning

Symbolic
Programming
(AI)

Reasoning
Learning

Uncertainty
Confidence

Soft Computing

Uncertainty
Confidence

Precise
Deterministic

19

Soft Computing

Stages of Reasoning
Functional
Programming

Complex Math

Computers

Humans

Logic

Evolution
Experience
Uncertainty

Symbolic
Programming
(AI)

Soft Computing

20

Page 10
10

8/28/2003

Soft Computing

Soft Computing Characteristics

Human Expertise (if-then rules, cases,
conventional knowledge representations)
Biologically inspired computing models (NN)
New optimization techniques (GA, simulated annealing)
Model-free learning (NN, CBR)
Fault tolerance (deletion of neuron, rule, or case)
Real-world applications (large scale with uncertainties)
21

Soft Computing

Soft Computing Entertainment

Star Trek
Kirk and Spock are the classic fuzzy
and crisp reasoners

Errand of Mercy episode

Klingon army attacks a neutral planet
Kirk and Spock beam down
Enterprise is chased away
Inhabitants are not concerned
Should they try and help?

Kirk
Fuzzy

Spock
Crisp

Spock odds of succeeding are 7,249.5 to 1

Kirk we should do it anyway
22

Page 11
11

8/28/2003

Soft Computing

Soft Computing in History

If a man will begin with certainties,
he will end with doubts,
but if he will be content to begin with doubts,
he shall end in certainties.
- Francis Bacon 1605
THE ADVANCEMENT OF LEARNING

23

Page 12
12