Introduction to Soft Computing

libyantawdryAI and Robotics

Oct 23, 2013 (3 years and 9 months ago)

127 views

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Introduction to Soft Computing

Docent Xiao
-
Zhi Gao

Department of Automation and Systems Technology

xiao
-
zhi.gao@aalto.fi

Institute of Intelligent Power Electronics


IPE

Page

‹#›

What is Soft Computing?


Unlike conventional (hard) computing,
soft computing applies a combination of
different methods: Fuzzy Logic (FL),
Neural Networks (NN), and Genetic
Algorithms (GA)


Soft computing mimics human thinking


For example, reasoning/inference in
decision making and imprecise information
processing

Over 100 000 publications today

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Zadeh’s Definition on 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 A. Zadeh, Father of Soft Computing,
University of California
-
Berkeley

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Lotfi A. Zadeh (1921
-
)

Director, Berkeley Initiative in Soft Computing

(BISC)

Professor Emeritus

University of California, Berkeley

Berkeley, California

USA

Institute of Intelligent Power Electronics


IPE

Page

‹#›

FL, NN, and GA


Fuzzy Logic


Fuzzy IF
-
THEN reasoning rules


Neural Networks


Capable of approximation and adaptation


Genetic Algorithms


All
-
purpose

derivative
-
free

optimization
method

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Fuzzy Systems (FS)


Fuzzy systems can approximate
decision making processes of human
under uncertain environments


Fuzzy systems are based on fuzzy sets
and fuzzy reasoning


fuzzy sets are generalization of crisp sets


fuzzy reasoning infers conclusions from
known facts using given fuzzy rules

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Fuzzy Systems (FS)


Modeling of human thinking


Numerical

values are
not

efficient


Linguistic

variables exist in real world

»
An example:


Chatting with a stranger on the phone


Estimation of your partner’s age:
(40? probability of 40? or
about

middle
-
aged
?)


about middle aged

(linguistic term)

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Fuzzy Systems (FS)


Fuzzy logic derives conclusions based
on
given

fuzzy IF
-
THEN rules and
known

facts


An example:

»
Given a fuzzy rule: IF bath is very hot, THEN
add a lot of cold water

»
Known fact: bath is a little hot

»
Conclusion: how much cold water should be
added?


Fuzzy conclusion: a small amount of cold water

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Structure of Fuzzy Systems (FS)

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Structure of Human Brain

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Human Neurons

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Neural Networks (NN)


Neural networks are highly simplified models of
human brain to deal with specific tasks


massively connected neurons


Several kinds of neural networks are proposed


feedforward neural networks


recurrent neural networks


supervised neural networks


unsupervised neural networks


Applications of neural networks are intensive


Institute of Intelligent Power Electronics


IPE

Page

‹#›

Neural Networks (Feedforward)

Back
-
Propagation
Neural Networks

(Multiple Layer
Perceptron)

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Neural Networks (Recurrent)

Hopfield Neural Networks

(Hopfield [1982] [1986])

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Charles Darwin (1809
-
1882)

Institute of Intelligent Power Electronics


IPE

Page

‹#›

’The Origin of Species’ by Darwin [1859]

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Species Evolution and Natural Selection

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Flow Chart of Genetic Algorithms

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Fusion of Soft Computing Methods


Soft computing methods are considered
complementary

rather than
competitive


There are various kinds of combinations
of soft computing methods


Fusion of soft computing methods not
only stays on
algorithm

level, but also
system

level

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Framework of Soft Computing

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Example I:

Genetic Algorithms + Fuzzy Logic

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Example II:

Neural Networks + Fuzzy Logic

ANFIS by R. Jang
[Jang 97]

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Model Construction:

Comparison between SC and HC


Mathematical models are linear and
non
-
linear functions

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Model Construction (Traditional Rules)


Rules with sharp boundaries

IF 0 ≤ x ≤ 1, THEN y=1

IF 1 ≤ x ≤ 2, THEN y=0.99



IF 9 ≤ x ≤ 10, THEN y=0

IF 0≤x ≤ 1, THEN y=f(x)

IF 1 ≤ x ≤ 2, THEN y=g(x)



IF 9 ≤ x ≤ 10, THEN y=h(x)

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Model Construction (Fuzzy Logic)


Multiple fuzzy rules are fired with each
input
simultaneously

IF x

0
, THEN y


1

IF x


5, THEN y


0.5

IF x


10, THEN y


0

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Model Construction (Neural Networks)


Model
-
free

approximation based on
training data

x=0, y=1.0

x=0.5, y=0.99



x=10, y=0

Institute of Intelligent Power Electronics


IPE

Page

‹#›

An Application Example (HC vs SC)

Inverted Pendulum

Difficult to solve with Hard Computing

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Neural Networks
-
based Solution
to Inverted Pendulum Control

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Neural Networks
-
based Solution
to Inverted Pendulum Control

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Fuzzy Logic
-
based Solution to
Inverted Pendulum Control

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Fuzzy Logic
-
based Solution to
Inverted Pendulum Control

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Power Control in Mobile Communications
Systems: A Case Study of HC and SC

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Mobile Power Control with Hard Computing

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Mobile Power Control with Soft Computing

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Features of Soft Computing


Advantages of soft computing methods


Take advantage of human intuition


Cope with

imprecision

and
uncertainty


Disadvantages of soft computing methods


Complex structures and algorithms


Heavy computation burden


Still not as widely used as Hard Computing in
engineering areas yet

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Soft Computing and Hard Computing


Soft computing and hard computing are
also
complementary

rather than
competitive
[Ovaska 99, 02, 04]

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Fusion of SC and HC

Institute of Intelligent Power Electronics


IPE

Page

‹#›

SC and HC (Independent)

Institute of Intelligent Power Electronics


IPE

Page

‹#›

SC and HC (Parallel)

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Fusion of SC & HC in Missile Autopilot

McDowell [1997]

Institute of Intelligent Power Electronics


IPE

Page

‹#›

SC and HC (Serial)

PostProcessing

PreProcessing

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Fusion of SC and HC in
Electric Load Forecasting

Kamiya

[2003]

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Fusion of SC and HC in Power
Control of Mobile Communications

Gao [2001]

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Fusion of SC and HC in Boiler
-
Turbines Systems Fault Diagnosis

Ben
-
Abdennour

[1996]

Institute of Intelligent Power Electronics


IPE

Page

‹#›

HC
-
Designed SC & SC
-
Designed HC

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Fusion of SC and HC in
Time Series Prediction


Akhmetov [2001]

General
Parameter
Method

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Fusion of SC and HC in Nuclear

Steam Generator Control Systems

Zhao[1997]

Institute of Intelligent Power Electronics


IPE

Page

‹#›

HC
-
Assisted SC & SC
-
Assisted HC

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Fusion of SC
and HC in
Large
-
Scale
Power Plant
Control

Kamiya

[2003]

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Problems of HC Solved by SC

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Problems of SC Solved by HC

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Features of Fusion of SC and HC


Fusion of soft computing and hard
computing can provide us with an
improved performance over employing
either of them
separately


Individual
strengths

are enhanced


Individual
drawbacks

are overcome


Fusion of soft computing and hard
computing leads to hybrid intelligent
systems

Institute of Intelligent Power Electronics


IPE

Page

‹#›

SC Applications: Control


Heavy industry
(Matsushita,
Siemens, Stora
-
Enso)


Home appliances
(Canon, Sony,
Goldstar, Siemens)


Automobiles
(Nissan, Mitsubishi,
Daimler
-
Chrysler,
BMW, Volkswagen)


Spacecrafts (NASA)

Institute of Intelligent Power Electronics


IPE

Page

‹#›

SC Applications: Business


hospital stay prediction


TV commercial slot evaluation


address matching


fuzzy cluster analysis


sales prognosis for mail order
house


multi
-
criteria optimization etc.


source: FuzzyTech


supplier evaluation for
sample testing


customer targeting


sequencing


scheduling


optimizing R&D


projects


knowledge
-
based
prognosis


fuzzy data analysis

Institute of Intelligent Power Electronics


IPE

Page

‹#›

SC Applications: Finance


Fuzzy scoring for mortgage applicants


creditworthiness assessment


fuzzy
-
enhanced score card for lease risk assessment


risk profile analysis


insurance fraud detection


cash supply optimization


foreign exchange trading


trading surveillance


investor classification etc.


Source: FuzzyTech

Institute of Intelligent Power Electronics


IPE

Page

‹#›

SC Applications: Robotics



Fukuda’s lab

http://www.mein.
nagoya
-
u.ac.jp

Joseph F.
Engelberger

‘We are proud to
announce that the
HelpMate

Robotic
Courier
has been acquired by
Pyxis Corporation


Entertainment
robot AIBO

Institute of Intelligent Power Electronics


IPE

Page

‹#›

SC Applications: Others




Statistics


Social sciences


Behavioural sciences


Biology


Medicine

Institute of Intelligent Power Electronics


IPE

Page

‹#›

SC Today (Zadeh)


Computing with Words (CW)


Theory of Information Granulation
(TFIG)


Computational Theory of Perceptions
(CTP)


Refer to [Zadeh 9
9,
01, 03]

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Institute of Intelligent Power Electronics


IPE

Page

‹#›

Conclusions


Soft computing methods are promising
in dealing with real
-
world problems


Theory of soft computing should be
further investigated


Fusion of soft computing and hard
computing is necessary for developing
high
-
performance and low
-
cost systems