Introduction Fuzzy Logic - Introduction
Adriano Cruz NCE e IM/UFRJ
[email protected]
Computers are useless, they can only give you answers. Pablo Picasso
J. Yen, R. Langari, “Fuzzy Logic: Intelligence, Control and Information”, Prentice Hall, 1999
J. R. Jang, C. Sun, E. Mizutani, “Neuro-Fuzzy and Soft Computing: A Computational Approach to Learning and Machine Intelligence, Prentice Hall, 1997
C. von Altrock, “Fuzzy Logic & NeuroFuzzy Applications Explained”, Prentice Hall PTR, 1995
No. 3
Summary
@2001 Adriano Cruz
NCE e IM - UFRJ
No. 4
Artificial Intelligence?
Introduction Fuzzy Sets Fuzzy Set Operations Fuzzy Systems
@2001 Adriano Cruz
No. 2
H. T. Nguyen, E. A. Walker, “A First Course in Fuzzy Logic”, Chapman & Hall/CRC, 2000 Bart Kosko, “Fuzzy Thinking”, Harper Collins Publishers, 1994, ISBN 0-00-654713-3 L. H. Tsoukalas, R. E. Uhig, “Fuzzy and Neural Approaches in Engineering”, John Wiley and Sons, Inc, 1997
Slides and notes: http://equipe.nce.ufrj.br/adriano/fuzzy/bibliogr-ic.htm
NCE e IM - UFRJ
NCE e IM - UFRJ
Bibliography 2
@2001 Adriano Cruz
Light travels faster than sound. That is the reason why some people look brighter until they start talking. Linux Journal
@2001 Adriano Cruz
Bibliography 1
Adriano Cruz NCE-IM UFRJ
[email protected]
NCE e IM - UFRJ
No. 5
“AI is the activity of providing such machines as computers with the ability to display behaviours that would be regarded as intelligent if it were observed in humans” (R. McLeod)
“AI is the study of agents that exist in an environment, perceive and act.” (S. Russel and P. Norvig)
@2001 Adriano Cruz
NCE e IM - UFRJ
No. 6
1
Artificial Intelligence?
AI emphasizes symbolic processing
Acts on higher levels of intelligence
AI seeks to understand
@2001 Adriano Cruz
NCE e IM - UFRJ
Computational Intelligence
No. 7
Computational Intelligence
Acts on lower levels of Intelligence Uses learning extensively Pattern recognition and heuristics play important roles
@2001 Adriano Cruz
NCE e IM - UFRJ
Computational Intelligence
Fuzzy Logic
Fuzzy Logic
Artificial Neural Networks
Artificial Neural Networks
Evolutionary Systems
Evolutionary Systems
Swarm Intelligence
Swarm Intelligence
Hybrid Systems
Hybrid Systems
@2001 Adriano Cruz
NCE e IM - UFRJ
No. 9
Fuzzy Logic
@2001 Adriano Cruz
NCE e IM - UFRJ
No. 10
Fuzzy Logic
Logic that deals mathematically with imprecise information usually employed by humans.
Multi-valued logic that extends Boolean logic usually employed in computer science.
@2001 Adriano Cruz
No. 8
NCE e IM - UFRJ
No. 11
Used to alleviate difficulties in developing and analysing complex control systems.
Function approximator
Decision systems
@2001 Adriano Cruz
NCE e IM - UFRJ
No. 12
2
Fuzzy Logic
Computational Intelligence
Who is greater than 1.80 m?
Who is tall?
Who weighs more than 100 kg?
Who is heavy?
The driver was heavy and tall.
@2001 Adriano Cruz
NCE e IM - UFRJ
No. 13
Artificial Neural Networks
ANN consist of many simple computing elements – usually simple nonlinear summing operations – highly connected by links of varying strength.
NCE e IM - UFRJ
No. 15
Artificial Neural Networks
Evolutionary Systems
Swarm Intelligence
Hybrid Systems
@2001 Adriano Cruz
NCE e IM - UFRJ
No. 14
ANNs are able to learn from examples.
Function approximators.
Solutions not always correct.
ANNs are able to generalize the acquired knowledge.
@2001 Adriano Cruz
NCE e IM - UFRJ
No. 16
Neural Networks
Neurons
@2001 Adriano Cruz
Fuzzy Logic
ANNs
Computational models that try to emulate the structure of the human brain wishing to reproduce at least some of its flexibility and power.
@2001 Adriano Cruz
NCE e IM - UFRJ
No. 17
@2001 Adriano Cruz
NCE e IM - UFRJ
No. 18
3
Structure
Training Weight values change during the training process Values are presented at the inputs and outputs are compared to the desired values. Wrong outputs cause weights to change in order to reduce the error Process is repeated with different inputs till the ANN is able to give the correct answers Hopefully the ANN will be able to give the correct answer even to inputs that were not trained. NCE e IM - UFRJ No. 20 @2001 Adriano Cruz
Inputs
Input layer
Weight Hidden layer
Matrix 1
@2001 Adriano Cruz
Weight Matrix 2
Output layer
Outputs
NCE e IM - UFRJ
No. 19
Computational Intelligence
Evolutionary Systems
Fuzzy Logic
Artificial Neural Networks
Evolutionary Systems
Swarm Intelligence
Hybrid Systems
@2001 Adriano Cruz
NCE e IM - UFRJ
No. 21
@2001 Adriano Cruz
The Metaphor
NCE e IM - UFRJ
NCE e IM - UFRJ
No. 22
Adaptation
The metaphor that lays behind GAs is the natural selection. The problem of each species in the nature is seek for the best adaptations in order to survive in a hostile environment that is in constant modification.
@2001 Adriano Cruz
ES are global search and optimization algorithms modelled from natural genetic principles such as natural selection. They are stochastic searching methods. Good solutions will survive and be combined by the natural selection process. At the end the most fit will survive.
No. 23
The sets of characteristics of an individual, that distinguishes from everybody else, defines its survival capacity.
These characteristics are determined by its genetic material.
@2001 Adriano Cruz
NCE e IM - UFRJ
No. 24
4
Mechanisms
GA Flux begin
The competition for scarce resources makes the apts survive and reproduce. Through reproduction the genes from individuals are transmitted to their descendants. This continuous process of selection and reproduction of the best individuals may conduct to more adpated individuals.
@2001 Adriano Cruz
NCE e IM - UFRJ
No. 25
Randomly Initial Population
Mutation
Current generation
Artificial Neural Networks
Evolutionary Systems
Swarm Intelligence
Hybrid Systems
@2001 Adriano Cruz
NCE e IM - UFRJ
@2001 Adriano Cruz
NCE e IM - UFRJ
NCE e IM - UFRJ
No. 26
Swarm Intelligence
No. 27
Swarm Intelligence (SI) is the property of a system whereby the collective behaviours of (unsophisticated) agents interacting locally with their environment cause coherent functional global patterns to emerge. SI provides a basis with which it is possible to explore collective (or distributed) problem solving without centralized control or the provision of a global model.
@2001 Adriano Cruz
NCE e IM - UFRJ
No. 28
Motivations
Distributed, no central control or data source; No (explicit) model of the environment; Perception of environment, i.e. sensing; Ability to change environment.
@2001 Adriano Cruz
Next Generatio
OK?
Characteristics of a swarm
Generates Sons
No
Fuzzy Logic
Selects Parents
Crossing
Computational Intelligence
Evaluates
No. 29
Robust nature of animal problem-solving – simple creatures exhibit complex behaviour; – behaviour modified by dynamic environment.
Emergent behaviour observed in: – bacteria – ants – bees – ...
@2001 Adriano Cruz
NCE e IM - UFRJ
No. 30
5
Computational Intelligence
Ant Colonies
Ants are behaviourally unsophisticated; collectively perform complex tasks. Ants have highly developed sophisticated sign-based stigmergy – communicate using pheromones; – trails are laid that can be followed by other ants.
Stigmergy is a method of indirect communication in a self-organising emergent system where its individual parts communicate with one another by modifying their local environment.
@2001 Adriano Cruz
NCE e IM - UFRJ
No. 31
Hybrid Systems
NCE e IM - UFRJ
Fuzzy Logic
Artificial Neural Networks
Evolutionary Systems
Swarm Intelligence
Hybrid Systems
@2001 Adriano Cruz
NCE e IM - UFRJ
No. 32
History
Each intelligent technique has its particular strengths and weakness and cannot be applied to universally to every problem. Mixing together these techniques systems improve the quality of the solutions and allows application to different tasks.
@2001 Adriano Cruz
No. 33
40s 50s 60s 70s
AI
ANNs
47 Cybernetics
43 Neuron Model
56 AI
57 Perceptron Adaline Madaline 74 Back80Propagation Self orgazing map 82 Hopfield 83 Boltzmann Mach
60 Lisp Expert Systems
80s 90s
@2001 Adriano Cruz
FL
EA
65 Fuzzy Sets 74 Fuzzy Control Genetic Algorithm 85 Fuzzy modelling (TSK model)
Neuro-Fuzzy modelling
NCE e IM - UFRJ
Immune modelling Genetic programming
No. 34
Aristotle
Why do we reason as we do?
@2001 Adriano Cruz
NCE e IM - UFRJ
No. 35
Macedonian philosopher who lived between 384 e 322 AC Studied under Plato in the Academy Creator of formal logic His father Nichomachus was court physician to King Amyntas Associates the spirit of observation and a classification instinct He was considered during the middle ages the philosopher He shaped much of the western mind.
@2001 Adriano Cruz
NCE e IM - UFRJ
No. 36
6
Aristotle X Buddha
Limitations of the Aristotle’ Aristotle’s Logic
Objects are separated on very clear categories One object either belongs to a category or another Either you are or not Helps to separate objects into well defined categories.
@2001 Adriano Cruz
NCE e IM - UFRJ
No. 37
Everything must either be or not be, whether in the present or in the future. Aristotle
I have not explained that the world is eternal or not eternal. I have not explained that the world is finite or infinite. The Buddha
@2001 Adriano Cruz
Why fuzzy logic?
NCE e IM - UFRJ
No. 38
Why fuzzy logic?
Every language is vague. All traditional logic habitually assumes that precise symbols are being employed. It is therefore not applicable to this terrestrial life, but only to an imagined celestial one. Everything is vague to a degree you do not realize till you have tried to make it precise.
As far as the laws of Mathematics refer to reality, they are not certain; and as far as they are certain, they do not refer to reality. Albert Einstein
Bertrand Russel @2001 Adriano Cruz
NCE e IM - UFRJ
No. 39
@2001 Adriano Cruz
How to classify?
NCE e IM - UFRJ
No. 40
To be or not to be?
Happy people Small rooms High temperatures Faster cars High tax rates High people
@2001 Adriano Cruz
NCE e IM - UFRJ
No. 41
Bertrand Russel, while trying to formalize Mathematic had difficulties due to the liar’s paradox. “I am lying.” If Eubulides‘ statement was true, then he is lying when he says “I am lying” and so he isn't, i.e. his statement is false. If his statement is false, then he isn't lying when he tells us he is, and so his statement is true.
@2001 Adriano Cruz
NCE e IM - UFRJ
No. 42
7
Answer: To be and not to be.
The Detractors
Consider the set of all sets that are not members of its own set. Is it a member of this set?
If it is a member then it is not, but if it is not then it is.
Fuzzy theory is wrong, wrong, and pernicious. What we need is more logical thinking, not less. The danger of fuzzy logic is that it will encourage the sort of imprecise thinking that has brought us so much trouble. Fuzzy logic is the cocaine of the science. Prof. William Kaham - U. Cal - Berkeley
@2001 Adriano Cruz
NCE e IM - UFRJ
No. 43
@2001 Adriano Cruz
The Detractors
NCE e IM - UFRJ
No. 45
Fuzzy Thinking
Lotfy Zadeh. “Fuzzy Sets”, Information na Control, 1965 Principle of Incompatibility
– As the complexity of a system increases, our ability to make precise yet significant descriptions about its behaviour diminishes until a threshold is reached beyond which precision and significance (or relevance) become almost mutually exclusive characteristics. Lofty Zadeh @2001 Adriano Cruz
Yes
Yes
@2001 Adriano Cruz
NCE e IM - UFRJ
NCE e IM - UFRJ
No. 46
Fuzzy Thinking No
No
No. 44
The Beginning
“Fuzzification” is a kind of scientific permissiveness. It tends to result in socially appealing slogans unaccompanied by the discipline of hard scientific work and patient observation. Prof. Rudolf Kalam - U. Florida - Gainesville
@2001 Adriano Cruz
NCE e IM - UFRJ
No. 47
If the interest rate is high and the deficit is high then there will be a recession If rush hour then diminish the interval between busses If the tyre skids then loose the brake a bit If the soil is very dry then water it for very long time
@2001 Adriano Cruz
NCE e IM - UFRJ
No. 48
8
Fuzzifying M easure Temp = 35º Temp = 48º Age = 35 Grade = 10.0 Grade = 8.5
Fuzzy Systems
Fuzzified Measure Tem p = high, µ high (t)=0.8 Tem p = high, µ high (t)=1.0 Idade = middle, µm iddle (i)=0.8 Grade = A, µA (t)=1.0 Grade = A, µA (t)=0.87
Y=F(X) Y=F(X)
X
Function F(x) is unknown @2001 Adriano Cruz
NCE e IM - UFRJ
No. 49
Approximation of Functions Y
NCE e IM - UFRJ
@2001 Adriano Cruz
No. 50
Fuzzy Aproximation Theorem
patches
Patches are pieces of knowledge about a problem Every patch corresponds to a rule or proposition If the speed is high then step on the break
X
NCE e IM - UFRJ
No. 51
Fuzzy Aproximation Theorem An additive fuzzy system F:X->Y uniformly approximates f:X->Y if X is compact and f is continuous. Bart Kosko
@2001 Adriano Cruz
NCE e IM - UFRJ
No. 52
Fuzzy Systems Rules
No. 53
Sets
Operators
Data Management Fuzzyfier
NCE e IM - UFRJ
@2001 Adriano Cruz
@2001 Adriano Cruz
Inference Engine
NCE e IM - UFRJ
Deffuzzifier
@2001 Adriano Cruz
No. 54
9
Advantages
Disadvantages
Use rules that express imprecision of the real world. Easy to understand, test and maintain Easy to be prototyped Robust. They operate even when there is lack of rules or wrong rules. Need less rules Parallel evaluation of rules Accumulate evidences in favour and against NCE e IM - UFRJ
@2001 Adriano Cruz
No. 55
Commercial Products
NCE e IM - UFRJ
No. 57
Products
Do not learn easily
Difficult to establish correct rules
Lack of precise mathematical model
@2001 Adriano Cruz
NCE e IM - UFRJ
No. 56
Sugeno designed a voice controlled system to operate an unmanned helicopter Anti-Lock Braking Systems: Nissan, Mitsubishi. Honda, Mazda, Hyunday, BMW, Bosch and Peugeot Suspension, transmission and fuel injector systems are usual. Hitachi uses approximately 150 rules to trade in Japanese bonds and futures Yamaichi Securities uses hundreds of rules to manage a stock fund Anaesthesia Control and Fuzzy Data Analysis for Cardio-Anaesthesia NCE e IM - UFRJ No. 58 @2001 Adriano Cruz
Questions?
Air conditioning
Mitsubishi, Hitachi, Sharp
Avoids temperature oscillations and saves energy
Electronic fuel injection
NOK/Nissan
Injection based on throttle, O2 tax, water temperature, RPM, etc
Steel
Nippon Steel
Mix inputs and controls time and temperature
Golf
Maruman Golf Club
Chooses clubs
Lifts
Fujitec
Improves response time based on traffic
@2001 Adriano Cruz
Need more tests and simulation
Commercial Products
Sendai subway: 16 stations and 13,5 km route, designed by Hitachi Washing machines that measure weight, saturation time and water clarity in order to program cycles Portable camcorders with automatic focus and anti-jitter Vacuum cleaners that measure air dust to set suction power Microwave ovens that measure temperature, humidity, weight of food to set time and power.
@2001 Adriano Cruz
NCE e IM - UFRJ
No. 59
Is fuzzy logic probability ?
Find a fuzzy product description.
Find fuzzy development tools.
Fuzzy Logic is a multi values logic. Find other examples.
@2001 Adriano Cruz
NCE e IM - UFRJ
No. 60
10