armin shmilovici ben-gurion university, israel [email protected]
TRANSCRIPT
Credits for some slidesS.Natarajan, Fuzzy Decision MakingJ.-S.R. Jang, Neuro-Fuzzy and Soft
ComputingM. Casey, Lectures on Artificial
Intelligence – CS364
Fuzzy Vs. ProbabilityWalking in the desert, close to being
dehydrated, you find two sealed bottles:The first labeled water - 95% probabilityThe second labeled water - 0.95% similarityWhich one will you choose to drink from???
3
UncertaintyInformation can be incomplete inconsistent,
uncertain, or all three. Uncertainty is defined as the lack of the exact
knowledge that would enable us to reach a perfectly reliable conclusion.
Some Sources of Uncertain Knowledge:Weak implicationsImprecise language. Our natural language is
ambiguous and imprecise.Unknown data. When the data is incomplete or
missingviews of different experts.
DefinitionMany decision-making and problem-solving tasks
are too complex to be understood quantitatively, however, people succeed by using knowledge that is imprecise rather than precise.
Fuzzy set theory resembles human reasoning in its use of approximate information and uncertainty to generate decisions.
It was specifically designed to mathematically represent uncertainty and vagueness and provide formalized tools for dealing with the imprecision intrinsic to many problems.
More DefinitionsFuzzy logic is a set of mathematical principles for knowledge
representation based on degrees of membership.
Unlike two-valued Boolean logic, fuzzy logic is multi-valued. It deals with degrees of membership and degrees of truth.
Fuzzy logic uses the continuum of logical values between 0 (completely false) and 1 (completely true). Instead of just black and white, it employs the spectrum of colours, accepting that things can be partly true and partly false at the same time.
(a) Boolean Logic. (b) Multi-valued Logic.0 1 10 0.2 0.4 0.6 0.8 100 1 10
Crisp Vs Fuzzy Sets
150 210170 180 190 200160
Height, cmDegree ofMembership
Tall Men
150 210180 190 200
1.0
0.0
0.2
0.4
0.6
0.8
160
Degree ofMembership
170
1.0
0.0
0.2
0.4
0.6
0.8
Height, cm
Fuzzy Sets
Crisp Sets
The x-axis represents the universe of discourse – the range of all possible values applicable to a chosen variable. In our case, the variable is the man height. According to this representation, the universe of men’s heights consists of all tall men.
The y-axis represents the membership value of the fuzzy set. In our case, the fuzzy set of “tall men” maps height values into corresponding membership values.
A Fuzzy Set has Fuzzy Boundaries In the fuzzy theory, fuzzy set A of universe X is defined by
function µA(x) called the membership function of set A
µA(x) : X {0, 1}, where µA(x) = 1 if x is totally in A;
µA(x) = 0 if x is not in A;
0 < µA(x) < 1 if x is partly in A.
This set allows a continuum of possible choices. For any element x of universe X, membership function µA(x) equals the degree to which x is an element of set A. This degree, a value between 0 and 1, represents the degree of membership, also called membership value, of element x in set A.
Fuzzy Set RepresentationFirst, we determine the membership functions. In
our “tall men” example, we can obtain fuzzy sets of tall, short and average men.
The universe of discourse – the men’s heights – consists of three sets: short, average and tall men. As you will see, a man who is 184 cm tall is a member of the average men set with a degree of membership of 0.1, and at the same time, he is also a member of the tall men set with a degree of 0.4. (see graph on next page)
Fuzzy Set Representation
150 210170 180 190 200160
Height, cmDegree ofMembership
Tall Men
150 210180 190 200
1.0
0.0
0.2
0.4
0.6
0.8
160
Degree ofMembership
Short Average ShortTall
170
1.0
0.0
0.2
0.4
0.6
0.8
Fuzzy Sets
Crisp Sets
Short Average
Tall
Tall
Fuzzy SetsFuzzy SetsFormal definition:
A fuzzy set A in X is expressed as a set of ordered pairs:
19/04/23 11
A x x x XA {( , ( ))| }
Universe oruniverse of discourse
Fuzzy setMembership
function(MF)
A fuzzy set is totally characterized by aA fuzzy set is totally characterized by amembership function (MF).membership function (MF).
Fuzzy Sets with Discrete UniversesFuzzy Sets with Discrete UniversesFuzzy set C = “desirable city to live in”X = {SF, Boston, LA} (discrete and nonordered)C = {(SF, 0.9), (Boston, 0.8), (LA, 0.6)}
Fuzzy set A = “sensible number of children”X = {0, 1, 2, 3, 4, 5, 6} (discrete universe)A = {(0, .1), (1, .3), (2, .7), (3, 1), (4, .6), (5, .2),
(6, .1)}
19/04/23 12
Fuzzy Sets with Cont. UniversesFuzzy Sets with Cont. Universes
Fuzzy set B = “about 50 years old”X = Set of positive real numbers (continuous)B = {(x, mB(x)) | x in X}
19/04/23 13
B xx
( )
1
150
10
2
Alternative NotationAlternative NotationA fuzzy set A can be alternatively denoted as
follows:
19/04/23 14
A x xAx X
i i
i
( ) /
A x xA
X
( ) /
X is discrete
X is continuous
Note that and integral signs stand for the union of membership grades; “/” stands for a marker and does not imply division.
Fuzzy PartitionFuzzy Partition
Fuzzy partitions formed by the linguistic values “young”, “middle aged”, and “old”:
19/04/23 15
lingmf.m
Linguistic Variables and HedgesThe range of possible values of a linguistic
variable represents the universe of discourse of that variable. For example, the universe of discourse of the linguistic variable speed might have the range between 0 and 220 km/h and may include such fuzzy subsets as very slow, slow, medium, fast, and very fast.
A linguistic variable carries with it the concept of fuzzy set qualifiers, called hedges.
Hedges are terms that modify the shape of fuzzy sets. They include adverbs such as very, somewhat, quite, more or less and slightly.
Linguistic Variables and Hedges
Short
Very Tall
Short Tall
Degree ofMembership
150 210180 190 200
1.0
0.0
0.2
0.4
0.6
0.8
160 170
Height, cm
Average
TallVery Short Very Tall
Linguistic Variables and HedgesHedge Mathematical
Expression
A little
Slightly
Very
Extremely
Hedge MathematicalExpression Graphical Representation
[A ( x )]1.3
[A ( x )]1.7
[A ( x )]2
[A ( x )]3
Operations on Fuzzy Sets
Complement
0x
1
( x )
0x
1
Containment
0x
1
0x
1
A B
Not A
A
Intersection
0x
1
0x
A B
Union0
1
A BA B
0x
1
0x
1
B
A
B
A
( x )
( x )
( x )
Note: Membership FunctionsFor the sake of convenience, usually a fuzzy set is
denoted as:
A = A(xi)/xi + …………. + A(xn)/xn
where A(xi)/xi (a singleton) is a pair “grade of membership” element, that belongs to a finite universe of discourse:
A = {x1, x2, .., xn}
Basic definitions & Terminology
Support(A) = {x X | A(x) > 0}
Core(A) = {x X | A(x) = 1}
Normality: core(A) A is a normal fuzzy set
Crossover(A) = {x X | A(x) = 0.5}
- cut: A = {x X | A(x) }
Strong - cut: A’ = {x X | A(x) > }
19/04/23 21
Set-Theoretic OperationsSet-Theoretic OperationsSubset:
Complement:
Union:
Intersection:
19/04/23 22
A B A B
C A B x x x x xc A B A B ( ) max( ( ), ( )) ( ) ( )
C A B x x x x xc A B A B ( ) min( ( ), ( )) ( ) ( )
A X A x xA A ( ) ( )1
Set-Theoretic OperationsSet-Theoretic Operations
19/04/23 23
subset.m
fuzsetop.m
MF FormulationMF FormulationTriangular MF:
19/04/23 24
trimf x a b cx ab a
c xc b
( ; , , ) max min , ,
0
Trapezoidal MF: trapmf x a b c dx ab a
d xd c
( ; , , , ) max min , , ,
1 0
Generalized bell MF: gbellmf x a b cx cb
b( ; , , )
1
12
Gaussian MF: gaussmf x a b c ex c
( ; , , )
12
2
MF FormulationMF Formulation
19/04/23 25
disp_mf.m
EqualityFuzzy set A is considered equal to a fuzzy set B,
IF AND ONLY IF (iff): A(x) = B(x), xX
A = 0.3/1 + 0.5/2 + 1/3B = 0.3/1 + 0.5/2 + 1/3
therefore A = B
InclusionInclusion of one fuzzy set into another fuzzy set.
Fuzzy set A X is included in (is a subset of) another fuzzy set, B X:
A(x) B(x), xX
Consider X = {1, 2, 3} and sets A and B
A = 0.3/1 + 0.5/2 + 1/3;B = 0.5/1 + 0.55/2 + 1/3
then A is a subset of B, or A B
CardinalityCardinality of a non-fuzzy set, Z, is the number of elements in
Z. BUT the cardinality of a fuzzy set A, the so-called SIGMA COUNT, is expressed as a SUM of the values of the membership function of A, A(x):
cardA = A(x1) + A(x2) + … A(xn) = ΣA(xi), for i=1..n
Consider X = {1, 2, 3} and sets A and B
A = 0.3/1 + 0.5/2 + 1/3;B = 0.5/1 + 0.55/2 + 1/3
cardA = 1.8
cardB = 2.05
Empty Fuzzy SetA fuzzy set A is empty, IF AND ONLY IF:
A(x) = 0, xX
Consider X = {1, 2, 3} and set A
A = 0/1 + 0/2 + 0/3
then A is empty
Alpha-cutAn -cut or -level set of a fuzzy set A X is an ORDINARY
SET A X, such that:
A={A(x), xX}.
Consider X = {1, 2, 3} and set A
A = 0.3/1 + 0.5/2 + 1/3
then A0.5 = {2, 3},
A0.1 = {1, 2, 3},
A1 = {3}
Normalized Fuzzy SetsA fuzzy subset of X is called normal if there exists at least
one element xX such that A(x) = 1.
A fuzzy subset that is not normal is called subnormal.
All crisp subsets except for the null set are normal. In fuzzy set theory, the concept of nullness essentially generalises to subnormality.
The height of a fuzzy subset A is the large membership grade of an element in A
height(A) = maxx(A(x))
Fuzzy Sets Core and SupportAssume A is a fuzzy subset of X:
the support of A is the crisp subset of X consisting of all elements with membership grade:
supp(A) = {x A(x) 0 and xX}
the core of A is the crisp subset of X consisting of all elements with membership grade:
core(A) = {x A(x) = 1 and xX}
Fuzzy Set Math OperationsaA = {aA(x), xX}
Let a =0.5, and A = {0.5/a, 0.3/b, 0.2/c, 1/d}
thenAa = {0.25/a, 0.15/b, 0.1/c, 0.5/d}
Aa = {A(x)a, xX}Let a =2, and
A = {0.5/a, 0.3/b, 0.2/c, 1/d}then
Aa = {0.25/a, 0.09/b, 0.04/c, 1/d}…
Classical Decision MakingAlternative actions {drill/not drill}Alternative states of nature {oil in
ground}Relations between action/state/outcomeUtility or objective function(s)Weakness:
Assume utility is known and accurateAssume relations are known and accurateAssumptions not correct ->solution not
accurate!
35
Fuzzy LogicFuzzy LogicSimulating the process of human Simulating the process of human
reasoningreasoning framework to computing with words and framework to computing with words and
perception, using linguistics variables.perception, using linguistics variables. Deals with uncertaintiesDeals with uncertainties Dreative decision-making processDreative decision-making process
36
Fuzzy Fuzzy Decision MakingDecision MakingA model for decision making in a fuzzy
environment.Object function and constraints are
characterized as their membership functions.The intersection of fuzzy constraints and
fuzzy objection function.
37
ExampleObjective function:
x should be larger than 10Constraint:
x should be in the vicinity of 11
38
Fuzzy decision-making method
Representation of the decision problem
Fuzzy set evaluation of the decision alternatives
Selection of the optimal alternative
The method consists of three main steps:
Different classical decision-making
39
Fuzzy Decision-Solution
Example-Job SelectionChoose one of four alternative job offers
{a1,a2,a3,a4}Following criteria
High Salary (G1)Interesting Job (C1/G2)Close driving time (C2)
Salaries in Eu {40k,45k,50k,60k}Time in minutes {42, 12, 18, 4}
Job Selection – Cont.Fuzzify the goals and constraints according to your personal utility functions. For example, for a linear salary utility function (membership function) such as 37K is a 0.0 membership in “High Salary”, and 64k is a 1.0 membership in “High Salary”,then
G1=0.11/a1+0.3/a2+0.48/a3+0.8/a4C1 is fuzzy by its nature, for exampleC1=0.4/a1+0.6/a2+0.2/a3+0.2/a4For a linear driving time utility function (membership
function) such as 8 minutes is a 0.0 membership in “Close driving time ”, and 50 minutes is a 1.0 membership in “Close driving time ”,then
C2=0.1/a1+0.9/a2+0.7/a3+1.0/a4Find D by minimizing across alternativesD=0.1/a1+0.3/a2+0.2/a3+0.2/a4Choose alternative which maximize D -> a2Note that a2 is the best, but its not good (only 0.2)
42
MULTIOBJECTIVE DECISION MAKINGMULTIOBJECTIVE DECISION MAKING
Most simple decision processes are based Most simple decision processes are based on a single objective, such as minimizing on a single objective, such as minimizing cost, maximizing profit, minimizing run cost, maximizing profit, minimizing run time, and so forth. Often, however, time, and so forth. Often, however, decisions must be made in an environment decisions must be made in an environment where more than one objective function where more than one objective function governs constraints on the problem, and governs constraints on the problem, and the relative value of each of these the relative value of each of these objectives is different. objectives is different.
For example, we are designing a new For example, we are designing a new computer, and we want simultaneously to computer, and we want simultaneously to minimize the cost, maximize CPU, minimize the cost, maximize CPU, maximize Random Access Memory (RAM), maximize Random Access Memory (RAM), and maximize reliability.and maximize reliability.
43
Multi-objective Decision Making
A = {a1,a2,…,an}: set of alternatives
O = {o1,o2,…,or}: set of objectives
The degree of membership of alternative a in Oj is given below.
Decision function:
The optimum decision a*
aaa
OOOD
rOOO
r
,...,min
...
1
21
aa DAa
D
max*
44
bbii is a parameter measuring how important is a parameter measuring how important is the objective to a decision maker for a is the objective to a decision maker for a given decisiongiven decision
D = M(OD = M(O11,b,b11) ∩ M(O) ∩ M(O22,b,b22) ∩ .. ∩ M(O) ∩ .. ∩ M(Orr,b,brr))This function is represented as the This function is represented as the
intersection of r tuples of a decision intersection of r tuples of a decision measure, M(Omeasure, M(Oii,b,bii), involving the objectives ), involving the objectives and preferences,and preferences,
The classical implication satisfies the The classical implication satisfies the requirements. Of preserving the linear requirements. Of preserving the linear ordering the preference set, and at the same ordering the preference set, and at the same time relates the two quantities in a logical time relates the two quantities in a logical way where negation is also accommodated. way where negation is also accommodated.
The decision measure for a particular The decision measure for a particular alternative, a, can be replaced with a alternative, a, can be replaced with a classical implication of the form M(Oclassical implication of the form M(Oii(a(aii), b), bii) ) = b= bi i ----> O> Oii(a) = b¯(a) = b¯ii V O V Oii(a)(a)
45
Multi-objective Decision Making
Define a set of preferences {P}Parameter bi is contained on set {P}
aaaa
Hence
aaa
ObCLet
aObD
aOb
aObbaOM
bOM
bOMbOMbOMD
r
iii
CCCAa
D
ObC
iii
ii
ii
iiii
ii
rr
,...,,minmax
,
,max
,
,
,...,,
21
'
*
'
'
'
2211
46
EXAMPLE – Multi-Objective DecisionEXAMPLE – Multi-Objective DecisionA construction engineer on a construction A construction engineer on a construction
project must retain the mass of soil form project must retain the mass of soil form sliding into a building site. Alternatives : sliding into a building site. Alternatives :
MSE - mechanically stabilized embankmentMSE - mechanically stabilized embankmentConc - a mass concrete spread wallConc - a mass concrete spread wallGab -a gabion wallGab -a gabion wallFour objectives that impact his decision:Four objectives that impact his decision:Cost Cost CCMaintainability MMaintainability MStandard S Standard S Environment E Environment E A = { MSE, Conc, Gab} = {aA = { MSE, Conc, Gab} = {a11, a, a22, a, a33}}O = {C, M, S, E} = {OO = {C, M, S, E} = {O11, O, O22, O, O33, O, O44}}P = {bP = {b11, b, b22, b, b33, b, b44} [0,1]} [0,1]
47
OO11 = { 0.4/MSE + 1/Conc + 0.1/Gab } = { 0.4/MSE + 1/Conc + 0.1/Gab }
OO2 2 = { 0.7/MSE + .8/Conc + 0.4/Gab} = { 0.7/MSE + .8/Conc + 0.4/Gab}
OO33 = {0.2/MSE + .4/Conc + 1/Gab } = {0.2/MSE + .4/Conc + 1/Gab }
OO44 = {1/MSE + 0.5/Conc + 0.5/Gab } = {1/MSE + 0.5/Conc + 0.5/Gab }
Mse conc Gab
1
.5
0O1 O2 O3 O4
48
49
50
Multiple-Choice Multiple-Criteria
niAA i ,...,2,1
ktCC t ,...,2,1
51
Example – Cooling an nuclear Reactor(Step1:representation of decision problem)
Decision Alternatives A = ｛ A1,A2,A3 ｝A1 ： flooding the reactor cavity)
A2 ： depressurizing the primary system
A3 ： doing nothing
Decision Criteria C= ｛ C1,C2,C3,C4,C5 ｝
C1 ： feasibility of the strategy C2 ：effectiveness of the strategy
C3 ： possibility of adverse effects C4 ： amount
of information needs
C5 ： compatibility with existing procedures
52
Step1:representation of decision problem(Cont.)
C1C2C3C4C5
A1
A2
A3
53
Step1:representation of decision problem (Cont.)
C1 C2 C3 C4 C5
A1A2 A
3
Hierarchical structure of an example problem
54
Fuzzy set evaluation of decision alternatives
The step consist of three activities
(1)Choosing sets of the preference ratings for the importance weights of the decision
Preference ratings include Linguistic variable and triangular fuzzy number
Example ： The preference ratings for the importance weights of the decision criteria can be defined as follows:T importance= ｛ very low, low, medium, high,very high ｝
55
Fuzzy set evaluation of decision alternatives(Cont.)
Where a,b and c are real numbers and a ≦ b ≦ c
The triangular fuzzy numbers are used as membership functions corresponding to the elements in term set
56
0 0.25 0.5 0.75 1
Linguistic variable and triangular fuzzy number
1
VL L M H VH
57
Fuzzy set evaluation of decision alternatives
Choosing sets of the preference ratings for the importance weights of the decision
58
Fuzzy set evaluation of decision alternatives(Cont.)
(2) Evaluating the importance weights of the criteria and the degrees of appropriateness of the decision alternatives
(3) Aggregating the weights of the decision criteria
59
Fuzzy set evaluation of decision alternatives(Cont.)
Substituting Sit and Wt with triangular fuzzy numbers, that is, Sit = (oit, pit, qit) and Wt =(at, bt, ct), Fi is approximated as
where
For i =1,2,…,n and t = 1,2,..,k
60
Evaluating and Aggregating
0.2375
0.5375
0.8625
0.2125
0.50.7625
0.2750.5625
0.8
61
This step includes two activities Prioritization of the decision alternatives
using the aggregated assessments Choice of the decision alternative with
highest priority as the optimal
Selection of the optimal alternative
62
Selection of the optimal alternative(Cont.)
α = index, the degree of optimism of the decision-maker
0≦ α ≦ 1
Larger Fi means the higher appropriateness of the decision alternative
α
The total integral value methodLet the total integral value for triangular fuzzy number F=(a, b, c)
63
Selection of the optimal alternative
If α =0 then A3 is Best!If α =0.5 then A3 is Best!If α =1.0 then A1 is Best!
0.54375) 2)0.3875) 2)0.7) 1)
0.49375) 3)0.35625) 3)0.63125) 3)
0.55) 1)0.41875) 1)0.68125) 2)
64
65
Fuzzy Ordering- contdy1 = 5 , y2 = 2 , y1 >= y2, usually no ambiguity in the ranking, so we have CRISP ordering
Situations where issues or actions are associated with uncertainty, random or fuzzy , ranking is necessary
Uncertainty in rank is random – we can use pdf ( probability density function), use Gaussian normal function involving standard deviation and solve
Plot of pdf function is there in the previous slide
Example of uncertain ranking – x1 height of Italians and x2 is the height of Swedes
Are Swedes taller than Italians ? Assign membership function µ1 for
Swedes and membership function µ2 for Italians.
66
Fuzzy Ordering
67
Fuzzy Expert SystemsUsed to capture human procedural
knowledge in the form of linguistic variableFuzzy rules: If oven’s temperature=HIGH
then cooking time=SHORT A fuzzy expert system can chain rules
together via the process of inferencing.Many successful application, especially for
control of nonlinear systems.
Fuzzy ReasoningFuzzy ReasoningSingle rule with multiple antecedentRule: if x is A and y is B then z is CFact: x is A’ and y is B’Conclusion: z is C’
Graphic Representation:
19/04/23 69
A B T-norm
X Y
w
A’ B’ C2
Z
C’
ZX Y
A’ B’
x is A’ y is B’ z is C’