tutorial on type-2 fuzzy sets and systems wcci 2016, vancouver
TRANSCRIPT
![Page 1: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/1.jpg)
Tutorial on Type-2 Fuzzy Sets and SystemsWCCI 2016, Vancouver
Jon Garibaldi, Robert John and Christian Wagner
Lab for Uncertainty in Data and Decision Making
L
U
C
I
D
http://lucidresearch.org
![Page 2: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/2.jpg)
Outline
• Type-2 Fuzzy sets – what are they?
• Interval Type-2
– Fuzzy Sets
– Inference
– Real World Example
• General Type-2
– Fuzzy Sets
– Inference
– Computational Aspects
– Control Example
– Design you own Type-2 FLSs
• Constructing Type-2 Fuzzy Sets from Data
– Building Type-2 FSs from Survey Data
– Application Example© University of Nottingham WCCI – Fuzz-IEEE 2016 2
![Page 3: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/3.jpg)
Linguistic Variable - Age
© University of Nottingham WCCI – Fuzz-IEEE 2016 3
0
1
0 20 40 60 80 Age
pensioner
old
middle-aged
young
teenagechild
infant
![Page 4: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/4.jpg)
Standard (Type-1) Fuzzy Set
© University of Nottingham WCCI – Fuzz-IEEE 2016 4
μ
![Page 5: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/5.jpg)
However …
© University of Nottingham WCCI – Fuzz-IEEE 2016 5
μ
An age of 20 has
membership of
precisely 0.200000 -
there is no
fuzziness here!
![Page 6: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/6.jpg)
Exercise – ‘Middle-Aged’?
© University of Nottingham WCCI – Fuzz-IEEE 2016 6
0 20 40 60 80 100
0.0
0.2
0.4
0.6
0.8
1.0
age
mu
Where do you put the centre of 'Middle-Aged'?
![Page 7: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/7.jpg)
Type-2 Fuzzy Sets
© University of Nottingham WCCI – Fuzz-IEEE 2016 7
μ
An age of 20 now
has a membership
of the type-2 fuzzy
set of between
0.10 and 0.55
![Page 8: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/8.jpg)
Basics on Interval Type-2
J.M. Mendel, R.I. John, F.Liu, “Interval Type-2 Fuzzy Logic Systems Made Simple”, IEEE Transactions Fuzzy Systems, 14(6):808-21, 2006
![Page 9: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/9.jpg)
Basic Definitions
© University of Nottingham WCCI – Fuzz-IEEE 2016 9
![Page 10: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/10.jpg)
Basic Definitions
© University of Nottingham WCCI – Fuzz-IEEE 2016 10
![Page 11: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/11.jpg)
Example Discrete IT2 Set
© University of Nottingham WCCI – Fuzz-IEEE 2016 11
![Page 12: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/12.jpg)
Vertical Slice
© University of Nottingham WCCI – Fuzz-IEEE 2016 12
![Page 13: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/13.jpg)
FOU
© University of Nottingham WCCI – Fuzz-IEEE 2016 13
![Page 14: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/14.jpg)
Upper and Lower MF
𝜇 𝐴 𝑥 =
© University of Nottingham WCCI – Fuzz-IEEE 2016 14
![Page 15: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/15.jpg)
Concepts IllustratedMENDEL et al.: INTERVAL TYPE-2 FUZZY LOGIC SYSTEMS MADE SIMPLE 811
Fig. 4. FOU (shaded), LMF (dashed), UMF (solid) and an embedded FS (wavy line) for IT2 FS .
Fig. 5. Example of an embedded IT2 FS associated with the T2 MF depictedin Fig. 2.
Set is the union of all the primary memberships of set in
(11), and, there are a total of . Note that acts as
the domain for .
An example of an embedded T1 FS is depicted in Fig. 4; it
is the wavy curve. Other examples of are and ,
.
Example 2: Fig. 5 depicts one of the possible
embedded IT2 FSs for the T2 MF that is depicted in Fig. 2.
Observe that the embedded T1 FS that is associated with this
embedded IT2 FS is
.
Comparing (11) and (12), we see that the embedded IT2 FS
can be represented in terms of the embedded T1 FS , as
(13)
with the understanding that this means putting a secondary
grade of 1 at all points of . We will make heavy use of this
new way to represent in the sequel.
So far we have emphasized the vertical-slice representation
(decomposition) of an IT2 FS as given in (6). Next, we provide
a different representation for such a fuzzy set that is in terms
of so-called wavy slices. This representation, which makes very
heavy use of embedded IT2 FSs (Definition 8), was first pre-
sented in [19] for an arbitrary T2 FS, and is the bedrock for the
rest of this paper. We state this result for a discrete IT2 FS.
Theorem 1 (Representation Theorem): For an IT2 FS, for
which and are discrete, is the union of all of its em-
bedded IT2 FSs, i.e.,
(14)
where
(15)
and
(16)
in which denotes the discretization levels of secondary vari-
able at each of the .
Comment 1: This theorem expresses as a union of simpler
T2 FSs, the . They are simpler because their secondary MFs
are singletons. Whereas (6) is a vertical slice representation of
, (14) is a wavy slice representation of .
Comment 2: A detailed proof of this theorem appears in [19].
Although it is important to have such a proof, we maintain that
the results in (14) are obvious using the following simple geo-
metric argument.
• The MF of an IT2 FS is three-dimensional (3-D) (e.g., Fig.
2). Each of its embedded IT2 FSs is a 3-D wavy slice (a
foil). Create all of the possible wavy slices and take their
union to reconstruct the original 3-D MF. Same points,
which occur in different wavy slices, only appear once in
the set-theoretic union.
© University of Nottingham WCCI – Fuzz-IEEE 2016 15
![Page 16: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/16.jpg)
Representing Type-2 Sets
• Vertical slice representation
– union of all vertical slices
• Horizontal slice representation
– union of all alpha planes (‘horizontal slices’)
• Wavy slice representation
– union of all embedded sets (‘wavy slices’)
• zSliced based representation
– used for general type-2: next lecture
© University of Nottingham WCCI – Fuzz-IEEE 2016 16
![Page 17: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/17.jpg)
Interval Type-2 Inference
Some material is taken from:J.M. Mendel
“Type-2 Fuzzy Sets and Systems: An Overview”Computational Intelligence Magazine
2(1):20-29, 2007
![Page 18: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/18.jpg)
Type-1 Fuzzy Inference Systems
© University of Nottingham WCCI – Fuzz-IEEE 2016 18
rulesfuzzifiercrisp
inputsdefuzzifier
crispoutputs
inferencefuzzy
input setsfuzzy
output sets
x y = f(x)
![Page 19: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/19.jpg)
Type-2 Fuzzy Inference Systems
© University of Nottingham WCCI – Fuzz-IEEE 2016 19
rulesfuzzifiercrisp
inputsdefuzzifier
crispoutputs
inferencetype-2fuzzy
input sets
type-2fuzzy
output sets
typereducer
type-1fuzzy
output sets
x[yl, yu] = f(x)
![Page 20: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/20.jpg)
Single Rule Inference
© University of Nottingham WCCI – Fuzz-IEEE 2016 20
![Page 21: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/21.jpg)
Rule Combination
© University of Nottingham WCCI – Fuzz-IEEE 2016 21
![Page 22: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/22.jpg)
Rule Combination
© University of Nottingham WCCI – Fuzz-IEEE 2016 22
![Page 23: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/23.jpg)
Type Reduction
© University of Nottingham WCCI – Fuzz-IEEE 2016 23
![Page 24: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/24.jpg)
Interval Type-2 Fuzzy Logic and Supply Chain Management
S Miller, M Gongora, JM Garibaldi, RI John
"Interval type-2 fuzzy modelling and stochastic search for real-world inventory management"
Soft Computing, 16(8), 1447-1459, 2012
![Page 25: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/25.jpg)
The Problem
• Managing the supply chain is (surprisingly still) difficult:
– There are large amounts of uncertainties
– There are conflicting objectives
– Different stake holders have different needs
© University of Nottingham WCCI – Fuzz-IEEE 2016 25
![Page 26: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/26.jpg)
Supply Chain Management (SCM)
• SCM is the management of material flows from the procurement of basic raw materials to final product delivery considering;
– information flows among whole processes of supply chains,
– material flows,
– long-term relations between customers and suppliers.
© University of Nottingham WCCI – Fuzz-IEEE 2016 26
![Page 27: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/27.jpg)
Typical Supply Chain
• A typical supply chain consists of five main components
– Customers,
– Retailers,
– Wholesalers/Distributors,
– Manufacturers,
– Component/Raw material suppliers.
© University of Nottingham WCCI – Fuzz-IEEE 2016 27
![Page 28: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/28.jpg)
Inventory Management
• Inventory management is an integrated approach to plan and control inventory while considering the whole network from suppliers to end users.
• It is essential for good inventory management:
– to avoid stock outs,
– to manage surplus stock.
• Purpose of inventory management is to find out:}
– How many units to order?
– When to order?
© University of Nottingham WCCI – Fuzz-IEEE 2016 28
![Page 29: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/29.jpg)
© University of Nottingham WCCI – Fuzz-IEEE 2016 29
![Page 30: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/30.jpg)
Stage-I: Ranking of Suppliers
• The decision makers selected the criteria relevant to the circumstance at hand from a list of criteria.
• They were evaluated in a linguistic way such as `low', `moderate', `high', `very high' to generate trapezoidal fuzzy sets for the importance of each criterion based on thoughts of the decision maker.
• The linguistic terms were converted into fuzzy weights using fuzzy membership functions.
© University of Nottingham WCCI – Fuzz-IEEE 2016 30
![Page 31: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/31.jpg)
© University of Nottingham WCCI – Fuzz-IEEE 2016 31
![Page 32: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/32.jpg)
Method Summary
• Performances were determined in the same manner of the criteria using linguistic terms such as `excellent', `very good', `good', `poor'
– the linguistic terms were converted into fuzzy weights using fuzzy membership functions
• The aggregate fuzzy scores for each supplier was calculated by aggregating all the pertinent criteria
– each of them was converted into crisp scores using centroid type-reduction and defuzzification methods
• Suppliers were ranked according to their crisp scores
• The risk values are calculated based on their scores© University of Nottingham WCCI – Fuzz-IEEE 2016 32
![Page 33: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/33.jpg)
© University of Nottingham WCCI – Fuzz-IEEE 2016 33
![Page 34: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/34.jpg)
Inventory Planning with Consideration of Supplier Risk
• Several assumptions are made to determine the form of problem
• The problem is formulated defining objectives, constraints and decision variables
• Objectives are scaled into one objective using two scalarisation method
– weighted sum and Tchebycheff approaches
• Six different scenario are generated using different weight settings
© University of Nottingham WCCI – Fuzz-IEEE 2016 34
![Page 35: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/35.jpg)
General Type-2 Fuzzy Sets
C.Wagner, H.Hagras, “Towards General Type-2
Fuzzy Logic Systems based on zSlices”, IEEE Trans Fuzzy Systems, 18(4):637-60, 2010
![Page 36: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/36.jpg)
General Type-2
© University of Nottingham WCCI – Fuzz-IEEE 2016 36
![Page 37: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/37.jpg)
Secondary View
© University of Nottingham WCCI – Fuzz-IEEE 2016 37
![Page 38: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/38.jpg)
zSlices
© University of Nottingham WCCI – Fuzz-IEEE 2016 38
![Page 39: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/39.jpg)
zSlices-based General Type-2 FSs
© University of Nottingham WCCI – Fuzz-IEEE 2016 39
![Page 40: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/40.jpg)
zSlices Illustration
© University of Nottingham WCCI – Fuzz-IEEE 2016 40
![Page 41: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/41.jpg)
From zSlices to General
© University of Nottingham WCCI – Fuzz-IEEE 2016 41
![Page 42: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/42.jpg)
zSlices based Inference
© University of Nottingham WCCI – Fuzz-IEEE 2016 42
![Page 43: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/43.jpg)
A Robotics Example
© University of Nottingham WCCI – Fuzz-IEEE 2016 43
![Page 44: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/44.jpg)
Three FLCs
• Type-1
• Interval Type-2
• zSlices based General Type-2
• All implemented using Juzzy, an open source Java library for T1, IT2, and zGT2 FLSs.
• Two inputs, one output.
• Same rule base for all:
© University of Nottingham WCCI – Fuzz-IEEE 2016 44
![Page 45: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/45.jpg)
Type-1 MFs
© University of Nottingham WCCI – Fuzz-IEEE 2016 45
Input: Front SonarNear and Far
Input: Side SonarNear and Far
Output: SteeringLeft, Straight, Right
![Page 46: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/46.jpg)
Type-1 Control Surface
© University of Nottingham WCCI – Fuzz-IEEE 2016 46
![Page 47: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/47.jpg)
Interval Type-2 MFs
© University of Nottingham WCCI – Fuzz-IEEE 2016 47
Input: Front SonarNear and Far
Input: Side SonarNear and Far
Output: SteeringLeft, Straight, Right
…
![Page 48: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/48.jpg)
Interval Type-2 Control Surface
© University of Nottingham WCCI – Fuzz-IEEE 2016 48
![Page 49: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/49.jpg)
zSlices based General Type-2 MFs
© University of Nottingham WCCI – Fuzz-IEEE 2016 49
Input: Front SonarNear and Far
Input: Side SonarNear and Far
Output: SteeringLeft, Straight, Right
…
![Page 50: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/50.jpg)
zSlices based General Type-2 MFs
© University of Nottingham WCCI – Fuzz-IEEE 2016 50
Input: Front SonarNear and Far
Input: Side SonarNear and Far
Output: Steer Left, Straight, Right
![Page 51: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/51.jpg)
zSlices based general Type-2 CS
© University of Nottingham WCCI – Fuzz-IEEE 2016 51
![Page 52: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/52.jpg)
FLCs in comparison
© University of Nottingham WCCI – Fuzz-IEEE 2016 52
Type-1 Interval Type-2 zGen. Type-2
![Page 53: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/53.jpg)
Generating Control Surfaces
© University of Nottingham WCCI – Fuzz-IEEE 2016 53
![Page 54: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/54.jpg)
Juzzy Online
An Online System for Type-1, Interval Type-2 and General-Type-2 Fuzzy Systems
![Page 55: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/55.jpg)
Different Types of Fuzzy Sets
• Type-1, interval type-2 and general type-2 fuzzy sets
• Modelling capability and complexity increase
– access, in particular to higher-order fuzzy sets and systems, is hampered by the need for programming skills and familiarity with theory
© University of Nottingham WCCI – Fuzz-IEEE 2016 55
![Page 56: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/56.jpg)
zSlice General Type-2 Fuzzy Sets
• Drastically reduced complexity & highly parallelizable
• Applications from robotic control to ambient intelligent agents, modelling of expert opinion, CWW, etc.
© University of Nottingham WCCI – Fuzz-IEEE 2016 56
Standard General Type-2 Fuzzy Set
zSlices based General Type-2 Fuzzy Set
![Page 57: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/57.jpg)
Fuzzy Logic Toolboxes
• There are a variety of toolkits available to develop Fuzzy Logic based applications, e.g.:– MATLAB® Fuzzy Logic Tool™ 2 Users’Guide, The MathWorks, Inc., Natick, USA,
March 2010. (type-1)
– J.R. Castro, O. Castillo, P. Melin, “An Interval Type-2 Fuzzy Logic Toolbox for Control Applications”, Proc. Int. Conference on Fuzzy Systems , pp. 1-6, July 2007, London, UK. (type-1, interval type-2)
– C. Wagner, S. Miller, J.M. Garibaldi, " A fuzzy toolbox for the R programming language," Proc. Int. Conference on Fuzzy Systems, pp.1185-1192, June 2011, Taipei, Taiwan. (type-1, interval type-2)
– C. Wagner, “Juzzy – A Java based Toolkit for Type-2 Fuzzy Logic”, Proc. IEEE Symp. on Advances in Type-2 Fuzzy Logic Systems, pp. 45-52, April 2013, Singapore. (type-1, interval type-2, general type-2)
– And other sources, in particular source code snippets…
• All require the installation of specific software and/or programming experience
© University of Nottingham WCCI – Fuzz-IEEE 2016 57
![Page 58: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/58.jpg)
JuzzyOnline – Browser Fuzzy Toolkit
• “A graphical, browser-based toolkit for the development of
– type-1, interval type-2 & general type-2 fuzzy systems.”
• Based on open-source Juzzy software
• Freely accessible online at: Link at end of presentation
– Completely graphical (no coding required)
– Compatible with most (all?) browsers & platforms, incl. tablets & smartphones
– Technologies: Java, Apache Tomcat, PostgreSQL, Microsoft Azure
© University of Nottingham WCCI – Fuzz-IEEE 2016 58
• Fully detailed outputs (e.g., type-reduced sets)
• Flexible figure generation (e.g., for papers!)
• Easy-sharing of FLSs via web-links (e.g., by email)
![Page 59: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/59.jpg)
Example – The tipping problem
• We would like to determine the amount of tip (as a percentage) one should give to the waiting staff based on two variables: the quality of the food and the level /quality of service provided by the member(s) of waiting staff
– Inputs
• Food: [0, 10]
• Service: [0, 10]
– Output
• Service(%): [0, 30]
• Three FLSs
– type-1, interval type-2 and zSlices based general type-2
© University of Nottingham WCCI – Fuzz-IEEE 2016 59
![Page 60: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/60.jpg)
Type-1 Fuzzy Logic System
• A link to an existing system: Type-1 FLS
• The type-1 membership functions:
© University of Nottingham WCCI – Fuzz-IEEE 2016 60
![Page 61: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/61.jpg)
Type-1 Fuzzy Logic System
• Changing, adding and removing membership functions:
© University of Nottingham WCCI – Fuzz-IEEE 2016 61
![Page 62: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/62.jpg)
Type-1 Fuzzy Logic System
• Adding, removing and editing rules:
© University of Nottingham WCCI – Fuzz-IEEE 2016 62
![Page 63: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/63.jpg)
Type-1 Fuzzy Logic System
• Inference view– here, two inputs: food=3, service =7, centroid
defuzzification
![Page 64: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/64.jpg)
Type-1 Fuzzy Logic System
• Control Surface View
– Two inputs, one output
© University of Nottingham WCCI – Fuzz-IEEE 2016 64
![Page 65: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/65.jpg)
Interval Type-2 FLS Highlights
• A link to an existing system: Interval Type-2 FLS
– The rules are identical to the type-1 FLS
• The interval type-2 membership functions:
© University of Nottingham WCCI – Fuzz-IEEE 2016 65
![Page 66: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/66.jpg)
Interval Type-2 Fuzzy Logic System
• Inference view
– here, two inputs: food=3, service =7
– Centroid type reduction
![Page 67: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/67.jpg)
General Type-2 FLS Highlights
• A link to an existing system: General Type-2 FLS
– The rules are identical to the type-1 FLS
• The zSlices based general type-2 membership functions:
– (here with 4 zSlices)
© University of Nottingham WCCI – Fuzz-IEEE 2016 67
![Page 68: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/68.jpg)
General Type-2 Fuzzy Logic System
• Inference view
– here, two inputs: food=3, service =7
– Centroid type reduction
![Page 69: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/69.jpg)
Navigation and Extra Features
© University of Nottingham WCCI – Fuzz-IEEE 2016 69
![Page 70: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/70.jpg)
Current and Future Development
• JuzzyOnline enables:
– Design & execution of type-1, interval and general type-2 FLSs
– No programming expertise or software required
– Easy sharing of FLSs
– Easy output and figure generation
• Current and future feature expansion:
– Other membership functions
– Detailed uncertainty visualisation
– Non-singleton fuzzification
• Suggestions?
© University of Nottingham WCCI – Fuzz-IEEE 2016 70
![Page 71: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/71.jpg)
Links
• http://juzzyonline.wagnerweb.net :
© University of Nottingham WCCI – Fuzz-IEEE 2016 71
![Page 72: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/72.jpg)
Constructing General Type-2 Fuzzy Sets
C Wagner, S Miller, JM Garibaldi, DT Anderson, TC Havens,
"From Interval-valued Data to General Type-2 Fuzzy Sets"
IEEE Transactions on Fuzzy Systems, 23(2), 248-269, 2015
![Page 73: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/73.jpg)
Aims
• In this work we aim to:
– create an accurate representation of interval-valued survey data
– capture the different types of uncertainty that are present
– keeping all information contained in the original data, and making no assumptions about the data
© University of Nottingham WCCI – Fuzz-IEEE 2016 73
![Page 74: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/74.jpg)
Survey Data
• The objective of our survey is to:
• Elicit opinions from a group of experts
• Study the dynamics (and variation) in decision making
– Aggregate sections of data to derive overall decisions.
• Model survey data in an expert system for use on new data
© University of Nottingham WCCI – Fuzz-IEEE 2016 74
![Page 75: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/75.jpg)
Survey Data
© University of Nottingham WCCI – Fuzz-IEEE 2016
Overall, how would you rate this eating place?
VERY
POOR
VERY
GOOD
Overall, how would you rate this eating place?
VERY
POOR
VERY
GOOD
Certain
Uncertain
To capture uncertainty, intervals are used:
The width of the ellipse denotes certainty.
75
![Page 76: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/76.jpg)
Fuzzy Sets
• We have proposed a novel fuzzy approach to interval modelling
• Fuzzyness in survey data could be:
– An individual’s certainty in their answer
– Variation in opinion between individuals (inter-expert)
– Variation in opinion of one individual over repeated surveys (intra-expert)
• In particular, we have proposed the use of zSlicesbased General Type-2 fuzzy sets
© University of Nottingham WCCI – Fuzz-IEEE 2016 76
![Page 77: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/77.jpg)
From Intervals to Fuzzy Sets
Expert A Expert B Expert C Expert D
1st Answer 40 80 40 85 20 80 25 75
2nd Answer 30 60 50 80 30 85 35 75
3rd Answer 35 70 45 95 25 75 30 70
© University of Nottingham WCCI – Fuzz-IEEE 2016
Here, 4 experts have been surveyed 3 times on the same subject.
77
![Page 78: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/78.jpg)
Viewing the intervals, it can be seen that there is some agreement.
Using the proposed method, this agreement can be used to produce a Type-1 fuzzy set.
From Intervals to Fuzzy Sets
Expert A
1st Answer 40 80
2nd Answer 30 60
3rd Answer 35 70
© University of Nottingham WCCI – Fuzz-IEEE 2016
1st
2nd
3rd
10 20 30 40 50 60 70 80 90 1000
78
![Page 79: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/79.jpg)
From Intervals to Fuzzy Sets
• To do this with the proposed method:
– Divide y into 3 sections.
– Compute 3 intervals in X for each level of agreement.
– Create a Type-1 fuzzy set using the intervals.
© University of Nottingham WCCI – Fuzz-IEEE 2016 79
![Page 80: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/80.jpg)
From Intervals to Fuzzy Sets
• If we repeat this process for all 4 experts:
• We have 4 fuzzy sets that model the intra-expert variation for each of the 4 experts.
© University of Nottingham WCCI – Fuzz-IEEE 2016 80
Expert A Expert B Expert C Expert D
![Page 81: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/81.jpg)
From Intervals to Fuzzy Sets
• Again, there is agreement between the 4 Type-1 fuzzy sets
• Using the proposed method, these sets can be used to create a General Type-2 Fuzzy set
© University of Nottingham WCCI – Fuzz-IEEE 2016 81
![Page 82: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/82.jpg)
From Intervals to Fuzzy Sets
• To do this:
– Divide Z into 4 sections
– Compute 4 Type-1 Fuzzy sets for each level of agreement
– Create a General Type-2 fuzzy set using the Type-1 fuzzy sets
© University of Nottingham WCCI – Fuzz-IEEE 2016 82
![Page 83: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/83.jpg)
From Intervals to Fuzzy Sets
• Resulting in a General Type-2 Fuzzy set that:
– Represents Inter- and Intra-expert variation in two distinct dimensions
– Makes no assumptions about the data, discarding no information
© University of Nottingham WCCI – Fuzz-IEEE 2016 83
![Page 84: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/84.jpg)
Current Work
• Current work has focused on uncertain intervals:
• In uncertain intervals, the endpoints are uncertain
• This uncertainty can be modelled with the FOU
• Step 1 results in an IT2 fuzzy set
• Step 2 results in a GT2 fuzzy set with different LMF and UMF
© University of Nottingham WCCI – Fuzz-IEEE 2016
Ā
xlĀ rĀ
B
x
:
llB
: rlB
: rrB
:lrB
:
84
![Page 85: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/85.jpg)
GT2 Application
S. Naqvi, S. Miller, J.M. Garibaldi“A General Type-II Similarity Based Model for Breast
Cancer Grading with FTIR Spectral Data”
FUZZ-IEEE 2014, Beijing, China
![Page 86: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/86.jpg)
86
Background
• Breast Cancer is one of the most frequent occurring cancers among women throughout the world
• Cancer diagnosis and prognosis
• Nottingham Prognostic Index (NPI)
• NPI= tumour diameter/5 + lymph node stage + tumour grade
– NPI categorizes patients in three prognosis groups Good, Intermediate & Poor
© University of Nottingham WCCI – Fuzz-IEEE 2016 86
![Page 87: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/87.jpg)
Nottingham Grading
System (NGS)
nuclear pleomorphismtubular formation
mitotic count
Grade
1
Less aggressive appearance of
tumour
2
Intermediateaggressive
appearance of tumour
3
Highlyaggressive
appearance of tumour
![Page 88: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/88.jpg)
Objective of Research
• FTIR (Fourier Transform Infra-red spectroscopy/microscopy) extracts a ‘molecular fingerprint’ of the sample after passing infra-red radiation though
– FTIR has been increasingly applied to the study of biomedical problems including cancer
– No two unique molecular structures produce the same infrared spectrum
• Can we characterise cancer grades using FTIR?
– Using advanced fuzzy methods to represent the high levels of uncertainty observed in this context
– Using the characterisation to automatically determine grade
© University of Nottingham WCCI – Fuzz-IEEE 2016 88
![Page 89: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/89.jpg)
Breast Tumour Tissue Microarray, Containing 40 Cases of Paired Breast Invasive Cancer
© University of Nottingham WCCI – Fuzz-IEEE 2016 89
Cancer Normal Cancer Normal
![Page 90: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/90.jpg)
Example of a 10x10 Core Section
© University of Nottingham WCCI – Fuzz-IEEE 2016 90
![Page 91: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/91.jpg)
Raw Spectra
© University of Nottingham WCCI – Fuzz-IEEE 2016 91
![Page 92: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/92.jpg)
Processed Spectra
© University of Nottingham WCCI – Fuzz-IEEE 2016 92
![Page 93: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/93.jpg)
Select five features for each grade
Features are based on absorbance values at different wave lengths
in the spectral region
Create 6 T-I fuzzy sets for each feature for each grade (G-I: 2
cases with 3 sets per case,
G-II: 6 sets from 6 cases, G-III: 6 sets from 6 cases)
Combine 6 T-I fuzzy sets into a zGT-II fuzzy set with cases
represented as z-axes (zSlices)
We have one zGT-II fuzzy set per feature per grade with 6 zSlices
Each zGT-II fuzzy set of a grade is considered as a prototype for
that grade for a particular feature
Select spectra from unseen data for any grade for each of the five
features
Create a T-I fuzzy set with 30 spectra per feature
Use weighted zsliced based similarity measure to find similarity
between created T-I fuzzy set and zSlices of prototype zGT-II
fuzzy sets for each feature for each grade.
A similarity score for zSlices for each feature for each grade is
obtained by this method
Based on the results of similarity measure, predict the grade of
unseen data by using the following methods
Maximum sum of summation of similarity scores
Majority vote
Discuss the results of similarity scores for each feature for each
grade and classification method
Describe a Grade profile for each grade on the basis of features
and similarity scores
Fig. 1. Block diagram of the model structure
location and the area covered by each region is shown in Fig.
2. For each feature, 2 absorbance values are used from each
spectrum to create an interval. Maximum peak height and
minimum absorbance values are used to create an interval.
For example, for feature 1, minimum absorbance value A
and maximum absorbance value B are combined to create
an interval (A, B ) as shown in Fig. 2. For feature 5, two
distinct peak heights have been used to create an interval.
Every spectrum used has a set of 5 interval values each for a
feature.
B. T-I Fuzzy Set Creation
For the second stage, T-I fuzzy sets have been created from
the interval data. We have initially selected 30 spectra to create
a T-I fuzzy set. As there are 30 values for each set, the primary
membership domain is divided into 30 sections ranging from
1/30 to 30/30. As we have 2 cases from G-I, 26 cases from G-
II and 6 cases of G-III, we have decided to create 6 T-I fuzzy
sets for each grade per feature from these cases. For G-I, three
regions from two cases have been selected making it 6 sets
in accordance with other grades. For G-II, we have selected
6 cases out of 26 and for G-III, spectra from all 6 cases have
been included. We have 6 sets of 30 spectra from each grade
per feature. In total we have 90 T-I fuzzy sets for all features
for all grades. The T-I fuzzy sets are created with the help of
the following Equation for interval data as described by Miller
et al [20].
Fig. 2. Regions and approximate locations of selected features
µ(A) = y1/
N[
i 1 = 1
A i 1
+ y2/
0
@N − 1[
i 1 = 1
N − 1[
i 2 = i 1 + 1
A i 1 \ A i 2
1
A
+ y3/
0
@N − 2[
i 1 = 1
N − 1[
i 2 = i 1 + 1
N[
i 3 = i 2 + 1
A i 1 \ A i 2 \ A i 3
1
A
+ . . .
+ yn /
1[
i 1 = 1
. . .
N[
i N = N
A i 1 \ . . . \ A i N
!
,
wher e yi = i / N
(4)
Where y is the degree of membership over the domain x.
It represents the number of intervals overlapping at a certain
point. An is a series of intervals where i 2 { 1....N } and N is
the number of the intervals. The T-I fuzzy set A is defined by
the membership function µ(A). In Equation 4, the ’ /’ sign
refers to degree of membership and is not a division sign
except for the last line and the addition symbol represents
the union and it is not the arithmetic addition. The T-I fuzzy
set is created by taking the union of all the intervals which are
associated with a membership of y1, the union of all possible
two tuple intersections of intervals are associated with y2 and
so on. Fig. 3 shows examples of created T-I fuzzy sets for
various features for all three grades. These sets aim to cover
the intra-case uncertainty found within spectra of a single case.
![Page 94: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/94.jpg)
Select five features for each grade
Features are based on absorbance values at different wave lengths
in the spectral region
Create 6 T-I fuzzy sets for each feature for each grade (G-I: 2
cases with 3 sets per case,
G-II: 6 sets from 6 cases, G-III: 6 sets from 6 cases)
Combine 6 T-I fuzzy sets into a zGT-II fuzzy set with cases
represented as z-axes (zSlices)
We have one zGT-II fuzzy set per feature per grade with 6 zSlices
Each zGT-II fuzzy set of a grade is considered as a prototype for
that grade for a particular feature
Select spectra from unseen data for any grade for each of the five
features
Create a T-I fuzzy set with 30 spectra per feature
Use weighted zsliced based similarity measure to find similarity
between created T-I fuzzy set and zSlices of prototype zGT-II
fuzzy sets for each feature for each grade.
A similarity score for zSlices for each feature for each grade is
obtained by this method
Based on the results of similarity measure, predict the grade of
unseen data by using the following methods
Maximum sum of summation of similarity scores
Majority vote
Discuss the results of similarity scores for each feature for each
grade and classification method
Describe a Grade profile for each grade on the basis of features
and similarity scores
Fig. 1. Block diagram of the model structure
location and the area covered by each region is shown in Fig.
2. For each feature, 2 absorbance values are used from each
spectrum to create an interval. Maximum peak height and
minimum absorbance values are used to create an interval.
For example, for feature 1, minimum absorbance value A
and maximum absorbance value B are combined to create
an interval (A, B ) as shown in Fig. 2. For feature 5, two
distinct peak heights have been used to create an interval.
Every spectrum used has a set of 5 interval values each for a
feature.
B. T-I Fuzzy Set Creation
For the second stage, T-I fuzzy sets have been created from
the interval data. We have initially selected 30 spectra to create
a T-I fuzzy set. As there are 30 values for each set, the primary
membership domain is divided into 30 sections ranging from
1/30 to 30/30. As we have 2 cases from G-I, 26 cases from G-
II and 6 cases of G-III, we have decided to create 6 T-I fuzzy
sets for each grade per feature from these cases. For G-I, three
regions from two cases have been selected making it 6 sets
in accordance with other grades. For G-II, we have selected
6 cases out of 26 and for G-III, spectra from all 6 cases have
been included. We have 6 sets of 30 spectra from each grade
per feature. In total we have 90 T-I fuzzy sets for all features
for all grades. The T-I fuzzy sets are created with the help of
the following Equation for interval data as described by Miller
et al [20].
Fig. 2. Regions and approximate locations of selected features
µ(A) = y1/
N[
i 1 = 1
A i 1
+ y2/
0
@N − 1[
i 1 = 1
N − 1[
i 2 = i 1 + 1
A i 1 \ A i 2
1
A
+ y3/
0
@N − 2[
i 1 = 1
N − 1[
i 2 = i 1 + 1
N[
i 3 = i 2 + 1
A i 1 \ A i 2 \ A i 3
1
A
+ . . .
+ yn /
1[
i 1 = 1
. . .
N[
i N = N
A i 1 \ . . . \ A i N
!
,
wher e yi = i / N
(4)
Where y is the degree of membership over the domain x.
It represents the number of intervals overlapping at a certain
point. An is a series of intervals where i 2 { 1....N } and N is
the number of the intervals. The T-I fuzzy set A is defined by
the membership function µ(A). In Equation 4, the ’ /’ sign
refers to degree of membership and is not a division sign
except for the last line and the addition symbol represents
the union and it is not the arithmetic addition. The T-I fuzzy
set is created by taking the union of all the intervals which are
associated with a membership of y1, the union of all possible
two tuple intersections of intervals are associated with y2 and
so on. Fig. 3 shows examples of created T-I fuzzy sets for
various features for all three grades. These sets aim to cover
the intra-case uncertainty found within spectra of asingle case.
![Page 95: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/95.jpg)
© University of Nottingham WCCI – Fuzz-IEEE 2016 95
![Page 96: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/96.jpg)
Similarity Measures for zGT2 Sets
• Representation of degree to which fuzzy sets are similar
• Method used is developed by McCulloch et al. (2013)– J McCulloch, C Wagner, U Aickelin, "Extending similarity measures of
interval type-2 fuzzy sets to general type-2 fuzzy sets", FUZZ-IEEE 2013, 1-8
• The method results in a similarity score between zero and one
– zero indicates completely disjoint sets and one indicates completely similar sets
© University of Nottingham WCCI – Fuzz-IEEE 2016 96
![Page 97: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/97.jpg)
Example of a Type-I Fuzzy Set
© University of Nottingham WCCI – Fuzz-IEEE 2016 97
Domain
µ
30 intervals
![Page 98: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/98.jpg)
Example zGT2 Set for a Feature
© University of Nottingham WCCI – Fuzz-IEEE 2016 98
Z Slices
µ
![Page 99: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/99.jpg)
Result Test Cases Grade 1
© University of Nottingham WCCI – Fuzz-IEEE 2016 99
Feature G-I G-II G-III
1 0.9264 0.8397 0.8235
2 0.8938 0.8035 0.7905
3 0.8122 0.8452 0.8790
4 0.6407 0.5194 0.5347
5 0.9001 0.8391 0.8719
Sum 4.1732 3.8469 3.8996
Majority vote
W L L
Feature G-I G-II G-III
1 0.9047 0.8424 0.8235
2 0.8681 0.7935 0.7905
3 0.7816 0.6838 0.7319
4 0.7653 0.7089 0.7102
5 0.9283 0.8684 0.8617
Sum 4.248 3.8970 3.9421
Majority vote
W L L
Case 1 Case 2
![Page 100: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/100.jpg)
Grade Profiles (G-I)
© University of Nottingham WCCI – Fuzz-IEEE 2016 100
Case 1 Case 2
![Page 101: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/101.jpg)
Summary of Classification Methods
© University of Nottingham WCCI – Fuzz-IEEE 2016 101
Type Cases Correct Incorrect
G-I 2 2 0
G-II 6 1 5
G-III 6 6 0
Type Cases Correct Incorrect
G-I 2 2 0
G-II 6 2 3 (1 Tie)
G-III 6 3 1 (2 Tie)
Summation
Majority Vote
![Page 102: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/102.jpg)
Conclusions
• Extracting features in terms of intervals from FTIR spectra
• A new method based on Type-II Fuzzy sets (zSlices approach) for breast cancer grade classification
• Results were appreciable for G-I and G-III
• G-II was found to be most difficult to categorise (More uncertainty between cases)
© University of Nottingham WCCI – Fuzz-IEEE 2016 102
![Page 103: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/103.jpg)
Summary and Questions
Jon Garibaldi, Robert John and Christian Wagner
Lab for Uncertainty in Data and Decision Making
L
U
C
I
D
http://lucidresearch.org
![Page 104: Tutorial on Type-2 Fuzzy Sets and Systems WCCI 2016, Vancouver](https://reader034.vdocument.in/reader034/viewer/2022042611/5881d0881a28ab71198b49a7/html5/thumbnails/104.jpg)
Lab for Uncertainty in Data and Decision Making
L
U
C
I
D
Jon Garibaldihttps://scholar.google.co.uk/citations?user=L_9C4v0AAAAJ
Robert Johnhttps://scholar.google.co.uk/citations?user=33ftCdEAAAAJ
Christian Wagnerhttps://scholar.google.co.uk/citations?user=tAQTJRIAAAAJ
http://lucidresearch.org