Generalization of Hierarchical Crisp Clustering Algorithms to Fuzzy - - PowerPoint PPT Presentation

▶
generalization of hierarchical crisp clustering
SMART_READER_LITE
LIVE PREVIEW

Generalization of Hierarchical Crisp Clustering Algorithms to Fuzzy - - PowerPoint PPT Presentation

Dec 19, 2023 378 likes •595 views

Generalization of Hierarchical Crisp Clustering Algorithms to Fuzzy Logic Mathias Bank Dr. Friedhelm Schwenker mathias.bank@uni-ulm.de friedhelm.schwenker@uni-ulm.de Faculty for Mathematics and Economics Institute of Neural Information

slide-1
SLIDE 1

Generalization of Hierarchical Crisp Clustering Algorithms to Fuzzy Logic

Mathias Bank mathias.bank@uni-ulm.de Faculty for Mathematics and Economics University of Ulm

  • Dr. Friedhelm Schwenker

friedhelm.schwenker@uni-ulm.de Institute of Neural Information Processing University of Ulm

slide-2
SLIDE 2

Generalization of Hierarchical Crisp Clustering Algorithms to Fuzzy Logic | 21.07.2010 Page 2

Fuzzy Logic

  • !

"#$

% % &&% '% (%

slide-3
SLIDE 3

Generalization of Hierarchical Crisp Clustering Algorithms to Fuzzy Logic | 21.07.2010 Page 3

Motivation for Hierarchical Fuzzy Clustering )(# * +# ,&&

slide-4
SLIDE 4

Generalization of Hierarchical Crisp Clustering Algorithms to Fuzzy Logic | 21.07.2010 Page 4

Precondition for different Clustering Approach

+

*#$

!-./ +01# +

*#,&&2 $

!-./ +01

  • #12

(

slide-5
SLIDE 5

Generalization of Hierarchical Crisp Clustering Algorithms to Fuzzy Logic | 21.07.2010 Page 5

Generalized agglomerative Clustering Algorithm

+3 +4 +5 +6 + +7 +3 3 .43/ .53/ .63/ .3/ .73/ +4 3 .54/ .64/ .4/ .74/ +5 3 .65/ .5/ .75/ +6 3 .6/ .76/ + 3 .7/ +7 3

  • !$
slide-6
SLIDE 6

Generalization of Hierarchical Crisp Clustering Algorithms to Fuzzy Logic | 21.07.2010 Page 6

Generalized agglomerative Clustering Algorithm

+3 +4 +5 +6 + +7 +3 3 .43/ .53/ .63/ .3/ .73/ +4 3 .54/ .64/ .4/ .74/ +5 3 .65/ .5/ .75/ +6 3 .6/ .76/ + 3 .7/ +7 3

  • !$
  • +#
slide-7
SLIDE 7

Generalization of Hierarchical Crisp Clustering Algorithms to Fuzzy Logic | 21.07.2010 Page 7

Generalized agglomerative Clustering Algorithm

+3 +4 +5 +6 + +7 +3 3

  • .53/ .63/ .3/ .73/

+4 3 .54/ .64/ .4/ .74/ +5 3 .65/ .5/ .75/ +6 3 .6/ .76/ + 3 .7/ +7 3

  • !$
  • +#
  • +#
slide-8
SLIDE 8

Generalization of Hierarchical Crisp Clustering Algorithms to Fuzzy Logic | 21.07.2010 Page 8

Generalized agglomerative Clustering Algorithm

+3 +4 +5 +6 + +7 $3 +3 3

  • .53/ .63/ .3/ .73/
  • +4

3 .54/ .64/ .4/ .74/

  • +5

3 .65/ .5/ .75/ .$35/ +6 3 .6/ .76/ .$36/ + 3 .7/ .$3/ +7 3 .$37/ $3 3

  • !$
  • +#
  • +#
  • 8
slide-9
SLIDE 9

Generalization of Hierarchical Crisp Clustering Algorithms to Fuzzy Logic | 21.07.2010 Page 9

Linkage metrics

$9+ $9$

0$0:

  • ;8

%

0$0:

  • ;8

%

  • *

"

slide-10
SLIDE 10

Generalization of Hierarchical Crisp Clustering Algorithms to Fuzzy Logic | 21.07.2010 Page 10

Linkage metrics

$9+ $9$

0$0:

  • ;8

%

0$0:

  • ;8

%

  • *

"

slide-11
SLIDE 11

Generalization of Hierarchical Crisp Clustering Algorithms to Fuzzy Logic | 21.07.2010 Page 11

Linkage metrics

  • $9+

$9$

0$0:

  • ;8

%

0$0:

  • ;8

%

  • *
slide-12
SLIDE 12

Generalization of Hierarchical Crisp Clustering Algorithms to Fuzzy Logic | 21.07.2010 Page 12

Fuzzyness specification

+3 +4 +5 +6 + +7 +3 3

  • <

.53/ < .63/ < .3/ < .73/ +4 3 < .54/ < .64/ < .4/ < .74/ +5 3 .65/ .5/ .75/ +6 3 .6/ .76/ + 3 .7/ +7 3

  • ,&&&&

(= +# >

.##/

$#

slide-13
SLIDE 13

Generalization of Hierarchical Crisp Clustering Algorithms to Fuzzy Logic | 21.07.2010 Page 13

Generalized agglomerative Clustering Algorithm

  • !$
  • +#
  • +#
  • (&&#
  • 8

.?.18// ;

slide-14
SLIDE 14

Generalization of Hierarchical Crisp Clustering Algorithms to Fuzzy Logic | 21.07.2010 Page 14

Fuzzy Membership Calculation

3 3 4 4 ( @ $ (@ @$

  • !"!#

$ %

!&'(!"

!"!#

$ %

!&'(!"

  • p: Parents
slide-15
SLIDE 15

) * &) *

slide-16
SLIDE 16

Generalization of Hierarchical Crisp Clustering Algorithms to Fuzzy Logic | 21.07.2010 Page 16

Generalized agglomerative Clustering Algorithm

  • !$
  • +#
  • +#
  • (&&#
  • 8

.?.18// ; +,- . $,&&1+,- . ) #&)#&

slide-17
SLIDE 17

Generalization of Hierarchical Crisp Clustering Algorithms to Fuzzy Logic | 21.07.2010 Page 17

Evaluation measure for external labeled data

(@ (@ ($ ($ + 4(4$ 4(4@ 6(4@4$ 6(4@4$3+

#

$-2'28

  • #

#

,&&

A&& ,&&

+

B @ A$C3.473455D336E/ A$C42.377635367337D5 / /0 0 1 023 ,&&/0 0 1 023 # ) * *#) * '

slide-18
SLIDE 18

Generalization of Hierarchical Crisp Clustering Algorithms to Fuzzy Logic | 21.07.2010 Page 18

Evaluation Results – COS-Quality for document parents

slide-19
SLIDE 19

Generalization of Hierarchical Crisp Clustering Algorithms to Fuzzy Logic | 21.07.2010 Page 19

Evaluation Results – Fuzziness per document

slide-20
SLIDE 20

Generalization of Hierarchical Crisp Clustering Algorithms to Fuzzy Logic | 21.07.2010 Page 20

Conclusion

(#$$&,&&0 +&& ! ,&&F 4G2G8

(#,&&$(

)!#.H&&/

'A

slide-21
SLIDE 21

Generalization of Hierarchical Crisp Clustering Algorithms to Fuzzy Logic | 21.07.2010 Page 21

Discussion