Adaptif Fuzzy-Neuro Generalized Learning Vector Quantization (FNGLVQ) = Adaptive Fuzzy-Neuro Generalized Learning Vector Quantization (FNGLVQ) / Andry Sunandar
| Main Author: | Andry Sunandar, author |
|---|---|
| Format: | Masters Doctoral |
| Terbitan: |
, 2013
|
| Subjects: | |
| Online Access: |
http://lib.ui.ac.id/file?file=digital/20330250-T-Andry Sunandar.pdf |
| ctrlnum |
20330250 |
|---|---|
| fullrecord |
<?xml version="1.0"?>
<dc schemaLocation="http://www.openarchives.org/OAI/2.0/oai_dc/ http://www.openarchives.org/OAI/2.0/oai_dc.xsd"><title>Adaptif Fuzzy-Neuro Generalized Learning Vector Quantization (FNGLVQ) = Adaptive Fuzzy-Neuro Generalized Learning Vector Quantization (FNGLVQ) / Andry Sunandar</title><creator>Andry Sunandar, author</creator><type>Thesis:Masters</type><place/><publisher/><date>2013</date><description><b><ABSTRAK</b><br>
Telah dilakukan penelitian terhadap pengembangan algoritma FNGLVQ
sehingga memiliki karakteristik adaptif terhadap data input sehingga besaran
perubahan vektor referensi memiliki besaran nilai yang adaptif. Karakteristik
adaptif didapatkan dengan melakukan modifikasi terhadap perubahan update
bobot dengan melakukan penurunan fungsi keanggotaan fuzzy tidak hanya
terhadap parameter mean (yang dilakukan pada FNGLVQ awal) namun
penurunan dilakukan terhadap kedua nilai min dan max sehingga besaran
perubahan nilai min dan max akan bervariasi (tidak konstan seperti FNGLVQ)
yang tergantung dari besaran input yang digunakan.
Karakteristik ini dapat meningkatkan akurasi dalam percobaan dalam
ketiga jenis data, yakni data EKG Aritmia, data pengenalan Aroma dengan 3
campuran, serta data Sleep secara keseluruhan, namun perbedaan nilai akurasi
terbesar didapatkan dari pengujian data pengenalan aroma 3 campuran.
Pengembangan karakteristik adaptif terhadap algoritma FNGLVQ
dilakukan dengan kedua jenis fungsi keanggotaan yakni fungsi keanggotaan
segitiga dan fungsi keanggotaan PI, dan FNGLVQ adaptif dengan fungsi
keanggotaan PI sedikit lebih baik dibandingkan FNGLVQ adaptif dengan fungsi
keanggotaan segitiga
<hr>
<b>ABSTRACT</b><br>
This research has been conducted on the development of FNGLVQ
algorithms which have adaptive characteristics to the input data so that the amount
of change in the reference vector has a magnitude of adaptive value. Adaptive
characteristics are obtained by modifying the update changes the weight by doing
a fuzzy membership function derivation. This is not only performed on the
parameters of the mean (which is done at the beginning FNGLVQ) but they are
derivated to both min and max values so that the amount of change in the weight
and is continued with min and max values will vary (not constant as in the case of
FNGLVQ) which in turn depends on the amount of inputs used.
These characteristics may increase the accuracy of the experiment in all
three types of data, including data Arrhythmia ECG, data recognition Aroma with
3 mix, as well as overall Sleep data, but the biggest difference is the accuracy of
values which have obtained from the test for 3 mixed aroma data recognition.
Development of adaptive characteristics of the algorithm FNGLVQ has
been performed with both types of membership functions namely triangular
membership functions and PI membership functions, and FNGLVQ PI adaptive
membership functions has been found to be slightly better than FNGLVQ
adaptive triangular membership functions</description><subject>Fuzzy systems</subject><subject>Computational intelligence</subject><identifier>20330250</identifier><source>http://lib.ui.ac.id/file?file=digital/20330250-T-Andry Sunandar.pdf</source><recordID>20330250</recordID></dc>
|
| format |
Thesis:Masters Thesis Thesis:Doctoral |
| author |
Andry Sunandar, author |
| title |
Adaptif Fuzzy-Neuro Generalized Learning Vector Quantization (FNGLVQ) = Adaptive Fuzzy-Neuro Generalized Learning Vector Quantization (FNGLVQ) / Andry Sunandar |
| publishDate |
2013 |
| topic |
Fuzzy systems Computational intelligence |
| url |
http://lib.ui.ac.id/file?file=digital/20330250-T-Andry Sunandar.pdf |
| contents |
<b><ABSTRAK</b><br>
Telah dilakukan penelitian terhadap pengembangan algoritma FNGLVQ
sehingga memiliki karakteristik adaptif terhadap data input sehingga besaran
perubahan vektor referensi memiliki besaran nilai yang adaptif. Karakteristik
adaptif didapatkan dengan melakukan modifikasi terhadap perubahan update
bobot dengan melakukan penurunan fungsi keanggotaan fuzzy tidak hanya
terhadap parameter mean (yang dilakukan pada FNGLVQ awal) namun
penurunan dilakukan terhadap kedua nilai min dan max sehingga besaran
perubahan nilai min dan max akan bervariasi (tidak konstan seperti FNGLVQ)
yang tergantung dari besaran input yang digunakan.
Karakteristik ini dapat meningkatkan akurasi dalam percobaan dalam
ketiga jenis data, yakni data EKG Aritmia, data pengenalan Aroma dengan 3
campuran, serta data Sleep secara keseluruhan, namun perbedaan nilai akurasi
terbesar didapatkan dari pengujian data pengenalan aroma 3 campuran.
Pengembangan karakteristik adaptif terhadap algoritma FNGLVQ
dilakukan dengan kedua jenis fungsi keanggotaan yakni fungsi keanggotaan
segitiga dan fungsi keanggotaan PI, dan FNGLVQ adaptif dengan fungsi
keanggotaan PI sedikit lebih baik dibandingkan FNGLVQ adaptif dengan fungsi
keanggotaan segitiga
<hr>
<b>ABSTRACT</b><br>
This research has been conducted on the development of FNGLVQ
algorithms which have adaptive characteristics to the input data so that the amount
of change in the reference vector has a magnitude of adaptive value. Adaptive
characteristics are obtained by modifying the update changes the weight by doing
a fuzzy membership function derivation. This is not only performed on the
parameters of the mean (which is done at the beginning FNGLVQ) but they are
derivated to both min and max values so that the amount of change in the weight
and is continued with min and max values will vary (not constant as in the case of
FNGLVQ) which in turn depends on the amount of inputs used.
These characteristics may increase the accuracy of the experiment in all
three types of data, including data Arrhythmia ECG, data recognition Aroma with
3 mix, as well as overall Sleep data, but the biggest difference is the accuracy of
values which have obtained from the test for 3 mixed aroma data recognition.
Development of adaptive characteristics of the algorithm FNGLVQ has
been performed with both types of membership functions namely triangular
membership functions and PI membership functions, and FNGLVQ PI adaptive
membership functions has been found to be slightly better than FNGLVQ
adaptive triangular membership functions |
| id |
IOS18069.20330250 |
| institution |
Universitas Indonesia |
| institution_id |
51 |
| institution_type |
library:university library |
| library |
Perpustakaan Universitas Indonesia |
| library_id |
492 |
| collection |
Repository Disertasi (Open) Universitas Indonesia |
| repository_id |
18069 |
| city |
KOTA DEPOK |
| province |
JAWA BARAT |
| repoId |
IOS18069 |
| first_indexed |
2022-12-14T04:28:40Z |
| last_indexed |
2022-12-14T04:28:40Z |
| recordtype |
dc |
| merged_child_boolean |
1 |
| _version_ |
1752202458230161408 |
| score |
17.538404 |