Path: Top <-> Seminar Internasional <-> IRECOMS 2016 <-> 3. technology, health, nutrition

TESTING DESIGN OF NEURAL NETWORK PARAMETERS IN OPTIMIZATION TRAINING ALGORITHM

Proceeding from JHPTUMP / 2017-03-08 19:52:03
Oleh : Hindayati Mustafidah and Suwarsito, Universitas Muhammadiyah Purwokerto (h.mustafidah@ump.ac.id)
Dibuat : 2017-01-12, dengan 3 file

Keyword : testing, optimal training algorithm, network parameter, backpropagation

Various fields of human life, either for research or solving technical issues such as forecasting, diagnostics, and pattern recognition has many applications that use Artificial Neural Network (ANN), especially backpropagation method. In this ANN method, there are the most important part that determines its performance is a training algorithm that is used. There are 12 training algorithms that can be used in the ANN method of backpropagation. This algorithm performance is affected by network parameters including the number of neurons in the input layer, the maximum allowable epoch, the learning rate, and the target error are used. Each parameter has variation value that would affect the output of the network. In this research, each training algorithm will be tested against variations in the number of neurons in the hidden layer to get the most optimal one. This research was conducted using mixed method, that is computer software development research with qualitative descirptive testing (using ANOVA statistical test). The final results will be obtained a design of testing for obtaining ANN parameter models with the most optimal training algorithm that can later be used as the basis for the application of ANN in solving the problem according to their characteristics.

Beri Komentar ?#(92) | Bookmark

PropertiNilai Properti
ID PublisherJHPTUMP
OrganisasiUniversitas Muhammadiyah Purwokerto
Nama KontakLembaga Publikasi Ilmiah dan Penerbitan
AlamatJl. Dukuhwaluh
KotaPurwokerto
DaerahJawa Tengah
NegaraIndonesia
Telepon0281-636751, 630463
Fax0281-637239
E-mail Administratordigilib@ump.ac.id
E-mail CKOdigilib@ump.ac.id

Print ...

Kontributor...

  • , Editor: digilib@ump.ac.id

Download...