|
|
vol. 2, nr. 1 (2003)
|
Methodic aspects of creating ANN models in agrophysical research
|
|
|
Zbigniew Ślipek, Sławomir Francik, Jarosław Frączek
|
|
| (get PDF )
|
|
|
Department of Machine Design, University of Agriculture, ul. Balicka 104, 30-149 Kraków |
|
|
vol. 2 (2003), nr. 1,
pp. 231-241
|
|
|
abstract:
The paper presents phases of creating a model with the use of artificial neural networks (ANN). The authors have analysed the key phases of model creation based on own research results. Recommendations that allow to ensure the maximum accuracy of the neural model performance were formulated: During creation of the SSN model it is recommended to test different net architectures as most as possible. Multiple-repeating of learning process increases probability to find a minimum global error of working of SSN in error hyperspace. Choosing the best net shall be carried out basing on values of mistakes for testing set to which belongs standards not taken into account during learning process.
|
|
keywords:
artificial neural network, modelling
|
|
original in:
Polish
|
|
|
|
|
|
