Title :
On the influence of the training set data preprocessing on neural networks training
Author :
V. Nedeljkovic;M. Milosavljevic
Author_Institution :
Fac. of Electr. Eng., Belgrade Univ., Yugoslavia
fDate :
6/14/1905 12:00:00 AM
Abstract :
A training set preprocessing procedure is proposed and its influence on neural network training performance is considered. The proposed procedure has two steps: firstly, training set data are reclassified by the 1-NN method; secondly, the data obtained in the previous step are condensed. The theoretical considerations of the proposed methods as well as some examples of neural networks training based on original and preprocessed training sets are given.
Keywords :
"Data preprocessing","Neural networks","Multi-layer neural network","Feedforward neural networks","Density functional theory","Pattern recognition","Robotics and automation","Automatic control","Backpropagation","Training data"
Conference_Titel :
Pattern Recognition, 1992. Vol.II. Conference B: Pattern Recognition Methodology and Systems, Proceedings., 11th IAPR International Conference on
Print_ISBN :
0-8186-2915-0
DOI :
10.1109/ICPR.1992.201716
