TY  - JOUR
T1  - A Novel Algorithm for Designing Three Layered Artificial Neural Networks
AU - , Suman Ahmmed AU - , Khondaker Abdullah-Al-Mamun AU - , Monirul Islam 
JO  - International Journal of Soft Computing
VL  - 2
IS  - 3
SP  - 450
EP  - 458
PY  - 2007
DA  - 2001/08/19
SN  - 1816-9503
DO  - ijscomp.2007.450.458
UR  - https://makhillpublications.co/view-article.php?doi=ijscomp.2007.450.458
KW  - ANN
KW  -ensemble of ANNs
KW  -correlations
KW  -generalization
KW  -overfitting
KW  -back-propagation
AB  - Architecture determination of Artificial Neural Networks (ANNs) is an important issue for the successful application of ANNs in many practical problems. It is well known that a three layered ANN can solve any kind of linear and nonlinear problems. This study proposes a new pruning algorithm, Architecture Designing by Correlation and Sensitivity Pruning (ADCSP), to determine the three layered near optimal ANN architectures automatically. The salient features of ADCSP are that it uses correlations, apply merging operation, uses computationally inexpensive formula, maintain its generalization ability and avoid overfitting. It has been tested extensively on a number of benchmark problems in machine learning and neural networks. The experimental results show that ADCSP can determine smaller architectures with good generalization ability compared to many other works.
ER  -