BulDML at Institute of Mathematics and Informatics >
International Journal ITA >
2007 >
Volume 14 Number 3 >

Please use this identifier to cite or link to this item:

Title: Growing Neural Networks Using Nonconventional Activation Functions
Authors: Bodyanskiy, Yevgeniy
Pliss, Iryna
Slipchenko, Oleksandr
Keywords: Ontogenic Artificial Neural Network
Orthogonal Activation Functions
Time-Series Forecasting
Issue Date: 2007
Publisher: Institute of Information Theories and Applications FOI ITHEA
Abstract: In the paper, an ontogenic artificial neural network (ANNs) is proposed. The network uses orthogonal activation functions that allow significant reducing of computational complexity. Another advantage is numerical stability, because the system of activation functions is linearly independent by definition. A learning procedure for proposed ANN with guaranteed convergence to the global minimum of error function in the parameter space is developed. An algorithm for structure network structure adaptation is proposed. The algorithm allows adding or deleting a node in real-time without retraining of the network. Simulation results confirm the efficiency of the proposed approach.
ISSN: 1313-0463
Appears in Collections:Volume 14 Number 3

Files in This Item:

File Description SizeFormat
ijita14-3-p13.pdf82.52 kBAdobe PDFView/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.


Valid XHTML 1.0!   Creative Commons License