Title :
Object-oriented and functional concepts in artificial neural network modelling
Author :
Kock, Gerd ; Serbedzija, N.B.
Author_Institution :
GMD FIRST, Berlin, Germany
Abstract :
We propose a strategy on how to specify artificial neural networks and outline a neural network description language supporting this strategy. Unit and network types are defined in an object-oriented manner. For the specifications of topologies we use structure relations which themselves are defined using functional concepts. The underlying model of this approach is an extension of the AXON model. To demonstrate the use of our description language a complete specification of a three-layer backpropagation model is presented.
Keywords :
backpropagation; formal specification; hardware description languages; multilayer perceptrons; object-oriented methods; AXON model; artificial neural network modelling; description language; functional concepts; neural network description language; neural network specification; object-oriented methods; structure relations; three-layer backpropagation model; topology specification; Artificial neural networks; Backpropagation; Computer languages; Intelligent networks; Libraries; Logistics; Nerve fibers; Network topology; Object oriented modeling; Time of arrival estimation;
Conference_Titel :
Neural Networks, 1993. IJCNN '93-Nagoya. Proceedings of 1993 International Joint Conference on
Print_ISBN :
0-7803-1421-2
DOI :
10.1109/IJCNN.1993.714062
