Please use this identifier to cite or link to this item: http://hdl.handle.net/20.500.11790/31
Title: An Algorithm to Transform an Artificial Neural Network into its Open Equation Form and its Potential Applications
Authors: Rinke, Wolfram C. 
Keywords: Artificial neural networks;Inversion;Model based control;Open equation transformation
Issue Date: 2015
Publisher: North Atlantic University Union
Source: International Journal of Neural Networks and Advanced Applications, 2, 28-33
Journal: International Journal of Neural Networks and Advanced Applications 
Abstract: During the last decades artificial neural networks have evolved to an accepted and proven technology for modelling and function approximation. Different kinds of network architectures exist to support certain domains and applications in an efficient way. This paper assumes the traditional multilayer feedforward artificial neural network (ANN) architecture with one input layer, one or more fully interconnected hidden layers and one output layer. Each layer consists of several classic perceptron nodes using a differentiable transfer function like the logistics function. Very often it is useful to have an ANN model in an open equation form available, that allows a deeper analysis of the model and to do more complex experiments and simulations. The following paper presents an algorithm that makes it possible to transform an ANN into its open form equivalent, called process model architecture network or PMA network. It has been used as an integral part in several industrial control projects. A PMA network can be used for system simulation, scenario analyses or inverse model based control. An example application is discussed.
URI: http://hdl.handle.net/20.500.11790/31
http://www.naun.org/main/NAUN/neural/2015/a0872607-208.pdf
ISSN: 2313-0563
Rights: info:eu-repo/semantics/closedAccess
Appears in Collections:Informationstechnologie und Informationsmanagement

SFX Query Show full item record

Page view(s) 50

109
Last Week
0
Last month
5
checked on May 18, 2021

Google ScholarTM

Check


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