Go to content Go to navigation Go to search

Training of the Piecewise Linear Neural Network Used for Process Control

Petr DOLEŽEL, Martin MARIŠKA, Libor HAVLÍČEK

Abstract


This paper deals with a training of special topology of feedforward artificial neural network – artificial neural network with linear saturated activation functions in hidden neurons. This network is suitable to nonlinear processes control system design. In first paragraphs, motivation of this work is described. Then, possibilities of the training are introduced and at the end of the paper, the experiment results, which prove eligibility of the approach, are presented.

Keywords


Piecewise-Linear Neural Network, Training Algorithm, Process Control, Backpropagation
Full paper

2013-11-07

commenting closed for this article

Automation of Day-to-Day Operational Processes of a Knowledge Engineer Applications Based on Ontology or Database