Extraction of Logical Rules from Data by Means of Piecewise-Linear Neural Networks

0404708 - UIVT-O 20020161 RIV DE eng C - Konferenční příspěvek (zahraniční konf.)
Holeňa, Martin
Extraction of Logical Rules from Data by Means of Piecewise-Linear Neural Networks.
Discovery Science. Berlin: Springer, 2002 - (Lange, S.; Satoh, K.; Smith, C.), s. 193-205. Lecture Notes in Computer Science, 2534. ISBN 3-540-00188-3. ISSN 0302-9743.
[International Conference on Algorithmic Learning Theory /13./, International Conference on Discovery Science /5./. Lübeck (DE), 24.11.2002-26.11.2002]
Grant CEP: GA ČR GA201/00/1489; GA AV ČR IAB2030007
Výzkumný záměr: AV0Z1030915
Klíčová slova: data mining * knowledge discovery * artificial neural networks * multilayer perceptrons * rule extraction * piecewise-linear neural networks
Kód oboru RIV: BA - Obecná matematika

The extraction of logical rules from data by means of artificial NN is receiving increasingly much attention. Paper presents a mathematically well founded approach based on piecewise-linear activation functions,which is suitable for the extraction of both kinds of rules. Basic properties of piecewise-linear NN are reviewed, an algorithm for the extraction of Boolean rules with that approach is given. A biological application in which the presented approach has been successfully employed is briefly sketched.
Trvalý link: http://hdl.handle.net/11104/0124947