Original title: Robustness Aspects of Knowledge Discovery
Authors: Kalina, Jan
Document type: Papers
Conference/Event: Znalosti 2013, Ostrava (CZ), 2013-10-13 / 2013-10-15
Year: 2013
Language: eng
Abstract: The sensitivity of common knowledge discovery methods to the presence of outlying measurements in the observed data is discussed as their major drawback. Our work is devoted to robust methods for information extraction from data. First, we discuss neural networks for function approximation and their sensitivity to the presence of noise and outlying measurements in the data. We propose to fit neural networks in a robust way by means of a robust nonlinear regression. Secondly, we consider information extraction from categorical data, which commonly suffers from measurement errors. To improve its robustness properties, we propose a regularized version of the common test statistics, which may find applications e.g. in pattern discovery from categorical data.
Keywords: machine learning; neural networks; outliers; robust estimation
Project no.: GA13-01930S (CEP)
Funding provider: GA ČR
Host item entry: Datakon a Znalosti 2013. Part II, ISBN 978-80-248-3189-3

Institution: Institute of Computer Science AS ČR (web)
Document availability information: Fulltext is available at the institute of the Academy of Sciences.
Original record: http://hdl.handle.net/11104/0225084

Permalink: http://www.nusl.cz/ntk/nusl-156693


The record appears in these collections:
Research > Institutes ASCR > Institute of Computer Science
Conference materials > Papers