Number of the records: 1
New Measure of Boolean Factor Analysis Quality
- 1.
SYSNO ASEP 0359156 Document Type C - Proceedings Paper (int. conf.) R&D Document Type Conference Paper Title New Measure of Boolean Factor Analysis Quality Author(s) Frolov, A. A. (RU)
Húsek, Dušan (UIVT-O) RID, SAI, ORCID
Polyakov, P.Y. (RU)Source Title Adaptive and Natural Computing Algorithms. Part I, 1. - Heidelberg : Springer, 2011 / Dobnikar A. ; Lotrič U. ; Šter B. - ISSN 0302-9743 - ISBN 978-3-642-20281-0 Pages s. 100-109 Number of pages 10 s. Action ICANNGA'2011. International Conference /10./ Event date 14.04.2011-16.04.2011 VEvent location Ljubljana Country SI - Slovenia Event type WRD Language eng - English Country DE - Germany Keywords Boolean factor analysis ; information gain ; expectation-maximization ; associative memory ; neural network application ; Boolean matrix factorization ; bars problem ; Hopfield neural network Subject RIV IN - Informatics, Computer Science R&D Projects GAP202/10/0262 GA ČR - Czech Science Foundation (CSF) GA205/09/1079 GA ČR - Czech Science Foundation (CSF) CEZ AV0Z10300504 - UIVT-O (2005-2011) UT WOS 000302389300011 EID SCOPUS 79955094881 DOI 10.1007/978-3-642-20282-7_11 Annotation Learning of objects from complex patterns is a long-term challenge in philosophy, neuroscience, machine learning, data mining, and in statistics. There are some approaches in literature trying to solve this difficult task consisting in discovering hidden structure of high-dimensional binary data and one of them is Boolean factor analysis. However there is no expert independent measure for evaluating this method in terms of the quality of solutions obtained, when analyzing unknown data. Here we propose information gain, model-based measure of the rate of success of individual methods. This measure presupposes that observed signals arise as Boolean superposition of base signals with noise. For the case whereby a method does not provide parameters necessary for information gain calculation we introduce the procedure for their estimation. Using an extended version of the ”Bars Problem” generation of typical synthetics data for such a task, we show that our measure is sensitive to all types of data model parameters and attains its maximum, when best fit is achieved. Workplace Institute of Computer Science Contact Tereza Šírová, sirova@cs.cas.cz, Tel.: 266 053 800 Year of Publishing 2012
Number of the records: 1