Dynamic Classifier Aggregation Using Fuzzy t-conorm Integral

0427236 - ÚI 2015 RIV US eng C - Konferenční příspěvek (zahraniční konf.)
Štefka, David - Holeňa, Martin
Dynamic Classifier Aggregation Using Fuzzy t-conorm Integral.
SITIS 2011. Proceedings of the 7th International Conference on Signal Image Technology & Internet Based Systems. Los Alamitos: IEEE Computer Society, 2011 - (Yetongnon, K.; Chbeir, R.; Dipanda, A.), s. 126-133. ISBN 978-1-4673-0431-3.
[SITIS 2011. International Conference on Signal Image Technology and Internet Based Systems /7./. Dijon (FR), 28.11.2011-01.12.2011]
Grant CEP: GA MŠMT ME 949; GA ČR GA201/08/0802
Výzkumný záměr: CEZ:AV0Z10300504
Klíčová slova: fuzzy t-conorm integral * fuzzy measure * dynamic classifier combining
Kód oboru RIV: IN - Informatika

Fuzzy integral is a general aggregation operator, which encompasses many common aggregation operators like weighted mean, ordered weighted mean, weighted minimum and maximum, etc. In classifier combining, it can be used to aggregate the outputs of the individual classifiers in the team with respect to a fuzzy measure, based on the classifier confidences. In practice, the Choquet integral and the Sugeno integral are used most often. However, they both belong to the more general family of fuzzy t-conorm integral. In this paper, we theoretically examine which fuzzy t-conorm integrals are useful for classifier aggregation, and we experimentally compare the individual methods on 23 benchmark datasets.
Trvalý link: http://hdl.handle.net/11104/0232831