Original title:
Lesk a bída optimálních stromů
Authors:
Savický, Petr ; Klaschka, Jan Document type: Papers Conference/Event: ROBUST'2002. Zimní škola JČMF /12./, Hejnice (CZ), 2002-01-21 / 2002-01-25
Year:
2002
Language:
cze Abstract:
Optimal classification trees have the smallest error on training data, given the number of leaves. Previous experiments suggest that the generalization properties of the optimal trees might be consistently at least as good as these of the trees grown by classical methods. The result presented in current paper demonstrate, however, that for some classification problems the optimal trees are outperformed by the classical ones.
Keywords:
generalizační vlastnosti; klasifikační stromy; Occamova břitva; optimální algoritmy Project no.: AV0Z1030915 (CEP), GA201/00/1482 (CEP) Funding provider: GA ČR Host item entry: Robust'2002. Sborník prací Dvanácté zimní školy JČMF, ISBN 80-7015-900-6
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/11104/0125079