eng application/pdf 6276182 bytes info:eu-repo/semantics/article Radonja, Pero Beograd: Institut za šumarstvo Key words: stem profile function, neural networks, standard error of modeling, stem volume. Ključne reči: profilna funkcija debla, neuronske mreže, standardna greška modeliranja, zapremina debla. Abstract: Stem profile modelling by neural networks, NN, by method based on the artificial intelligence is presented in this paper. The programs newff, train and sim, which are part of the program package MATLAB were used. The method of obtaining stem profile models or stem profile functions based on the use of NN, was illustrated with several examples. Since the modelling is based on the use of modified Brink’s function, MBF, proved to be most suitable method until now, we compared these two methods. The size of the standard error of modelling was the base of the comparison. The data from the enclosed tables show the superiority of the modelling method based on NN. Izvod: U radu je prikazano modeliranje profilne funkcije debla pomoću neuronskih mreža, NM, odnosno pomoću postupka baziranog na veštačkoj inteligenciji. Iskorišćeni su programi koji se nalaze u programskom paketu MATLAB. Postupak dobijanja modela profilne funkcije na osnovu korišćenja NM, ilustrovan je sa više primera. S obzirom da se do sada kao najpogodniji postupak pokazalo modeliranje bazirano na primeni modifikovane Brinkove funkcije, MBF, uporedili smo ova dva postupka. Veličina standardne greške modeliranja bila je osnov poređenja. Podaci iz priloženih tabela jasno pokazuju superiornost postupka modeliranja baziranog na NM. All rights reserved Sustainable forestry volume: 57-58 startpage: 82 endpage: 92 2008 https://unilib.phaidrabg.rs/o:9455 ISSN: 1821-1046 Stem profile modeling using neural networks Modeliranje profilne funkcije debla pomoću neuronskih mreža