srp Zbornik radova Instituta za šumarstvo number: 48-49 startpage: 104 endpage: 114 Institut za šumarstvo All rights reserved application/pdf 3450875 bytes Radonja, Pero info:eu-repo/semantics/article duglazija, smreka, modeliranje, predviđanje, neuronske mreže. Douglas-fir, spruce, modeling, prediction, neural networks. 2003 PREGLED POSTIGNUTIH VISINA DUGLAZIJE U SRBIJI GRUPISANIH PREMA NADMORSKIM VISINAMA ZASADA U radu je dat pregled postignutih visina duglazije (Pseudotsuga menziesii /Mirb./ Franco) u Srbiji grupisanih prema nadmorskoj visini zasada. U prvom delu rada posmatrane su desetogodišnje duglazije. U drugom delu rada u cilju grube provere dobijenih modela posmatrane su visine desetogodisnjih i petnaestogodisnjih smrča (Picea excelsa) na osnovu podataka iz literature. Pokazano je da se kod modeliranja posmatranih procesa mogu vrlo efikasno iskoristiti modeli bazirani na veštačkoj inteligenciji odnosno na neuronskim mrežama (NM). Prikazana je primena NM različitih struktura. U procesu modeliranja iskorišćene su dvoslojne NM bazirane na Levenberg-Marquardt-ovom algoritmu kao i NM sa radijalnim neuronima. In this paper, the review of the reached heights of Douglas-fir (Pseudotsuga menziesii Mirb. Franco), in Serbia grouped with regard to the altitudes of the stands is done. In the first part of the paper plant heights of ten-year Douglas-fir is analyzed. In the second part of the paper, in the aim of obtained the coarse models verification, plant heights of ten-year and fifteen-year spruce (Picea exce/sa), based on data from literature, are considered. It is shown that artificial intelligence, i.e., neural networks (NN), based models, can be used very effectively in process of modeling. The application of NN of different structure, is presented. A two layers NN based on the Levenberg-Marquardt algorithm and radial basis function NN are used in process of modeling. https://unilib.phaidrabg.rs/o:6576 cobiss:COBISS.SR-ID 27739138 ISSN: 0354-1894