Markučič, Joško (2016) Generalizacijsko svojstvo unaprijedne neuronske mreže. = Generalisation property of feed forward neural network. Undergraduate thesis , Sveučilište u Zagrebu, Fakultet strojarstva i brodogradnje, UNSPECIFIED. Mentor: Majetić, Dubravko.
|
Text
markucic_2016_preddiplomski.pdf - Published Version Jezik dokumenta:Croatian Download (950kB) | Preview |
Abstract (Croatian)
U ovom radu prikazan je postupak učenja umjetne neuronske mreže s povratnim rasprostiranjem pogreške u zadatku prikaza generalizacijskog svojstva neuronske mreže na primjeru ponašanja nelinearnog i linearnog dinamičkog člana prvog reda. Glavna značajka umjetnih neuronskih mreža je mogućnost estimacije funkcija koje ovise o velikom broju ulaza koji nisu općenito poznati. Uz sam postupak učenja u ovom radu je prikazano i testiranje mreže koje se provodi nakon samog postupka učenja. Programska podrška kojom je implementirana opisana umjetna neuronska mreža i postupak učenja u cijelosti je izrađena u programskom paketu MATLAB. Opis, učenje i glavne značajke umjetnih neuronskih mreža prikazane su i detaljno upisane u uvodu, a zatim je u nastavku prikazan način učenja i rada zadane mreže.
Abstract
This paper shows the feed forward neural networks with error back-propagation generalisation property of nonlinear and linear first order dynamic. The main attribute of artificial neural networks is the ability to estimate or approximate functions that can depend on a large number of inputs that are generally unknown. Aside from the learning process, this paper also shows the neural network testing process which comes right after the learning stage. The program implementing the described artificial network and learning process has been entirely made in MATLAB. Description, main attributes and learning of artificial neural networks are shown and described in the introduction, after which the learning process is shown.
| Item Type: | Thesis (Undergraduate thesis) |
|---|---|
| Uncontrolled Keywords: | neuronske mreže; umjetna inteligencija; umjetni neuron; učenje neuronske mreže; testiranje neuronske mreže; težinski faktori; zamah prvog reda; zamah drugog reda; povratno rasprostiranje pogreške; generalizacijsko svojstvo; dinamički član prvog reda |
| Keywords (Croatian): | neural networks; artificial intelligence; artificial neuron; artificial network learning; artificial network testing; neural network weights; first order momentum; second order momentum; error back-propagation; generalisation property; first order dynamic term |
| Subjects: | TECHNICAL SCIENCE > Mechanical Engineering |
| Divisions: | 900 Department of Robotics and Production System Automation > 910 Chair of Engineering Automation |
| Date Deposited: | 20 Sep 2016 09:02 |
| Last Modified: | 26 Mar 2020 12:15 |
| URI: | http://repozitorij.fsb.hr/id/eprint/6422 |
Actions (login required)
 |
View Item |
Download Statistics
Download Statistics