Neuronska mreža u učenju kvadriranja brojeva

Špoljarić, Andreja (2015) Neuronska mreža u učenju kvadriranja brojeva. = Neural network in squaring numbers. Undergraduate thesis , Sveučilište u Zagrebu, Fakultet strojarstva i brodogradnje, UNSPECIFIED. Mentor: Majetić, Dubravko.

Preview

Text Špoljarić_2015_zavrsni_preddiplomski.pdf - Published Version Jezik dokumenta:Croatian Download (2MB) | Preview

Abstract (Croatian)

U ovom radu prikazan je postupak učenja neuronske mreže s povratnim rasprostiranjem pogreške u zadatku prikaza generalizacijskog svojstva neuronske mreže za primjer kvadriranja prirodnih brojeva. Općenito gledano neuronske mreže spadaju u jedno mnogo šire područje nazvano umjetna inteligencija. Glavna karakteristika umjetnih neuronskih mreža je u tome što se one ne oslanjaju isključivo na determinsitičke matematičke postupke zbog čega svi ulazi u sustav ne moraju biti u potpunosti točni da bi on mogao pravilno raditi. Osim samog postupka učenja u ovom je radu prikazano i testiranje mreže koje se provodi nakon samog procesa učenja, a cijela programska podrška, uključujući i prikladno grafičko sučelje, načinjeni su pomoću programskog paketa MATLAB. Opis, učenje i glavne karakteristike umjetnih neuronskih mreža prikazane su i detaljno opisane u uvodu, a zatim u nastavku slijedi prikaz i način učenja i rada zadane mreže.

Abstract

This paper deals with the functioning of artificial neural networks with error-back propagation whose goal is to square numbers. Generally neural networks belong to much wider area that is called artificial intelligence. The main characteristic of artificial neural networks is that they do not rely solely on deterministic mathematical procedures for which all inputs into the system may not be completely accurate so that it could work properly. In addition to the learning process in this paper, showing and testing the network carried out after the learning process, and the entire program support, including appropriate graphical interface, are made using the software package MATLAB. Description, teaching and main characteristics of artificial neural networks are shown and described in detail in the introduction. After that a presentation of the learning mode and the default operation network is given.

Item Type:	Thesis (Undergraduate thesis)
Uncontrolled Keywords:	neuronske mreže; umjetna inteligencija; umjetni neuron; kvadriranje; učenje neuronske mreže; testiranje neuronske mreže; težinski faktori; zamah prvog reda; zamah drugog reda; povratno rasprostiranje pogreške
Keywords (Croatian):	neural networks; artificial intelligence; artificial neuron; squaring; neural network learning; neural network testing; weighting factors; first order momentum; second order momentum; regenerative propagation errors
Subjects:	TECHNICAL SCIENCE > Mechanical Engineering
Divisions:	900 Department of Robotics and Production System Automation > 910 Chair of Engineering Automation
Date Deposited:	02 Mar 2015 11:01
Last Modified:	10 Apr 2020 16:26
URI:	http://repozitorij.fsb.hr/id/eprint/3311

Actions (login required)

View Item