Artificial Neural Network - Mjölnir

Amos Marcel Iversen & Rune Bæk Jensen

Studenteropgave: Fagmodulprojekt

Abstrakt

This project seeks to gain first-hand experience in applying the fundamentals of artificial neural networks
(ANN). To achieve this, a Java application has been developed using object-oriented programming and
model-view-controller as the software architecture.
The project initially elaborates the basics of ANNs, before converting these into application requirements
and a class diagram, with the purpose of serving as draft. Based on this, the application is then developed
and explained, before subsequently being tested and evaluated. The ANN functionalities have been tested
on the Mnist data set and show promising results for simple pattern recognition with accuracy reaching
upwards of 90%. For a more complete evaluation of the application, further testing on e.g. other data
types is imperative. Although the application is functional, some planned features are only partly- or yet
to be fully implemented. Still, the application is scalable and flexible however for future development a
more managed objective is vital, in order to assess whether to continue further progress on the application
as is, or a complete rework is more beneficial

UddannelserDatalogi, (Bachelor/kandidatuddannelse) Bachelor
SprogEngelsk
Udgivelsesdato18 dec. 2018
Antal sider90
VejledereHenning Christiansen

Emneord

  • AI
  • Artificial neural network
  • Datalogi
  • Computer science