Please use this identifier to cite or link to this item: http://hdl.handle.net/10609/107666
Title: Xarxa neuronal amb topologia evolutiva per resoldre el problema del camí més curt d'un graf
Author: Vilaseca Giralt, Joan Antoni
Tutor: Isern, David  
Others: Ventura, Carles  
Abstract: This TFG has performed a study of the functioning of neuro-evolutionary neural networks and the properties and characteristics of the NEAT method, which basically consists in the evolution of the topology from a minimum structure and the maintenance of a historical record through a number of innovations in order to preserve the innovations by classifying the networks into different classes called species. The goal was to create a Java implementation of this method that would allow us to study its details and apply it to the problem of finding the shortest path in a graph. That is why the alternatives to solve the problem through a neural network have been studied and it has been decided to implement a new algorithm that is based on applying the Neat network recursively and in this way go through the graph. The project could be of interest in areas where dynamically varying graphs are used, such as Internet packet routing or stand-alone driving. The implementation has used the Java language and to adapt it to the problem of the shortest path the original methods have been modified to use it recursively and add some problem-specific help (no return on the same path, cycle control) to improve results. We have found that this implementation is able to solve the problem on small graphs, while on larger graphs or with higher grade on its vertices it is necessary to improve their behavior.
Keywords: neuroevolution
speciation
especiación
shortest path problem
Document type: info:eu-repo/semantics/bachelorThesis
Issue Date: Dec-2019
Publication license: http://creativecommons.org/licenses/by-nc-nd/3.0/es/  
Appears in Collections:Bachelor thesis, research projects, etc.

Files in This Item:
File Description SizeFormat 
jvilasecagTFG1219memòria.pdfMemòria del TFG2,01 MBAdobe PDFThumbnail
View/Open
repositoriCodiGitHub.txt39 BTextView/Open