Please use this identifier to cite or link to this item:
Title: Definición, tipologías y casos de uso de Graph Neural Networks para el aprendizaje basado en relaciones
Author: Moure Ortega, Alfonso
Tutor: Isern Alarcón, David
Others: Ventura Royo, Carles  
Keywords: deep learning
graph neural network
node classification
Issue Date: 4-Jan-2022
Publisher: Universitat Oberta de Catalunya (UOC)
Abstract: The evolution of machine learning and, in particular, of deep learning, has accelerated drastically in the last decade. However, this development has been focused on the use of datasets where each sample is independent from the rest, without hierarchy, order or any other kind of relationship. This project is focused on exploring the evolution, contexts, use cases and different methods to perform machine learning operations over graphs using graph neural networks. This work starts by reviewing the history and theoretical knowledge about neural networks, exploring different approximations, frameworks and types of problems. To do so, this work presents a classification of models based on its approach to learning. Different kinds of problems are presented that can be solved with this particular type of neural networks: node and edge classification, graph classification, predicting changes in the knowledge domain or generating new graphs based on examples. Following the review of these theoretical aspects, a collection of implementations is presented in a Python notebook using spectral and spatial approximations to graphs to classify nodes, as well as a link prediction model. Also, to be able to compare results, an implementation based on a traditional model that doesn't attend to data structure is used to classify nodes. At the end, all results are compared and a collection of conclusions is presented to address the motivation of using this type of models, showing that they can outperform those based on euclidean spaces when structured data is present.
Language: Spanish
Appears in Collections:Bachelor thesis, research projects, etc.

Files in This Item:
File Description SizeFormat 
alfonsomoureTFG0122video.mp4Presentación en formato vídeo del proyecto474.09 MBMP4View/Open
alfonsomoureTFG0122memoria.pdfMemoria del TFG1.64 MBAdobe PDFView/Open
alfonsomoureTFG0122presentación.pdfPresentación del TFG1.16 MBAdobe PDFView/Open

This item is licensed under a Creative Commons License Creative Commons