Please use this identifier to cite or link to this item: https://hdl.handle.net/10356/144969
Title: Graph neural network with knowledge graph
Authors: Ang, Qi Xuan
Keywords: Engineering::Computer science and engineering
Issue Date: 2020
Publisher: Nanyang Technological University
Project: SCSE19-0679
Abstract: Knowledge Graphs contain factual information about the world, and providing a structural representation of this information. However, current knowledge graphs only contains a subset of the available information in the world. Link Prediction approaches aims to uncover the unknown information through predicting new links between existing entities in a Knowledge Graph, and is a key focus in Statistical Relational Learning (SRL). Current existing approaches to link prediction includes Tensor and Neural factorization methods, representing entities with low-dimensional representations. More recently, there has been works on investigating the use of Graph Convolutional Neural Network for learning the knowledge graph embeddings. In this report, we introduced a novel deep learning architecture inspired by works of Rela- tional Graph Convolutional Network (RGCN) and Gated Graph Convolutional Network (GatedGCN) for solving link prediction tasks in Knowledge Graphs. We focus on a range of Knowledge Graphs with different scale where our model predicts the edge labels be- tween any two connecting nodes in the graph. Our approach is able to outperform the baseline models on most of the Knowledge Graphs used in our experiments, indicating the increased capability of our model through distilling important features within RGCN and GatedGCN architecture.
URI: https://hdl.handle.net/10356/144969
Fulltext Permission: restricted
Fulltext Availability: With Fulltext
Appears in Collections:SCSE Student Reports (FYP/IA/PA/PI)

Files in This Item:
File Description SizeFormat 
FYP Ang Qi Xuan_U1621610H.pdf
  Restricted Access
874.16 kBAdobe PDFView/Open

Page view(s)

171
Updated on Jan 21, 2022

Download(s) 50

63
Updated on Jan 21, 2022

Google ScholarTM

Check

Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.