Please use this identifier to cite or link to this item:
Title: Benchmarking graph neural networks with OGB datasets
Authors: Muddineni, Krishnavyas
Keywords: Engineering::Computer science and engineering
Issue Date: 2021
Publisher: Nanyang Technological University
Source: Muddineni, K. (2021). Benchmarking graph neural networks with OGB datasets. Final Year Project (FYP), Nanyang Technological University, Singapore.
Abstract: Recently, graph neural networks gained increasing popularity across various domains. In the absence of a protocol, it is challenging to evaluate the performance of a Graph neural network. To address these challenges, a benchmarking infrastructure has been developed to quantify the performance. However, the datasets on which the performance of the GNNs evaluated are small to medium sized. In order to solve this issue, Open Graph Benchmark introduced real world datasets of different sizes. In this report, we benchmark the performance of various GNNs on datasets introduced by Open graph benchmark(OGB). Precisely, we benchmarked the performance of GNNs for link prediction and graph classification tasks. GNNs are still complex for evaluating on a single large graphs. We utilised a sampling technique to evaluate GNN performances on large graphs. In case of graph classification, the GNNs are able to perform well.
Fulltext Permission: restricted
Fulltext Availability: With Fulltext
Appears in Collections:SCSE Student Reports (FYP/IA/PA/PI)

Files in This Item:
File Description SizeFormat 
  Restricted Access
367.88 kBAdobe PDFView/Open

Page view(s)

Updated on Jan 22, 2022


Updated on Jan 22, 2022

Google ScholarTM


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