Graph neural networks with a distribution of parametrized graphs
Traditionally, graph neural networks have been trained using a single observed graph. However, the observed graph represents only one possible realization. In many applications, the graph may encounter uncertainties, such as having erroneous or missing edges, as well as edge weights that provide lit...
| Main Authors: | , , , |
|---|---|
| Other Authors: | |
| Format: | Conference Paper |
| Language: | English |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/178599 https://icml.cc/Conferences/2024 |
| _version_ | 1826115105944764416 |
|---|---|
| author | Lee, See Hian Ji, Feng Xia, Kelin Tay, Wee Peng |
| author2 | School of Electrical and Electronic Engineering |
| author_facet | School of Electrical and Electronic Engineering Lee, See Hian Ji, Feng Xia, Kelin Tay, Wee Peng |
| author_sort | Lee, See Hian |
| collection | NTU |
| description | Traditionally, graph neural networks have been trained using a single observed graph. However, the observed graph represents only one possible realization. In many applications, the graph may encounter uncertainties, such as having erroneous or missing edges, as well as edge weights that provide little informative value. To address these challenges and capture additional information previously absent in the observed graph, we introduce latent variables to parameterize and generate multiple graphs. The parameters follow an unknown distribution to be estimated. We propose a formulation in terms of maximum likelihood estimation of the network parameters. Therefore, it is possible to devise an algorithm based on Expectation- Maximization (EM). Specifically, we iteratively determine the distribution of the graphs using a Markov Chain Monte Carlo (MCMC) method, incorporating the principles of PAC-Bayesian theory. Numerical experiments demonstrate improvements in performance against baseline models on node classification for both heterogeneous and homogeneous graphs. |
| first_indexed | 2024-10-01T03:49:46Z |
| format | Conference Paper |
| id | ntu-10356/178599 |
| institution | Nanyang Technological University |
| language | English |
| last_indexed | 2024-10-01T03:49:46Z |
| publishDate | 2024 |
| record_format | dspace |
| spelling | ntu-10356/1785992024-07-05T15:38:53Z Graph neural networks with a distribution of parametrized graphs Lee, See Hian Ji, Feng Xia, Kelin Tay, Wee Peng School of Electrical and Electronic Engineering School of Physical and Mathematical Sciences 41st International Conference on Machine Learning (ICML 2024) Computer and Information Science Graph neural networks Graph representation learning Traditionally, graph neural networks have been trained using a single observed graph. However, the observed graph represents only one possible realization. In many applications, the graph may encounter uncertainties, such as having erroneous or missing edges, as well as edge weights that provide little informative value. To address these challenges and capture additional information previously absent in the observed graph, we introduce latent variables to parameterize and generate multiple graphs. The parameters follow an unknown distribution to be estimated. We propose a formulation in terms of maximum likelihood estimation of the network parameters. Therefore, it is possible to devise an algorithm based on Expectation- Maximization (EM). Specifically, we iteratively determine the distribution of the graphs using a Markov Chain Monte Carlo (MCMC) method, incorporating the principles of PAC-Bayesian theory. Numerical experiments demonstrate improvements in performance against baseline models on node classification for both heterogeneous and homogeneous graphs. Info-communications Media Development Authority (IMDA) Ministry of Education (MOE) National Research Foundation (NRF) Submitted/Accepted version 2024-07-01T01:01:17Z 2024-07-01T01:01:17Z 2024 Conference Paper Lee, S. H., Ji, F., Xia, K. & Tay, W. P. (2024). Graph neural networks with a distribution of parametrized graphs. 41st International Conference on Machine Learning (ICML 2024). 2640-3498 https://hdl.handle.net/10356/178599 https://icml.cc/Conferences/2024 en MOE-T2EP20221-0003 MOE-T2EP20120-0013 MOE-T2EP20220-0002 © The Author(s). Published by ICML. All rights reserved. This article may be downloaded for personal use only. Any other use requires prior permission of the copyright holder. The Version of Record is available online at https://proceedings.mlr.press/. application/pdf |
| spellingShingle | Computer and Information Science Graph neural networks Graph representation learning Lee, See Hian Ji, Feng Xia, Kelin Tay, Wee Peng Graph neural networks with a distribution of parametrized graphs |
| title | Graph neural networks with a distribution of parametrized graphs |
| title_full | Graph neural networks with a distribution of parametrized graphs |
| title_fullStr | Graph neural networks with a distribution of parametrized graphs |
| title_full_unstemmed | Graph neural networks with a distribution of parametrized graphs |
| title_short | Graph neural networks with a distribution of parametrized graphs |
| title_sort | graph neural networks with a distribution of parametrized graphs |
| topic | Computer and Information Science Graph neural networks Graph representation learning |
| url | https://hdl.handle.net/10356/178599 https://icml.cc/Conferences/2024 |
| work_keys_str_mv | AT leeseehian graphneuralnetworkswithadistributionofparametrizedgraphs AT jifeng graphneuralnetworkswithadistributionofparametrizedgraphs AT xiakelin graphneuralnetworkswithadistributionofparametrizedgraphs AT tayweepeng graphneuralnetworkswithadistributionofparametrizedgraphs |