Neighbor-anchoring adversarial graph neural networks

Graph neural networks (GNNs) have witnessed widespread adoption due to their ability to learn superior representations for graph data. While GNNs exhibit strong discriminative power, they often fall short of learning the underlying node distribution for increased robustness. To deal with this, inspi...

Full description

Saved in:
Bibliographic Details
Main Authors: Liu, Zemin, Fang, Yuan, Liu, Yong, Zheng, Vincent Wenchen
Other Authors: School of Computer Science and Engineering
Format: Article
Language:English
Published: 2023
Subjects:
Online Access:https://hdl.handle.net/10356/172860
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
id sg-ntu-dr.10356-172860
record_format dspace
spelling sg-ntu-dr.10356-1728602023-12-27T02:25:43Z Neighbor-anchoring adversarial graph neural networks Liu, Zemin Fang, Yuan Liu, Yong Zheng, Vincent Wenchen School of Computer Science and Engineering Joint NTU-UBC Research Centre of Excellence in Active Living for the Elderly (LILY) Engineering::Computer science and engineering Generative Adversarial Network Graph Neural Network Graph neural networks (GNNs) have witnessed widespread adoption due to their ability to learn superior representations for graph data. While GNNs exhibit strong discriminative power, they often fall short of learning the underlying node distribution for increased robustness. To deal with this, inspired by generative adversarial networks (GANs), we investigate the problem of adversarial learning on graph neural networks, and propose a novel framework named NAGNN (i.e., Neighbor-anchoring Adversarial Graph Neural Networks) for graph representation learning, which trains not only a discriminator but also a generator that compete with each other. In particular, we propose a novel neighbor-anchoring strategy, where the generator produces samples with explicit features and neighborhood structures anchored on a reference real node, so that the discriminator can perform neighborhood aggregation on the fake samples to learn superior representation. The advantage of our neighbor-anchoring strategy can be demonstrated both theoretically and empirically. Furthermore, as a by-product, our generator can synthesize realistic-looking features, enabling potential applications such as automatic content summarization. Finally, we conduct extensive experiments on four public benchmark datasets, and achieve promising results under both quantitative and qualitative evaluations. Agency for Science, Technology and Research (A*STAR) This work was supported by the Agency for Science, Technology and Research (A*STAR) through AME Programmatic Funds under Grant A20H6b0151. 2023-12-27T02:25:43Z 2023-12-27T02:25:43Z 2023 Journal Article Liu, Z., Fang, Y., Liu, Y. & Zheng, V. W. (2023). Neighbor-anchoring adversarial graph neural networks. IEEE Transactions On Knowledge and Data Engineering, 35(1), 784-795. https://dx.doi.org/10.1109/TKDE.2021.3087970 1041-4347 https://hdl.handle.net/10356/172860 10.1109/TKDE.2021.3087970 2-s2.0-85145168819 1 35 784 795 en A20H6b0151 IEEE Transactions on Knowledge and Data Engineering © 2021 IEEE. All rights reserved.
institution Nanyang Technological University
building NTU Library
continent Asia
country Singapore
Singapore
content_provider NTU Library
collection DR-NTU
language English
topic Engineering::Computer science and engineering
Generative Adversarial Network
Graph Neural Network
spellingShingle Engineering::Computer science and engineering
Generative Adversarial Network
Graph Neural Network
Liu, Zemin
Fang, Yuan
Liu, Yong
Zheng, Vincent Wenchen
Neighbor-anchoring adversarial graph neural networks
description Graph neural networks (GNNs) have witnessed widespread adoption due to their ability to learn superior representations for graph data. While GNNs exhibit strong discriminative power, they often fall short of learning the underlying node distribution for increased robustness. To deal with this, inspired by generative adversarial networks (GANs), we investigate the problem of adversarial learning on graph neural networks, and propose a novel framework named NAGNN (i.e., Neighbor-anchoring Adversarial Graph Neural Networks) for graph representation learning, which trains not only a discriminator but also a generator that compete with each other. In particular, we propose a novel neighbor-anchoring strategy, where the generator produces samples with explicit features and neighborhood structures anchored on a reference real node, so that the discriminator can perform neighborhood aggregation on the fake samples to learn superior representation. The advantage of our neighbor-anchoring strategy can be demonstrated both theoretically and empirically. Furthermore, as a by-product, our generator can synthesize realistic-looking features, enabling potential applications such as automatic content summarization. Finally, we conduct extensive experiments on four public benchmark datasets, and achieve promising results under both quantitative and qualitative evaluations.
author2 School of Computer Science and Engineering
author_facet School of Computer Science and Engineering
Liu, Zemin
Fang, Yuan
Liu, Yong
Zheng, Vincent Wenchen
format Article
author Liu, Zemin
Fang, Yuan
Liu, Yong
Zheng, Vincent Wenchen
author_sort Liu, Zemin
title Neighbor-anchoring adversarial graph neural networks
title_short Neighbor-anchoring adversarial graph neural networks
title_full Neighbor-anchoring adversarial graph neural networks
title_fullStr Neighbor-anchoring adversarial graph neural networks
title_full_unstemmed Neighbor-anchoring adversarial graph neural networks
title_sort neighbor-anchoring adversarial graph neural networks
publishDate 2023
url https://hdl.handle.net/10356/172860
_version_ 1787136536229183488