Please use this identifier to cite or link to this item: https://doi.org/10.1109/tkde.2019.2957786
DC FieldValue
dc.titleGraph Adversarial Training: Dynamically Regularizing Based on Graph Structure
dc.contributor.authorFENG FULI
dc.contributor.authorXIANGNAN HE
dc.contributor.authorJIE TANG
dc.contributor.authorCHUA TAT SENG
dc.date.accessioned2020-04-21T01:03:49Z
dc.date.available2020-04-21T01:03:49Z
dc.date.issued2019-12-01
dc.identifier.citationFENG FULI, XIANGNAN HE, JIE TANG, CHUA TAT SENG (2019-12-01). Graph Adversarial Training: Dynamically Regularizing Based on Graph Structure. IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING : 1. ScholarBank@NUS Repository. https://doi.org/10.1109/tkde.2019.2957786
dc.identifier.isbn10414347
dc.identifier.issn15582191
dc.identifier.urihttps://scholarbank.nus.edu.sg/handle/10635/166767
dc.description.abstractRecent efforts show that neural networks are vulnerable to small but intentional perturbations on input features in visual classification tasks. Due to the additional consideration of connections between examples (\eg articles with citation link tend to be in the same class), graph neural networks could be more sensitive to the perturbations, since the perturbations from connected examples exacerbate the impact on a target example. Adversarial Training (AT), a dynamic regularization technique, can resist the worst-case perturbations on input features and is a promising choice to improve model robustness and generalization. However, existing AT methods focus on standard classification, being less effective when training models on graph since it does not model the impact from connected examples. In this work, we explore adversarial training on graph, aiming to improve the robustness and generalization of models learned on graph. We propose Graph Adversarial Training (GraphAT), which takes the impact from connected examples into account when learning to construct and resist perturbations. We give a general formulation of GraphAT, which can be seen as a dynamic regularization scheme based on the graph structure. To demonstrate the utility of GraphAT, we employ it on a state-of-the-art graph neural network model --- Graph Convolutional Network (GCN). We conduct experiments on two citation graphs (Citeseer and Cora) and a knowledge graph (NELL), verifying the effectiveness of GraphAT which outperforms normal training on GCN by 4.51% in node classification accuracy. Codes are available via: https://github.com/fulifeng/GraphAT.
dc.description.urihttps://arxiv.org/pdf/1902.08226.pdf
dc.language.isoen
dc.publisherIEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING
dc.subjectGraph adversarial training
dc.subjectNeural Networks
dc.subjectGraph Convolutional Network
dc.typeArticle
dc.contributor.departmentDEPT OF COMPUTER SCIENCE
dc.description.doi10.1109/tkde.2019.2957786
dc.description.sourcetitleIEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING
dc.description.page1
dc.published.statePublished
dc.grant.idAISG-100E-2018-012
dc.grant.idR-252-300-002-490
dc.grant.fundingagencyIMDA
dc.grant.fundingagencyNRF
Appears in Collections:Staff Publications
Elements

Show simple item record
Files in This Item:
File Description SizeFormatAccess SettingsVersion 
Graph Adversarial Training Dynamically Regularizing Based on Graph Structure.pdf614.28 kBAdobe PDF

OPEN

Post-printView/Download

SCOPUSTM   
Citations

19
checked on Jan 30, 2023

WEB OF SCIENCETM
Citations

3
checked on Sep 22, 2021

Page view(s)

287
checked on Feb 2, 2023

Download(s)

7
checked on Feb 2, 2023

Google ScholarTM

Check

Altmetric


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.