GLOBECOM 2020 - 2020 IEEE Global Communications Conference 2020
DOI: 10.1109/globecom42002.2020.9322311
|View full text |Cite
|
Sign up to set email alerts
|

Unsupervised User Identity Linkage via Graph Neural Networks

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
3
1

Citation Types

0
4
0

Year Published

2022
2022
2024
2024

Publication Types

Select...
6
1

Relationship

0
7

Authors

Journals

citations
Cited by 11 publications
(4 citation statements)
references
References 18 publications
0
4
0
Order By: Relevance
“…RW-based methods deal with heterogeneous graphs by defining metapaths [140] which specifies the type of each node in the walk, such as (user, item, user), so that each random walk sampled (also called "instance") with the same metapath can be projected to the same vector space. Examples on the dynamic graph are THINE [141] and HDGNN [142] which sample instances and encode them through the attention layer and bidirectional RNN layers, respectively. FIGURE 13: Handling heterogeneous nodes.…”
Section: H Encoding Heterogeneous Graphsmentioning
confidence: 99%
“…RW-based methods deal with heterogeneous graphs by defining metapaths [140] which specifies the type of each node in the walk, such as (user, item, user), so that each random walk sampled (also called "instance") with the same metapath can be projected to the same vector space. Examples on the dynamic graph are THINE [141] and HDGNN [142] which sample instances and encode them through the attention layer and bidirectional RNN layers, respectively. FIGURE 13: Handling heterogeneous nodes.…”
Section: H Encoding Heterogeneous Graphsmentioning
confidence: 99%
“…The authors of [36] proposed a novel model for the UUIL problem, which minimized the Earth Mover's Distance through two optimization methods: a generative adversarial network and an orthogonal matrix transformation model. The authors of [37] proposed a novel unsupervised network alignment named NWUIL that used a Gaussian distribution to embed each vertex in the network and to preserve both the network topological information and the uncertainties of vertex representation. The authors of [22] proposed a fully unsupervised network alignment framework named GAlign.…”
Section: Related Workmentioning
confidence: 99%
“…However, collecting sufficient aligned user pairs as annotations is inevitable. Some unsupervised methods are capable of addressing the UA problem without any labeled data [13,15,[30][31][32][33][34][35][36]. Liu et al [30] proposed an improved n-gram model that can automatically generate training data according to an evaluation of the username rareness.…”
Section: Related Workmentioning
confidence: 99%
“…A user identification algorithm FRUI-P was reported by Zhou et al [35], which introduced the concept of "friend relationships" without prior knowledge for solving the UA problem. Zhou et al [36] proposed capturing node distribution in Wasserstein space and reformulating the UA task as an optimal network transport problem in a fully unsupervised manner. Recently, Li et al [37] studied user's check-in records, and jointly considered user's spatial-temporal information (e.g., location and time) to link identical users without any annotations.…”
Section: Related Workmentioning
confidence: 99%