OpenHGNN: An Open Source Toolkit for Heterogeneous Graph Neural Network

Han, Hu; Zhao, Tianyu; Yang, Cheng; Hongyi, Zhang; Liu, Yaoqi; Wang, Xiao; Shi, Chuan

doi:10.1145/3511808.3557664

Cited by 18 publications

(5 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We use the public ACM and DBLP datasets in [28]. ACM and DBLP datasets are both bibliographic networks of academic publications.…”

Section: Datasetsmentioning

confidence: 99%

“…ACM. The ACM dataset from [28], also known as HGBn-ACM, is utilized in this study. It consists of four types of entities: 5959 authors (A), 3025 papers (P), 1902 terms (T), and 56 subjects (S).…”

Section: Datasetsmentioning

confidence: 99%

“…DBLP. The DBLP dataset, originating from [28], is known as HGNn-DBLP. It consists of four types of entities: 4057 authors (A), 14,328 papers (P), 7723 terms (T), and 56 venues (V).…”

Section: Datasetsmentioning

confidence: 99%

“…The data presented in this study are openly available in OpenHGNN Public at https://doi.org/10.1145/3511808.3557664 (accessed on 21 October 2022) , reference number [28].…”

Section: Data Availability Statementmentioning

confidence: 99%

See 3 more Smart Citations

Investigating Trace Equivalences in Information Networks

2023

Electronics

View full text Add to dashboard Cite

Equivalences are widely used and have achieved much success in concurrent systems. Meanwhile, information networks are ubiquitous for representing many complex systems and have similar characteristics and properties to concurrent systems such that they both can be described by graphs. In order to simplify information networks, we introduce equivalence to information networks, specifically leveraging the trace equivalence to reduce the complexity of these networks. In this paper, we first define the concept of trace and trace equivalence in information networks, drawing on the similar concept of concurrent systems. We then propose a computational method for determining whether two nodes are trace equivalent in an information network. With the help of this method, we derive trace-equivalent networks from original networks. Experiments show that we are able to reduce the number of nodes in the ACM and DBLP datasets by at most 65.21% and 46.68%, respectively. Running the PathSim algorithm on the original and derived networks, the mean error is 0.0728 in ACM and 0.0446 in DBLP. Overall, the results indicate that the derived networks have fewer nodes and edges than the original networks, yet still capture the same or similar information. By using trace equivalence, we are able to simplify information networks and improve their efficiency while preserving most of their informational content.

show abstract

“…We use the public ACM and DBLP datasets in [28]. ACM and DBLP datasets are both bibliographic networks of academic publications.…”

Section: Datasetsmentioning

confidence: 99%

Section: Datasetsmentioning

confidence: 99%

“…DBLP. The DBLP dataset, originating from [28], is known as HGNn-DBLP. It consists of four types of entities: 4057 authors (A), 14,328 papers (P), 7723 terms (T), and 56 venues (V).…”

Section: Datasetsmentioning

confidence: 99%

“…The data presented in this study are openly available in OpenHGNN Public at https://doi.org/10.1145/3511808.3557664 (accessed on 21 October 2022) , reference number [28].…”

Section: Data Availability Statementmentioning

confidence: 99%

See 2 more Smart Citations

Investigating Trace Equivalences in Information Networks

2023

Electronics

View full text Add to dashboard Cite

show abstract

“…The novel model of metapath2vec was proposed to capture the structural and semantic relationships between nodes and edges [23] . Hyperbolic spaces were introduced, and hyperbolic geometry was used to embed the nodes in the network [24] . Although these works provide a lowdimensional representation of nodes, they operate on the existing, rather than the original network.…”

Section: Introductionmentioning

confidence: 99%

BSIN: A Behavior Schema of Information Networks Based on Approximate Bisimulation

Hu,

2024

Tsinghua Sci. Technol.

View full text Add to dashboard Cite

Information networks are becoming increasingly important in practice. However, their escalating complexity is gradually impeding the efficiency of data mining. A novel network schema called the Behavior Schema of Information Networks (BSIN) is proposed to address this issue. This work defines the behavior of nodes as connected paths in BSIN, proposes a novel function distinguish behavior differences, and introduces approximate bisimulation into the acquisition of quotient sets for node types. The major highlight of BSIN is its ability to directly obtain a high-efficiency network on the basis of approximate bisimulation, rather than reducing the existing information network. It provides an effective representation of information networks, and the resulting novel network has a simple structure that more efficiently expresses semantic information than current network representations. The theoretical analysis of the connected paths between the original and the obtained networks demonstrates that errors are controllable; and semantic information is approximately retained. Case studies show that BSIN yields a simple network and is highly cost-effective.

show abstract