Article
Computer Science, Artificial Intelligence
Thanh Le, Nam Le, Bac Le
Summary: Knowledge graphs play an important role in intelligent information systems, and link prediction is a crucial technique to address the issue of missing connections between entities in knowledge graphs. This paper proposes the ConvRot model, which integrates a 2D convolution to capture local interactions among entities and relations while preserving intuitiveness. The proposed method achieves significant improvements on standard benchmark datasets.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Computer Science, Artificial Intelligence
Guillaume Salha-Galvan, Johannes F. Lutzeyer, George Dasoulas, Romain Hennequin, Michalis Vazirgiannis
Summary: This paper introduces a high-accuracy community-preserving message passing scheme and corresponding training and optimization strategies for improving both link prediction and community detection tasks. The effectiveness of the proposed method is validated through experiments.
Article
Computer Science, Artificial Intelligence
Lei Cai, Jundong Li, Jie Wang, Shuiwang Ji
Summary: This research focuses on the graph link prediction task and proposes a novel approach that uses deep learning to extract features from subgraphs. By utilizing line graphs, the authors solve the link prediction problem as a node classification task instead of a graph classification task, resulting in improved performance and efficiency.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2022)
Article
Computer Science, Information Systems
Jianxing Zheng, Zifeng Qin, Suge Wang, Deyu Li
Summary: This paper presents an explainable method for friend link recommendation in recommender systems. It leverages the similarity of user pairs using fusion embedding and incorporates external knowledge semantics. The proposed method considers both direct and indirect factors in predicting friend links and provides a good interpretation of the recommendation results.
INFORMATION SCIENCES
(2022)
Article
Computer Science, Artificial Intelligence
Yifan Lu, Mengzhou Gao, Huan Liu, Zehao Liu, Wei Yu, Xiaoming Li, Pengfei Jiao
Summary: NOH is a neural network model that utilizes structural information and estimates overlapped neighborhood from a heterogeneous graph for link prediction. It addresses the limitations of existing models that only consider low-order pairwise relations and node attributes, neglecting higher-order group interactions. Experimental results on four real-world datasets demonstrate that NOH consistently achieves state-of-the-art performance on link prediction.
PATTERN RECOGNITION
(2023)
Article
Computer Science, Interdisciplinary Applications
Yiyuan Pu, Daniel Beck, Karin Verspoor
Summary: This study explores the framing of literature-based discovery (LBD) as link prediction and graph embedding learning in the context of Alzheimer's Disease (AD). A four-stage approach is proposed to create and analyze an AD-specific knowledge graph and predict new knowledge based on time-sliced link prediction. The results show that neural network graph-embedding link prediction methods have promise for LBD, but the prediction setting is extremely challenging.
JOURNAL OF BIOMEDICAL INFORMATICS
(2023)
Article
Computer Science, Artificial Intelligence
Chaobo He, Junwei Cheng, Xiang Fei, Yu Weng, Yulong Zheng, Yong Tang
Summary: Link prediction in attributed networks is challenging due to the need to effectively utilize community structure and attribute information. In this paper, we propose a novel CPAGCN method that combines AGCN and MLP to tackle this task. CPAGCN outperforms several strong competitors in link prediction, as demonstrated by extensive experiments on six real-world attributed networks.
KNOWLEDGE-BASED SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Asan Agibetov
Summary: The traditional method of learning neural graph embeddings by minimizing pointwise mutual information has limitations in capturing information from pairs of nodes with low co-occurrence. To address this issue, we propose an improved approach that incorporates information from unlikely node pairs and significantly improves link prediction performance.
PATTERN RECOGNITION
(2023)
Article
Computer Science, Information Systems
Chenji Huang, Yixiang Fang, Xuemin Lin, Xin Cao, Wenjie Zhang
Summary: In this article, we propose a novel prediction model called ABLE, which utilizes the Attention mechanism and BiLSTM for Embedding, to improve the performance of meta-path prediction in heterogeneous information networks. Experimental results show that ABLE outperforms existing methods on multiple real datasets.
ACM TRANSACTIONS ON KNOWLEDGE DISCOVERY FROM DATA
(2022)
Article
Computer Science, Artificial Intelligence
Jinyin Chen, Xueke Wang, Xuanheng Xu
Summary: Dynamic network link prediction is a hot topic in network science, and our proposed model GC-LSTM, combining GCN and LSTM, can predict both added and removed links, making it more practical in reality.
APPLIED INTELLIGENCE
(2022)
Article
Mathematics
Libin Chen, Luyao Wang, Chengyi Zeng, Hongfu Liu, Jing Chen
Summary: Current graph-embedding methods mainly focus on static homogeneous graphs, but real networks have various types of nodes and temporal interactions. This paper proposes a dynamic heterogeneous graph-embedding method called DHGEEP, which predicts the evolution of dynamic heterogeneous networks by simulating the evolution of graph and node dynamics and capturing semantic and structural information. Experimental results demonstrate the excellent performance of DHGEEP.
Article
Computer Science, Information Systems
Manling Li, Yuanzhuo Wang, Denghui Zhang, Yantao Jia, Xueqi Cheng
Summary: This research proposes a hierarchy-constrained link prediction method, called hTransM, which is based on knowledge graph embedding methods. By detecting hierarchical structures, hTransM can adaptively determine the optimal margin and has been proven effective through theoretical analysis and experiments.
IEEE TRANSACTIONS ON BIG DATA
(2022)
Article
Computer Science, Artificial Intelligence
Jun Zhao, Minglai Shao, Hong Wang, Xiaomei Yu, Bo Li, Xudong Liu
Summary: This paper proposes a Cyber Threat Prediction model called CTP-DHGL, based on Dynamic Heterogeneous Graph Learning, to predict potential cyber threats. The model analyzes public security-related data and outperforms baseline models in learning the evolutionary patterns of cyber threats and predicting potential risks.
KNOWLEDGE-BASED SYSTEMS
(2022)
Article
Chemistry, Multidisciplinary
Junhui Chen, Feihu Huang, Jian Peng
Summary: Heterogeneous graph embedding has become popular in recent years, but most existing methods do not utilize all auxiliary information effectively. The Multi-Subgraph based Graph Convolution Network (MSGCN) proposed in this study shows superior performance in multi-class node classification tasks through utilizing topology, semantic, and feature information.
APPLIED SCIENCES-BASEL
(2021)
Article
Computer Science, Artificial Intelligence
Luca Gallo, Vito Latora, Alfredo Pulvirenti
Summary: Research on graph representation learning has been highly focused on single-layer graphs, and there is limited research on representation learning of multilayer structures without known inter-layer links. This study proposes MultiplexSAGE, a generalized algorithm capable of embedding multiplex networks and reconstructing intra-layer and inter-layer connectivity. Experimental analysis reveals that the quality of embedding is strongly influenced by the density and randomness of the graph's links in both simple and multiplex networks.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
