宋熙然 Xiran Song

Graduate Student of Computer Science

Huazhong University of Science and Technology

Biography

Xiran Song is a third-year graduate student in the School of Computer Science and Technology, Huazhong University of Science and Technology (HUST), advised by Professor Hong Huang and Senior Researcher Jianxun Lian. Previously, he obtained his B.S. degree from HUST in 2021. His research interests include data mining and graph representation learning.

Interests

Data Mining
Graph Representation Learning

Education

MEng in Computer Science, 2021 - 2024 (anticipated)
Huazhong University of Science and Technology
BEng in Computer Science, 2017 - 2021
Huazhong University of Science and Technology

Featured Publications

Xiran Song, Jianxun Lian, Hong Huang, Zihan Luo, Wei Zhou, Xue Lin, Mingqi Wu, Chaozhuo Li, Xing Xie, Hai Jin

April, 2023 In WWW'23

xGCN: An Extreme Graph Convolutional Network for Large-scale Social Link Prediction

Abstract: Graph neural networks (GNNs) have seen widespread usage across multiple real-world applications, yet in transductive learning, they still face challenges in accuracy, efficiency, and scalability, due to the extensive number of trainable parameters in the embedding table and the paradigm of stacking neighborhood aggregations.

Publications

Wei Zhou, Hong Huang, Ruize Shi, Xiran Song, Xue Lin, Xiao Wang, Hai Jin (2023). Temporal Heterogeneous Information Network Embedding via Semantic Evolution. TKDE'23.

PDF Cite

Xiran Song, Jianxun Lian, Hong Huang, Zihan Luo, Wei Zhou, Xue Lin, Mingqi Wu, Chaozhuo Li, Xing Xie, Hai Jin (2023). xGCN: An Extreme Graph Convolutional Network for Large-scale Social Link Prediction. In WWW'23.

PDF Cite Code Video

Xiran Song, Jianxun Lian, Hong Huang, Mingqi Wu, Hai Jin, Xing Xie (2022). Friend Recommendations with Self-Rescaling Graph Neural Networks. In KDD'22.

PDF Cite Code Video

Projects

XGCN

XGCN is a light-weight and easy-to-use library for large-scale GNN embedding. XGCN includes xGCN - an implementation for the TheWebConf 2023 paper: xGCN: An Extreme Graph Convolutional Network for Large-scale Social Link Prediction - which achieves remarkable accuracy and efficiency on large graphs, including an industrial dataset with 100 million nodes.