2019 Fiscal Year Research-status Report
Bipartite Graph Embedding: As A Unified Framework
| Project/Area Number |
19K20352
|
|---|
| Research Institution | National Institute of Advanced Industrial Science and Technology |
|---|
Principal Investigator |
劉 欣 国立研究開発法人産業技術総合研究所, 情報・人間工学領域, 研究員 (20803935)
|
|---|
| Project Period (FY) |
2019-04-01 – 2021-03-31
|
| Keywords | Graph Embedding / Graph Neural Network / Community Detection / Bipartite Graph / Node Ranking / Recommender System |
|---|
| Outline of Annual Research Achievements |
We have proposed a GNN based inductive framework to learn node embeddigns and approximate betweenness centrality. Our model is up to 44% improvement and up to 37 times faster compared with the-state-of-the methods. We have also proposed Recurrent Translation-Based Network (RTN) for top-N sparse sequential recommendation. Moreover, we presented an approach to recommend citations via knowledge graph embedding. Finally, we have studied how much graph structure is preserved by the current graph embedding techniques. In total, we have published eight papers, including two papers in top international conferences (CIKM2019, JCDL2020) and three papers in journals with impact factor (Entropy, IEEE Access, Computer Science and Information Systems).
|
| Current Status of Research Progress |
Current Status of Research Progress
1: Research has progressed more than it was originally planned.
Reason
The current progress of the project is going well. There are no delays. We are working now on 1) Bipartite graph embedding based on GCN and its application in Emergency Medical Service prediction, 2) Heterogeneous graph embedding. We have finished some of the works and submitted the papers.
|
| Strategy for Future Research Activity |
We will focus on graph embedding techniques under imperfect settings. For example, real-world graph data are often incomplete and contains missing features. In another example, real-world graph data are dynamic and changes with time. These imply that existing methods cannot be used directly. We will develop new approaches for practical usage of graph embedding techniques under these settings.
We will also continue applying graph embedding methods for solving various practical problems and applications.
|
| Causes of Carryover |
We failed to attend TheWebConf2020, which was canceled due to COVID-19. Considering the current special situation, we will use the travel expenses to buy computing and storage devices in order to carry out the project in remote working fashion.
|
