Research

Graph data is widely used as a fundamental data model for representing relationships between data entities. In recent years, graph data has grown enormously in scale, and processing such data requires a huge amount of computation time, which has become a barrier to its applications. KDE-SALT conducts research and development on ultrafast data processing, data mining, search, and learning for large-scale graph data.

Major Research Results:

• Fast graph clustering [IJCAI'19] [VLDB'15] [AAAI'13]
• Fast subgraph search [IJCAI'24] [IJCAI'23]
• Fast PageRank-based algorithms [SIGMOD'13] [ICDE'13] [AAAI'13] [KDD'12]

Data management technologies such as databases are essential infrastructure that supports society, and handling large-scale data demands even higher performance. In addition, data may have complex structures, as in real-world graph data, so new data management technologies for such complex data are urgently needed. KDE-SALT conducts research and development on high-performance query processing for large-scale data, graph database technologies, and data cleaning technologies.

Major Research Results:

• Ultrafast query processing [IEICE'25] [ISWC'21] [WWW'18]
• Compressed search for graph data [IEICE'25] [IEICE'24]
• Data documentation for sound data sharing [GRADES-NDA'25]

Artificial intelligence has spread widely throughout society and is used in a variety of decision-making processes. In recent years, not only the accuracy of analysis and inference but also how efficiently data can be processed has become increasingly important. Toward efficient artificial intelligence, we develop fast model training algorithms and efficient graph learning models, and study techniques for accelerating vector search, which underpins retrieval-augmented generation (RAG).

Major Research Results:

• Fast Affinity Propagation [AAAI'21] [KDD'15]
• Fast training of regression models [AAAI'16]
• Fast and accurate graph classification models [IEICE'24]

Data science and artificial intelligence have spread widely, and large-scale data processing is needed in many domains. KDE-SALT applies fast data processing technologies in practice to other fields such as chemistry, medicine, and engineering.

Major Research Results:

• Graph-based molecular search for AI-driven drug discovery [J. Cheminfo.‘25] [IJCAI'24] [iiWAS'23]
• Graph-based music recommendation using playlists [ASONAM'23]
• Automatic sleep stage classification [Sci. Rep.‘19] [ICDE Workshop'17]
• Similar activity detection using sensor data [iiWAS'23]