05月26日(Fri) 09:30〜11:10 Q会場(ウインクあいち-8F 展示場)
| 演題番号 | 4Q1-2in2 |
|---|---|
| 題目 | Deploying exploration in proximity indices for link collection problem |
| 著者 | Zhang Guoxi(Department of Intelligence Science and Technology, Graduate School of Informatics, Kyoto University) 鹿島 久嗣(京都大学大学院情報学研究科知能情報学専攻) |
| 時間 | 05月26日(Fri) 09:30〜11:10 |
| 概要 | Link analysis has attracted attention from wide range of disciplines as it reveals latent knowledge from relational data. In applications like drug discovery or reactome analysis, identifying unobserved relations from data may reduce cost of wet experiments. In this paper, we study the problem of link collection. Given fixed set of nodes and a set of links, an algorithm for this problem iteratively outputs pair of nodes as queries. Queries are labeled by an annotator and links among them are utilized in further query calculation. We examine the effect of deploying several exploration strategies into proximity indices. In our experiments modifications improve performance of proximity indices on several datasets. Our work provides insights about how exploration can be deployed into proximity indices for graph mining tasks. |
| 論文 | PDFファイル |