Supervised learning for inhomogeneous set of graphs
| Project/Area Number |
26330242
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 一般 |
|---|
| Research Field |
Intelligent informatics
|
|---|
| Research Institution | Hokkaido University |
|---|
Principal Investigator |
|
|---|
| Project Period (FY) |
2014-04-01 – 2017-03-31
|
| Project Status |
Completed (Fiscal Year 2016)
|
| Budget Amount *help |
¥4,030,000 (Direct Cost: ¥3,100,000、Indirect Cost: ¥930,000)
Fiscal Year 2016: ¥1,040,000 (Direct Cost: ¥800,000、Indirect Cost: ¥240,000)
Fiscal Year 2015: ¥1,040,000 (Direct Cost: ¥800,000、Indirect Cost: ¥240,000)
Fiscal Year 2014: ¥1,950,000 (Direct Cost: ¥1,500,000、Indirect Cost: ¥450,000)
|
|---|
| Keywords | 機械学習 / グラフ / 潜在構造 / 特徴表現 |
|---|
| Outline of Final Research Achievements |
When supervised learning over graphs is applied to, for example, real molecular graphs in QSAR, it suffers from the 'inhomogeneity' originated from mixing different data sources and different underlying mechanisms. To address this problem, we conducted research on the following four topics: 1) develop and analyze computational methods for simultaneous learning of predictive model and relevant subgraph features among all possible ones; 2) analyze the properties of feature space of subgraph indicators with real datasets, in particular, boolean structures, correlation structures, and redundancy; 3) develop computational methods for learning decision and regression trees over all possible subgraph features, and its ensemble learning by boosting; 4) develop a relaxed feature representation by introducing wildcard labels to node and edge labels of graphs.
|
Report
(4 results)
Research Products
(25 results)
