Information geometrical approach for distribution-integration data analysis
| Project/Area Number |
19500136
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Intelligent informatics
|
|---|
| Research Institution | National Institute of Advanced Industrial Science and Technology |
|---|
Principal Investigator |
AKAHO Shotaro National Institute of Advanced Industrial Science and Technology, 脳神経情報研究部門, 情報数理研究グループ長 (40356340)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
KAMISHIMA Toshihiro 独立行政法人産業技術総合研究所, 脳神経情報研究部門, 研究員 (50356820)
|
|---|
| Project Period (FY) |
2007 – 2009
|
| Project Status |
Completed (Fiscal Year 2009)
|
| Budget Amount *help |
¥4,550,000 (Direct Cost: ¥3,500,000、Indirect Cost: ¥1,050,000)
Fiscal Year 2009: ¥1,430,000 (Direct Cost: ¥1,100,000、Indirect Cost: ¥330,000)
Fiscal Year 2008: ¥1,430,000 (Direct Cost: ¥1,100,000、Indirect Cost: ¥330,000)
Fiscal Year 2007: ¥1,690,000 (Direct Cost: ¥1,300,000、Indirect Cost: ¥390,000)
|
|---|
| Keywords | 情報幾何学 / 次元縮約 / クラスタリング / ベイズ推定 / 協調フィルタリング |
|---|
| Research Abstract |
Novel methods of distribution-integration data analysis were constructed based on information geometrical framework. First, three extensions of the exponential family dimension reduction have been developed : application to mixture distributions that are not exponential family, Bayesian inference by probabilistic formulation, and simultaneous optimization of dimension reduction and clustering. Next, new machine learning paradigm "taming" which integrates small number of high-quality data and large number of low-quality data based on transfer learning was proposed.
|
Report
(4 results)
Research Products
(24 results)
