Structuring of un-labeled data by machine recognition and its applications
| Project/Area Number |
26330286
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 一般 |
|---|
| Research Field |
Soft computing
|
|---|
| Research Institution | Waseda University |
|---|
Principal Investigator |
|
|---|
| Project Period (FY) |
2014-04-01 – 2017-03-31
|
| Project Status |
Completed (Fiscal Year 2016)
|
| Budget Amount *help |
¥4,810,000 (Direct Cost: ¥3,700,000、Indirect Cost: ¥1,110,000)
Fiscal Year 2016: ¥1,560,000 (Direct Cost: ¥1,200,000、Indirect Cost: ¥360,000)
Fiscal Year 2015: ¥1,560,000 (Direct Cost: ¥1,200,000、Indirect Cost: ¥360,000)
Fiscal Year 2014: ¥1,690,000 (Direct Cost: ¥1,300,000、Indirect Cost: ¥390,000)
|
|---|
| Keywords | 尤度最適化 / alpha-HMMアルゴリズム / alpha-EMアルゴリズム / 類似動画像検索 / 脳波認証 / P300波形 / エグゼンプラー / フレームシグネチャ / 競合学習 / 数値ラベル / ビッグデータ構造化 / 機械学習 / M-distance |
|---|
| Outline of Final Research Achievements |
We addressed the structuring of unorganized information that causes information flood. The theory here is the optimization of the likelihood. The first aspect is the theoretical development of the likelihood maximization method. The second is to present novel applications by paying attention to the likelihood.
On the theoretical contribution, the alpha-HMM estimation method (hidden Markov model) was presented. This method includes the traditional log-HMM method (Baum-Welch method) as a special case showing a fast convergence. The second contribution is the alpha-EM method (expectation-maximization). We found shotgun optimization methods that are the fastest.On the applications, we addressed the similar-video retrieval. We extract exemplar frames that form a numerical label. By applying the M-distance (Matsuyama-Moriwaki distance), we succeeded in finding similar videos containing unauthorized scenes. We made another application on the fraudulent PIN holders with high performance.
|
Report
(4 results)
Research Products
(13 results)
