| Project/Area Number |
17K00227
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 一般 |
|---|
| Research Field |
Perceptual information processing
|
|---|
| Research Institution | Chiba University |
|---|
Principal Investigator |
Mori Yasukuni 千葉大学, 大学院工学研究院, 助教 (40361414)
|
|---|
| Project Period (FY) |
2017-04-01 – 2020-03-31
|
| Project Status |
Completed (Fiscal Year 2019)
|
| Budget Amount *help |
¥2,860,000 (Direct Cost: ¥2,200,000、Indirect Cost: ¥660,000)
Fiscal Year 2019: ¥910,000 (Direct Cost: ¥700,000、Indirect Cost: ¥210,000)
Fiscal Year 2018: ¥1,040,000 (Direct Cost: ¥800,000、Indirect Cost: ¥240,000)
Fiscal Year 2017: ¥910,000 (Direct Cost: ¥700,000、Indirect Cost: ¥210,000)
|
|---|
| Keywords | 特徴選択 / 深層学習 / 特徴抽出 / データマイニング / ビッグデータ / 機械学習 |
|---|
| Outline of Final Research Achievements |
In this study, I propose a new layer model for feature selection using deep learning, which produces excellent results in various fields. The proposed method adds a layer called a feature selection layer in which units are arranged on a one-to-one basis with each feature to the input layer of the network model used, and performs learning using training data. As a result of learning, it is expected that the weight of the unit corresponding to the feature which effectively acts on the task increases and the weight of the unnecessary feature decreases in the target task. Therefore, by using the value of this weight, it is possible to select effective features for the target task.
|
| Academic Significance and Societal Importance of the Research Achievements |
本研究で提案した特徴選択手法を利用することで，従来の手法では非常に難しかった，超高次元のデータに対しても特徴の選択をすることが可能になった．これにより，例えば，数千を超える特徴集合の中から，注目している識別タスクに有効に作用する重要な特徴を選別することが可能になり，これまで以上に探索的データ解析における新たな知見が得られる可能性を見出した．
|
