Multivariate analysis of multi-domain data considering the association between data vectors
Project/Area Number |
16H02789
|
Research Category |
Grant-in-Aid for Scientific Research (B)
|
Allocation Type | Single-year Grants |
Section | 一般 |
Research Field |
Statistical science
|
Research Institution | Kyoto University (2017-2019) Osaka University (2016) |
Principal Investigator |
|
Co-Investigator(Kenkyū-buntansha) |
清水 昌平 滋賀大学, データサイエンス学部, 教授 (10509871)
|
Project Period (FY) |
2016-04-01 – 2020-03-31
|
Project Status |
Completed (Fiscal Year 2019)
|
Budget Amount *help |
¥16,900,000 (Direct Cost: ¥13,000,000、Indirect Cost: ¥3,900,000)
Fiscal Year 2019: ¥2,730,000 (Direct Cost: ¥2,100,000、Indirect Cost: ¥630,000)
Fiscal Year 2018: ¥2,860,000 (Direct Cost: ¥2,200,000、Indirect Cost: ¥660,000)
Fiscal Year 2017: ¥3,120,000 (Direct Cost: ¥2,400,000、Indirect Cost: ¥720,000)
Fiscal Year 2016: ¥8,190,000 (Direct Cost: ¥6,300,000、Indirect Cost: ¥1,890,000)
|
Keywords | 多変量解析 / パターン認識 / グラフ埋め込み / 次元削減 / 分散表現 / ニューラルネットワーク / マルチモーダル / 自然言語処理 / 漸近理論 / 画像検索 |
Outline of Final Research Achievements |
Multi-domain association data consists of data vectors from various types of sources (called domains), such as images, tags, documents, etc., and the strength of associations between data vectors. Conventional multivariate analysis has dealt with one-to-one vector correspondence, and therefore, it cannot represent flexible data structures. In this study, we describe the relationship between vectors as a graph (network). Then, we have proposed and developed methods of information integration via dimensionality reduction, which preserves the graph structure as much as possible.
|
Academic Significance and Societal Importance of the Research Achievements |
関連性データのグラフ構造をなるべく保存するようにデータベクトルを変換することをグラフ埋め込みという.正準相関分析など従来の多変量解析を一般化したグラフ埋め込み手法を提案し,画像と単語の相互検索などのタスクで有効性を確認した.ニューラルネットワークによる非線形変換を用いたグラフ埋め込み法を提案し,さらに外れ値の影響を軽減するロバスト化を行った.ベクトル間の内積とそれを発展させたニューラルネットワークモデルによって表現できる類似度関数のクラスを明らかにした.
|
Report
(5 results)
Research Products
(64 results)