New development in hubness-aware metric learning in high dimensional data
| Project/Area Number |
15H02749
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Intelligent informatics
|
|---|
| Research Institution | Nara Institute of Science and Technology |
|---|
Principal Investigator |
SHIMBO Masashi 奈良先端科学技術大学院大学, 先端科学技術研究科, 准教授 (90311589)
|
|---|
| Project Period (FY) |
2015-04-01 – 2019-03-31
|
| Project Status |
Completed (Fiscal Year 2018)
|
| Budget Amount *help |
¥16,120,000 (Direct Cost: ¥12,400,000、Indirect Cost: ¥3,720,000)
Fiscal Year 2018: ¥3,770,000 (Direct Cost: ¥2,900,000、Indirect Cost: ¥870,000)
Fiscal Year 2017: ¥3,770,000 (Direct Cost: ¥2,900,000、Indirect Cost: ¥870,000)
Fiscal Year 2016: ¥3,770,000 (Direct Cost: ¥2,900,000、Indirect Cost: ¥870,000)
Fiscal Year 2015: ¥4,810,000 (Direct Cost: ¥3,700,000、Indirect Cost: ¥1,110,000)
|
|---|
| Keywords | ベクトル空間モデル / 高次元空間 / 近傍検索 / 表現学習 / 知識グラフ / 多変量回帰 / ベクトル表現モデル / 知識獲得 / 知識表現 / データマイニング / 距離・類似度尺度 / 高次元データ / グラフ埋め込み / 距離学習 / リンク解析 / 類似度 / 機械学習 / 人工知能 / ベクタースペースモデル / ゼロショット学習 / ドメイン対応 / 回帰 |
|---|
| Outline of Final Research Achievements |
A theoretical and empirical investigation was carried out on the similarity/dissimilarity measures in high-dimensional space. In particular, a special focus was placed on analyzing the hubness phenomenon. On the basis of this analysis, we have proposed approaches for sevearl applications.
Our major contributions are (i) the analysis of regression-based object-matching over two domains, and the proposal of a simple yet effective method dervied from the results of the analysis; (ii) application of our analysis for (i) to improving (single-domain) k-nearest neighbor classification; and (iii) analysis of knowledge graph embedding methods (in particular those embed objects in complex-valued space).
|
| Academic Significance and Societal Importance of the Research Achievements |
成果の概要で述べた貢献 (i) については, 従来の手法の欠点を理論的分析に基づいて明らかにし, 単純だが効果的な手法を提案した. 従来法と計算コストは変わらないが, 精度を高めることができる. (ii) は (i) の応用であるが, やはり簡単かつ効率的であり, 距離学習に比較してはるかに小さな計算量で, 精度向上が可能である. (iii) は知識グラフ埋め込みに関する研究である. 最近, 知識グラフを有効活用した自然言語処理での様々な応用が試みられており, われわれの分析および精度向上は, これら実応用の精度向上の一助になると考えている.
|
Report
(5 results)
Research Products
(17 results)
-
-
-
-
-
-
-
-
-
-
-
-
-
-
[Presentation] Ridge regression, hubness, and zero-shot learning2015
Author(s)
Yutaro Shigeto, Ikumi Suzuki, Kazuo Hara, Masashi Shimbo, Yuji Matsumoto
Organizer
2015 European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECMLPKDD 2015)
Place of Presentation
Porto, Porgugal
Year and Date
2015-09-07
Related Report
Int'l Joint Research
-
-
-
