Empirical Bayes Kernels: Unsupervised Kernel Learning
Project/Area Number |
25540100
|
Research Category |
Grant-in-Aid for Challenging Exploratory Research
|
Allocation Type | Multi-year Fund |
Research Field |
Intelligent informatics
|
Research Institution | Kyoto University |
Principal Investigator |
CUTURI Marco 京都大学, 情報学研究科, 准教授 (80597344)
|
Project Period (FY) |
2013-04-01 – 2015-03-31
|
Project Status |
Completed (Fiscal Year 2014)
|
Budget Amount *help |
¥3,770,000 (Direct Cost: ¥2,900,000、Indirect Cost: ¥870,000)
Fiscal Year 2014: ¥1,430,000 (Direct Cost: ¥1,100,000、Indirect Cost: ¥330,000)
Fiscal Year 2013: ¥2,340,000 (Direct Cost: ¥1,800,000、Indirect Cost: ¥540,000)
|
Keywords | 機械学習 / 距離学習 / ヒストグラムデータ / Metric learning / Kernel learning / Aitchison geometry / Probability Simplex / Contrastive Divergence / Stochastic Optimization / metric learning / kernel methods / probability simplex / empirical Bayes |
Outline of Final Research Achievements |
Our goal in this work was to provide a principled approach to carry out kernel/metric learning in an unsupervised way, to take advantage of large datasets of unlabeled data. We investigated this research avenue by focusing mostly on histogram data (bags-of-features). Using a combination of 3 known approaches by Aitchison, Lebanon and Hinton, we were able to propose different algorithms which perform at state-of-the art level or directly outperform competing approaches.
|
Report
(3 results)
Research Products
(7 results)