2009 Fiscal Year Final Research Report
SYSTEMATIZATION AND GENERALIZATION OF DISCRIMINATIVE TRAINING
| Project/Area Number |
19300064
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Perception information processing/Intelligent robotics
|
|---|
| Research Institution | Doshisha University |
|---|
Principal Investigator |
KATAGIRI Shigeru Doshisha University, 理工学部, 教授 (40396114)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
OHSAKI Miho 同志社大学, 理工学部, 准教授 (30313927)
NAKAMURA Atsushi 日本電信電話(株), NTTコミュニケーション科学基礎研究所, 主幹研究員 (50396206)
MCDERMOTT Erik (株)国際電気通信基礎技術研究所, 研究員 (40447024)
|
|---|
| Project Period (FY) |
2007 – 2009
|
| Keywords | 一般化確率的降下法 / 最小分類誤り学習法 / 識別学習 / パターン認識 / サポートベクタマシン / ブースティング |
|---|
| Research Abstract |
The systematization of discriminative training methods was conducted based on equivalence between misclassification measure and functional margin and on a unified formalization of microscopic- and macroscopic-unit classification. In addition, a geometric margin was derived for a general class of discriminant functions, and a new MCE method was developed by employing the geometric margin as the misclassification measure. The incremental-MCE method was also formalized that consisted of stepwise MCE training for the elemental weak classifiers. The results were implemented in educational software named DISCERN.
|
Research Products
(9 results)
