A study on classifier design by integration of the probabilistic evidences and it applications to image recognitions

2015 Fiscal Year Final Research Report

• Project/Area Number: 23500211
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Perception information processing/Intelligent robotics
• Research Institution: Hiroshima University
• Principal Investigator: Kurita Takio 広島大学, 工学(系)研究科(研究院), 教授 (10356941)
• Co-Investigator(Kenkyū-buntansha): HIDAKA AKINORI 東京電機大学, 理工学部, 助教 (70553519)
• Project Period (FY): 2011-04-28 – 2016-03-31
• Keywords: 判別カーネル / 事後確率 / カーネル学習 / 判別分析 / 画像認識 / 特徴抽出 / サポートベクターマシン

Outline of Final Research Achievements

It is known that the optimum nonlinear discriminant mapping can be defined as a linear combination of the Bayesian a posterior probabilities and the coefficients of the linear combination are obtained by solving the eigenvalue problem of the matrices defined by using the Bayesian a posterior probabilities. We derived the optimum kernel function (discriminant kernel) which were used in the optimum nonlinear discriminate mapping. It is also defined by using the posterior probabilities. For real applications, we have to estimate the posterior probabilities. In this study, we proposed several nonlinear discriminant analysis methods in which the probabilities were estimated by multinomial logistic regression or K nearest neighbors methods etc. To improve the generalization ability, the probabilities were estimated with regularization. Several applications of the image recognition or image retrieval were also tried.

Free Research Field

情報工学