Ensemble learning method using structure information and its application

2016 Fiscal Year Final Research Report

• Project/Area Number: 25730018
• Research Category: Grant-in-Aid for Young Scientists (B)
• Allocation Type: Multi-year Fund
• Research Field: Statistical science
• Research Institution: Future University-Hakodate
• Principal Investigator: Takenouchi Takashi 公立はこだて未来大学, システム情報科学部, 准教授 (50403340)
• Project Period (FY): 2013-04-01 – 2017-03-31
• Keywords: アンサンブル学習 / 判別分析

Outline of Final Research Achievements

Multiclass classification problems sometimes require huge computational cost and then a major (and efficient) approach for the problem is to integrate multiple binary classifier. In this framework, we proposed a general framework of ensemble, which includes various kinds of conventional integration-based methods as special cases.
We proposed an ensemble-based method for the Multi-task problem. The proposed method is based on a combination of the Itakura-Saito distance and an extended model, rather than the conventional combination of the Kullback-Leibler divergence and statistical models. We revealed statistical properties of the proposed method and investigated validity of the proposed method.

Free Research Field

統計的機械学習