A Study on nonlinear classification methods based on the optimization of the structure of the reproducing kernel Hilbert space

2015 Fiscal Year Final Research Report

• Project/Area Number: 25870811
• Research Category: Grant-in-Aid for Young Scientists (B)
• Allocation Type: Multi-year Fund
• Research Field: Intelligent informatics; Statistical science
• Research Institution: University of Tsukuba
• Principal Investigator: HINO Hideitsu 筑波大学, システム情報系, 助教 (10580079)
• Project Period (FY): 2013-04-01 – 2016-03-31
• Keywords: 機械学習 / 応用統計 / パターン認識

Outline of Final Research Achievements

Kernel method is one of the most important methods in machine learning. Its effectiveness depends on the kernel function and its parameter. We developed methods for optimizing the kernel function based on the data distribution of the intrinsic high dimensional space associated with the kernel function. We also proposed a similarity measure for the given data and the newly observed data based on the notion of information content in the given data, and applied the proposed measure for classifying a set of data and finding outliers from the observed dataset. We applied the developed methods to speaker recognition, hand gesture recognition, protein structure classification, and change point detection from time series data.

Free Research Field

機械学習と応用統計