Presentation 2004/6/18
Kernel density estimation and mode detection on the high dimensional hypersphere
Shigeyuki OBA, Shin ISHII,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Radial basis function (RBF) kernels are often used for kernel density estimation in a Euclidian space. We consider another family of kernels which can be applied to kernel density estimation in a hyperspherical space ; correlation kernels. A typical example of the family is von Mises-Fisher kernel. Based on correlation kernels, kernel density estimations can be done by using data points distributed on a hypersphere. We developped an algorithm that detects effectively the modes of the estimated kernel density. It resembles the mean-shift algorithm for RBF kernels, proposed by Comaniciu and Meer (2002), but is applicable to any correlation kernels. We apply it to clustering analysis of high dimensional data sets including gene expression profiles, and show the advantage over that based on Gaussian kernels.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) kernel density estimation / gene expression analysis / clustering / mean-shift / von Mises-Fisher distribution
Paper # NC2004-28
Date of Issue

Conference Information
Committee NC
Conference Date 2004/6/18(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Neurocomputing (NC)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Kernel density estimation and mode detection on the high dimensional hypersphere
Sub Title (in English)
Keyword(1) kernel density estimation
Keyword(2) gene expression analysis
Keyword(3) clustering
Keyword(4) mean-shift
Keyword(5) von Mises-Fisher distribution
1st Author's Name Shigeyuki OBA
1st Author's Affiliation Graduate School of Information Science, Nara Institute of Science and Technology()
2nd Author's Name Shin ISHII
2nd Author's Affiliation Graduate School of Information Science, Nara Institute of Science and Technology
Date 2004/6/18
Paper # NC2004-28
Volume (vol) vol.104
Number (no) 140
Page pp.pp.-
#Pages 6
Date of Issue