CHOSUN

효율적인 고차원 데이터 처리를 위한 N차원 볼의 체적 집중에 대한 분석

Metadata Downloads
Author(s)
김성복 윤형석
Issued Date
2019
Keyword
High-dimensional data N-dimensional ball Volume concentration Data sample distribution Machine learning
Abstract
This paper presents the volume concentration analysis of an n-dimensional ball (n-ball) defined in Euclidean space, near the surface and the equator, for efficient high-dimensional data processing. To quantify the volume concentration of an n-ball, two measures are defined: one measure as the volume ratio of a whole n-ball to the differential slice near the surface, and the other as the volume ratio of a whole n-ball to the differential slice near the equator. Without direct computation of the volumes of n-dimensional geometrical objects, both measures can be obtained as a function of the dimension of ball and the thickness of differential slice. According to computer simulation results for an n-ball, the surface volume concentration and the equator volume concentration show the changing patterns similar to each other, but the surface volume concentration is significantly stronger than the equator volume concentration. Finally, by interpreting the volume concentration of an n-ball as the distribution of high-dimensional data samples, a theoretical basis is provided for the planning of efficient high-dimensional data processing.
Alternative Title
Volume Concentration Analysis of an N-dimensional Ball for Efficient High Dimensional Data Processing
Alternative Author(s)
Sungbok Kim Hyungseok Yun
Publisher
조선대학교 공학기술연구원
Type
Research Laboratory
ISSN
2005-3142
URI
https://oak.chosun.ac.kr/handle/2020.oak/19333
https://www.chosun.ac.kr/user/indexSub.do?codyMenuSeq=23376167&siteId=riet&dum=dum&boardId=168878&page=3&command=view&boardSeq=315802&chkBoxSeq=&categoryId=&categoryDepth=&search=&column=null&searchDate1=&searchDate2=&selColumn=&myList=
Appears in Collections:
2019 > Vol 12, No 4
Authorize & License
  • AuthorizeOpen
Files in This Item:
  • There are no files associated with this item.

Items in Repository are protected by copyright, with all rights reserved, unless otherwise indicated.