Localized centering: Reducing hubness in large-sample data

Kazuo Hara, Ikumi Suzuki, Masashi Shimbo, Kei Kobayashi, Kenji Fukumizu, Milos Radovanovic

Research output: Chapter in Book/Report/Conference proceedingConference contribution

14 Citations (Scopus)

Abstract

Hubness has been recently identified as a problematic phenomenon occurring in high-dimensional space. In this paper, we address a different type of hubness that occurs when the number of samples is large. We investigate the difference between the hubness in highdimensional data and the one in large-sample data. One finding is that centering, which is known to reduce the former, does not work for the latter. We then propose a new hub-reduction method, called localized centering. It is an extension of centering, yet works effectively for both types of hubness. Using real-world datasets consisting of a large number of documents, we demonstrate that the proposed method improves the accuracy of knearest neighbor classification.

Original languageEnglish
Title of host publicationProceedings of the 29th AAAI Conference on Artificial Intelligence, AAAI 2015 and the 27th Innovative Applications of Artificial Intelligence Conference, IAAI 2015
PublisherAI Access Foundation
Pages2645-2651
Number of pages7
Volume4
ISBN (Electronic)9781577357025
Publication statusPublished - 2015 Jun 1
Externally publishedYes
Event29th AAAI Conference on Artificial Intelligence, AAAI 2015 and the 27th Innovative Applications of Artificial Intelligence Conference, IAAI 2015 - Austin, United States
Duration: 2015 Jan 252015 Jan 30

Other

Other29th AAAI Conference on Artificial Intelligence, AAAI 2015 and the 27th Innovative Applications of Artificial Intelligence Conference, IAAI 2015
CountryUnited States
CityAustin
Period15/1/2515/1/30

ASJC Scopus subject areas

  • Software
  • Artificial Intelligence

Cite this

Hara, K., Suzuki, I., Shimbo, M., Kobayashi, K., Fukumizu, K., & Radovanovic, M. (2015). Localized centering: Reducing hubness in large-sample data. In Proceedings of the 29th AAAI Conference on Artificial Intelligence, AAAI 2015 and the 27th Innovative Applications of Artificial Intelligence Conference, IAAI 2015 (Vol. 4, pp. 2645-2651). AI Access Foundation.

Localized centering : Reducing hubness in large-sample data. / Hara, Kazuo; Suzuki, Ikumi; Shimbo, Masashi; Kobayashi, Kei; Fukumizu, Kenji; Radovanovic, Milos.

Proceedings of the 29th AAAI Conference on Artificial Intelligence, AAAI 2015 and the 27th Innovative Applications of Artificial Intelligence Conference, IAAI 2015. Vol. 4 AI Access Foundation, 2015. p. 2645-2651.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Hara, K, Suzuki, I, Shimbo, M, Kobayashi, K, Fukumizu, K & Radovanovic, M 2015, Localized centering: Reducing hubness in large-sample data. in Proceedings of the 29th AAAI Conference on Artificial Intelligence, AAAI 2015 and the 27th Innovative Applications of Artificial Intelligence Conference, IAAI 2015. vol. 4, AI Access Foundation, pp. 2645-2651, 29th AAAI Conference on Artificial Intelligence, AAAI 2015 and the 27th Innovative Applications of Artificial Intelligence Conference, IAAI 2015, Austin, United States, 15/1/25.
Hara K, Suzuki I, Shimbo M, Kobayashi K, Fukumizu K, Radovanovic M. Localized centering: Reducing hubness in large-sample data. In Proceedings of the 29th AAAI Conference on Artificial Intelligence, AAAI 2015 and the 27th Innovative Applications of Artificial Intelligence Conference, IAAI 2015. Vol. 4. AI Access Foundation. 2015. p. 2645-2651
Hara, Kazuo ; Suzuki, Ikumi ; Shimbo, Masashi ; Kobayashi, Kei ; Fukumizu, Kenji ; Radovanovic, Milos. / Localized centering : Reducing hubness in large-sample data. Proceedings of the 29th AAAI Conference on Artificial Intelligence, AAAI 2015 and the 27th Innovative Applications of Artificial Intelligence Conference, IAAI 2015. Vol. 4 AI Access Foundation, 2015. pp. 2645-2651
@inproceedings{e3dd8d36ed48427e8f4575269e8ca55f,
title = "Localized centering: Reducing hubness in large-sample data",
abstract = "Hubness has been recently identified as a problematic phenomenon occurring in high-dimensional space. In this paper, we address a different type of hubness that occurs when the number of samples is large. We investigate the difference between the hubness in highdimensional data and the one in large-sample data. One finding is that centering, which is known to reduce the former, does not work for the latter. We then propose a new hub-reduction method, called localized centering. It is an extension of centering, yet works effectively for both types of hubness. Using real-world datasets consisting of a large number of documents, we demonstrate that the proposed method improves the accuracy of knearest neighbor classification.",
author = "Kazuo Hara and Ikumi Suzuki and Masashi Shimbo and Kei Kobayashi and Kenji Fukumizu and Milos Radovanovic",
year = "2015",
month = "6",
day = "1",
language = "English",
volume = "4",
pages = "2645--2651",
booktitle = "Proceedings of the 29th AAAI Conference on Artificial Intelligence, AAAI 2015 and the 27th Innovative Applications of Artificial Intelligence Conference, IAAI 2015",
publisher = "AI Access Foundation",
address = "United States",

}

TY - GEN

T1 - Localized centering

T2 - Reducing hubness in large-sample data

AU - Hara, Kazuo

AU - Suzuki, Ikumi

AU - Shimbo, Masashi

AU - Kobayashi, Kei

AU - Fukumizu, Kenji

AU - Radovanovic, Milos

PY - 2015/6/1

Y1 - 2015/6/1

N2 - Hubness has been recently identified as a problematic phenomenon occurring in high-dimensional space. In this paper, we address a different type of hubness that occurs when the number of samples is large. We investigate the difference between the hubness in highdimensional data and the one in large-sample data. One finding is that centering, which is known to reduce the former, does not work for the latter. We then propose a new hub-reduction method, called localized centering. It is an extension of centering, yet works effectively for both types of hubness. Using real-world datasets consisting of a large number of documents, we demonstrate that the proposed method improves the accuracy of knearest neighbor classification.

AB - Hubness has been recently identified as a problematic phenomenon occurring in high-dimensional space. In this paper, we address a different type of hubness that occurs when the number of samples is large. We investigate the difference between the hubness in highdimensional data and the one in large-sample data. One finding is that centering, which is known to reduce the former, does not work for the latter. We then propose a new hub-reduction method, called localized centering. It is an extension of centering, yet works effectively for both types of hubness. Using real-world datasets consisting of a large number of documents, we demonstrate that the proposed method improves the accuracy of knearest neighbor classification.

UR - http://www.scopus.com/inward/record.url?scp=84960130900&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84960130900&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:84960130900

VL - 4

SP - 2645

EP - 2651

BT - Proceedings of the 29th AAAI Conference on Artificial Intelligence, AAAI 2015 and the 27th Innovative Applications of Artificial Intelligence Conference, IAAI 2015

PB - AI Access Foundation

ER -