Kernel VA-files for relevance feedback retrieval

Douglas R. Heisterkamp, Jing Peng

Research output: Chapter in Book/Report/Conference proceedingConference contributionResearchpeer-review

6 Citations (Scopus)

Abstract

Many data partitioning index methods perform poorly in high dimensional space and do not support relevance feedback retrieval. The vector approximation file (VA-File) approach overcomes some of the difficulties of high dimensional vector spaces, but cannot be applied to relevance feedback retrieval using kernel distances in the data measurement space. This paper introduces a novel KVA-File (kernel VA-File) that extends VA-File to kernel-based retrieval methods. A key observation is that kernel distances may be non-linear in the data measurement space but is still linear in an induced feature space. It is this linear invariance in the induced feature space that enables KVA-File to work with kernel distances. An efficient approach to approximating vectors in an induced feature space is presented with the corresponding upper and lower distance bounds. Thus an effective indexing method is provided for kernel-based relevance feedback image retrieval methods. Experimental results using large image data sets (approximately 100,000 images with 463 dimensions of measurement) validate the efficacy of our method.

Original languageEnglish
Title of host publicationMMDB 2003
Subtitle of host publicationProceedings of the First ACM International Workshop on Multimedia Databases
EditorsS.-C. Chen, M.-L. Shyo
Pages48-54
Number of pages7
StatePublished - 1 Dec 2003
EventMMDB 2003: Proceedings of the First ACM International Workshop on Multimedia Databases - New Orleans, LA, United States
Duration: 7 Nov 20037 Nov 2003

Other

OtherMMDB 2003: Proceedings of the First ACM International Workshop on Multimedia Databases
CountryUnited States
CityNew Orleans, LA
Period7/11/037/11/03

Fingerprint

Feedback
Image retrieval
Vector spaces
Invariance

Keywords

  • Content-based image retrieval
  • Flexible metrics
  • Indexing
  • Kernel methods
  • Relevance feedback
  • VA-Files

Cite this

Heisterkamp, D. R., & Peng, J. (2003). Kernel VA-files for relevance feedback retrieval. In S-C. Chen, & M-L. Shyo (Eds.), MMDB 2003: Proceedings of the First ACM International Workshop on Multimedia Databases (pp. 48-54)
Heisterkamp, Douglas R. ; Peng, Jing. / Kernel VA-files for relevance feedback retrieval. MMDB 2003: Proceedings of the First ACM International Workshop on Multimedia Databases. editor / S.-C. Chen ; M.-L. Shyo. 2003. pp. 48-54
@inproceedings{3c6a6e344db541d7b6a55d864599c95a,
title = "Kernel VA-files for relevance feedback retrieval",
abstract = "Many data partitioning index methods perform poorly in high dimensional space and do not support relevance feedback retrieval. The vector approximation file (VA-File) approach overcomes some of the difficulties of high dimensional vector spaces, but cannot be applied to relevance feedback retrieval using kernel distances in the data measurement space. This paper introduces a novel KVA-File (kernel VA-File) that extends VA-File to kernel-based retrieval methods. A key observation is that kernel distances may be non-linear in the data measurement space but is still linear in an induced feature space. It is this linear invariance in the induced feature space that enables KVA-File to work with kernel distances. An efficient approach to approximating vectors in an induced feature space is presented with the corresponding upper and lower distance bounds. Thus an effective indexing method is provided for kernel-based relevance feedback image retrieval methods. Experimental results using large image data sets (approximately 100,000 images with 463 dimensions of measurement) validate the efficacy of our method.",
keywords = "Content-based image retrieval, Flexible metrics, Indexing, Kernel methods, Relevance feedback, VA-Files",
author = "Heisterkamp, {Douglas R.} and Jing Peng",
year = "2003",
month = "12",
day = "1",
language = "English",
isbn = "1581137265",
pages = "48--54",
editor = "S.-C. Chen and M.-L. Shyo",
booktitle = "MMDB 2003",

}

Heisterkamp, DR & Peng, J 2003, Kernel VA-files for relevance feedback retrieval. in S-C Chen & M-L Shyo (eds), MMDB 2003: Proceedings of the First ACM International Workshop on Multimedia Databases. pp. 48-54, MMDB 2003: Proceedings of the First ACM International Workshop on Multimedia Databases, New Orleans, LA, United States, 7/11/03.

Kernel VA-files for relevance feedback retrieval. / Heisterkamp, Douglas R.; Peng, Jing.

MMDB 2003: Proceedings of the First ACM International Workshop on Multimedia Databases. ed. / S.-C. Chen; M.-L. Shyo. 2003. p. 48-54.

Research output: Chapter in Book/Report/Conference proceedingConference contributionResearchpeer-review

TY - GEN

T1 - Kernel VA-files for relevance feedback retrieval

AU - Heisterkamp, Douglas R.

AU - Peng, Jing

PY - 2003/12/1

Y1 - 2003/12/1

N2 - Many data partitioning index methods perform poorly in high dimensional space and do not support relevance feedback retrieval. The vector approximation file (VA-File) approach overcomes some of the difficulties of high dimensional vector spaces, but cannot be applied to relevance feedback retrieval using kernel distances in the data measurement space. This paper introduces a novel KVA-File (kernel VA-File) that extends VA-File to kernel-based retrieval methods. A key observation is that kernel distances may be non-linear in the data measurement space but is still linear in an induced feature space. It is this linear invariance in the induced feature space that enables KVA-File to work with kernel distances. An efficient approach to approximating vectors in an induced feature space is presented with the corresponding upper and lower distance bounds. Thus an effective indexing method is provided for kernel-based relevance feedback image retrieval methods. Experimental results using large image data sets (approximately 100,000 images with 463 dimensions of measurement) validate the efficacy of our method.

AB - Many data partitioning index methods perform poorly in high dimensional space and do not support relevance feedback retrieval. The vector approximation file (VA-File) approach overcomes some of the difficulties of high dimensional vector spaces, but cannot be applied to relevance feedback retrieval using kernel distances in the data measurement space. This paper introduces a novel KVA-File (kernel VA-File) that extends VA-File to kernel-based retrieval methods. A key observation is that kernel distances may be non-linear in the data measurement space but is still linear in an induced feature space. It is this linear invariance in the induced feature space that enables KVA-File to work with kernel distances. An efficient approach to approximating vectors in an induced feature space is presented with the corresponding upper and lower distance bounds. Thus an effective indexing method is provided for kernel-based relevance feedback image retrieval methods. Experimental results using large image data sets (approximately 100,000 images with 463 dimensions of measurement) validate the efficacy of our method.

KW - Content-based image retrieval

KW - Flexible metrics

KW - Indexing

KW - Kernel methods

KW - Relevance feedback

KW - VA-Files

UR - http://www.scopus.com/inward/record.url?scp=13444307692&partnerID=8YFLogxK

M3 - Conference contribution

SN - 1581137265

SN - 9781581137262

SP - 48

EP - 54

BT - MMDB 2003

A2 - Chen, S.-C.

A2 - Shyo, M.-L.

ER -

Heisterkamp DR, Peng J. Kernel VA-files for relevance feedback retrieval. In Chen S-C, Shyo M-L, editors, MMDB 2003: Proceedings of the First ACM International Workshop on Multimedia Databases. 2003. p. 48-54