Dimensionality Reduction Through Sub-Space Mapping for Nearest Neighbour Algorithms

T R Payne, P Edwards

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

5 Citations (Scopus)
4 Downloads (Pure)

Abstract

Many learning algorithms make an implicit assumption that all the attributes present in the data are relevant to a learning task. However, several studies have demonstrated that this assumption rarely holds; for many supervised learning algorithms, the inclusion of irrelevant or redundant attributes can result in a degradation in classification accuracy. While a variety of different methods for dimensionality reduction exist, many of these are only appropriate for datasets which contain a small number of attributes (e.g. < 20). This paper presents an alternative approach to dimensionality reduction, and demonstrates how it can be combined with a Nearest Neighbour learning algorithm. We present an empirical evaluation of this approach, and contrast its performance with two related techniques; a Monte-Carlo wrapper and an Information Gain-based filter approach.

Original languageEnglish
Title of host publicationMachine Learning: ECML 2000
Subtitle of host publication11th European Conference on Machine Learning Barcelona, Catalonia, Spain, May 31 – June 2, 2000 Proceedings
EditorsRamon López de Mantaras, Enric Plaza
Place of PublicationBerlin, Germany
PublisherSpringer-Verlag
Pages331-343
Number of pages13
Volume1810
ISBN (Electronic)978-3-540-45164-8
ISBN (Print)978-3-540-67602-7
DOIs
Publication statusPublished - 2000

Publication series

NameLecture Notes in Artificial Intelligence
PublisherSpringer

Keywords

  • machine learning
  • feature selection
  • nearest-neighbour

Cite this

Payne, T. R., & Edwards, P. (2000). Dimensionality Reduction Through Sub-Space Mapping for Nearest Neighbour Algorithms. In R. L. de Mantaras, & E. Plaza (Eds.), Machine Learning: ECML 2000: 11th European Conference on Machine Learning Barcelona, Catalonia, Spain, May 31 – June 2, 2000 Proceedings (Vol. 1810, pp. 331-343). (Lecture Notes in Artificial Intelligence). Berlin, Germany: Springer-Verlag. https://doi.org/10.1007/3-540-45164-1_35

Dimensionality Reduction Through Sub-Space Mapping for Nearest Neighbour Algorithms. / Payne, T R ; Edwards, P .

Machine Learning: ECML 2000: 11th European Conference on Machine Learning Barcelona, Catalonia, Spain, May 31 – June 2, 2000 Proceedings. ed. / Ramon López de Mantaras; Enric Plaza. Vol. 1810 Berlin, Germany : Springer-Verlag, 2000. p. 331-343 (Lecture Notes in Artificial Intelligence).

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

Payne, TR & Edwards, P 2000, Dimensionality Reduction Through Sub-Space Mapping for Nearest Neighbour Algorithms. in RL de Mantaras & E Plaza (eds), Machine Learning: ECML 2000: 11th European Conference on Machine Learning Barcelona, Catalonia, Spain, May 31 – June 2, 2000 Proceedings. vol. 1810, Lecture Notes in Artificial Intelligence, Springer-Verlag, Berlin, Germany, pp. 331-343. https://doi.org/10.1007/3-540-45164-1_35
Payne TR, Edwards P. Dimensionality Reduction Through Sub-Space Mapping for Nearest Neighbour Algorithms. In de Mantaras RL, Plaza E, editors, Machine Learning: ECML 2000: 11th European Conference on Machine Learning Barcelona, Catalonia, Spain, May 31 – June 2, 2000 Proceedings. Vol. 1810. Berlin, Germany: Springer-Verlag. 2000. p. 331-343. (Lecture Notes in Artificial Intelligence). https://doi.org/10.1007/3-540-45164-1_35
Payne, T R ; Edwards, P . / Dimensionality Reduction Through Sub-Space Mapping for Nearest Neighbour Algorithms. Machine Learning: ECML 2000: 11th European Conference on Machine Learning Barcelona, Catalonia, Spain, May 31 – June 2, 2000 Proceedings. editor / Ramon López de Mantaras ; Enric Plaza. Vol. 1810 Berlin, Germany : Springer-Verlag, 2000. pp. 331-343 (Lecture Notes in Artificial Intelligence).
@inproceedings{ceec7bf46d3a4288bf6d120047c733f5,
title = "Dimensionality Reduction Through Sub-Space Mapping for Nearest Neighbour Algorithms",
abstract = "Many learning algorithms make an implicit assumption that all the attributes present in the data are relevant to a learning task. However, several studies have demonstrated that this assumption rarely holds; for many supervised learning algorithms, the inclusion of irrelevant or redundant attributes can result in a degradation in classification accuracy. While a variety of different methods for dimensionality reduction exist, many of these are only appropriate for datasets which contain a small number of attributes (e.g. < 20). This paper presents an alternative approach to dimensionality reduction, and demonstrates how it can be combined with a Nearest Neighbour learning algorithm. We present an empirical evaluation of this approach, and contrast its performance with two related techniques; a Monte-Carlo wrapper and an Information Gain-based filter approach.",
keywords = "machine learning, feature selection, nearest-neighbour",
author = "Payne, {T R} and P Edwards",
year = "2000",
doi = "10.1007/3-540-45164-1_35",
language = "English",
isbn = "978-3-540-67602-7",
volume = "1810",
series = "Lecture Notes in Artificial Intelligence",
publisher = "Springer-Verlag",
pages = "331--343",
editor = "{de Mantaras}, {Ramon L{\'o}pez} and Enric Plaza",
booktitle = "Machine Learning: ECML 2000",

}

TY - GEN

T1 - Dimensionality Reduction Through Sub-Space Mapping for Nearest Neighbour Algorithms

AU - Payne, T R

AU - Edwards, P

PY - 2000

Y1 - 2000

N2 - Many learning algorithms make an implicit assumption that all the attributes present in the data are relevant to a learning task. However, several studies have demonstrated that this assumption rarely holds; for many supervised learning algorithms, the inclusion of irrelevant or redundant attributes can result in a degradation in classification accuracy. While a variety of different methods for dimensionality reduction exist, many of these are only appropriate for datasets which contain a small number of attributes (e.g. < 20). This paper presents an alternative approach to dimensionality reduction, and demonstrates how it can be combined with a Nearest Neighbour learning algorithm. We present an empirical evaluation of this approach, and contrast its performance with two related techniques; a Monte-Carlo wrapper and an Information Gain-based filter approach.

AB - Many learning algorithms make an implicit assumption that all the attributes present in the data are relevant to a learning task. However, several studies have demonstrated that this assumption rarely holds; for many supervised learning algorithms, the inclusion of irrelevant or redundant attributes can result in a degradation in classification accuracy. While a variety of different methods for dimensionality reduction exist, many of these are only appropriate for datasets which contain a small number of attributes (e.g. < 20). This paper presents an alternative approach to dimensionality reduction, and demonstrates how it can be combined with a Nearest Neighbour learning algorithm. We present an empirical evaluation of this approach, and contrast its performance with two related techniques; a Monte-Carlo wrapper and an Information Gain-based filter approach.

KW - machine learning

KW - feature selection

KW - nearest-neighbour

U2 - 10.1007/3-540-45164-1_35

DO - 10.1007/3-540-45164-1_35

M3 - Conference contribution

SN - 978-3-540-67602-7

VL - 1810

T3 - Lecture Notes in Artificial Intelligence

SP - 331

EP - 343

BT - Machine Learning: ECML 2000

A2 - de Mantaras, Ramon López

A2 - Plaza, Enric

PB - Springer-Verlag

CY - Berlin, Germany

ER -