Denoising and dimension reduction in feature space

Mikio L. Braun, Joachim Buhmann, Klaus Muller

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

11 Citations (Scopus)

Abstract

We show that the relevant information about a classification problem in feature space is contained up to negligible error in a finite number of leading kernel PCA components if the kernel matches the underlying learning problem. Thus, kernels not only transform data sets such that good generalization can be achieved even by linear discriminant functions, but this transformation is also performed in a manner which makes economic use of feature space dimensions. In the best case, kernels provide efficient implicit representations of the data to perform classification. Practically, we propose an algorithm which enables us to recover the subspace and dimensionality relevant for good classification. Our algorithm can therefore be applied (1) to analyze the interplay of data set and kernel in a geometric fashion, (2) to help in model selection, and to (3) de-noise in feature space in order to yield better classification results.

Original languageEnglish
Title of host publicationAdvances in Neural Information Processing Systems
Pages185-192
Number of pages8
Publication statusPublished - 2007 Dec 1
Externally publishedYes
Event20th Annual Conference on Neural Information Processing Systems, NIPS 2006 - Vancouver, BC, Canada
Duration: 2006 Dec 42006 Dec 7

Other

Other20th Annual Conference on Neural Information Processing Systems, NIPS 2006
CountryCanada
CityVancouver, BC
Period06/12/406/12/7

Fingerprint

Economics

ASJC Scopus subject areas

  • Computer Networks and Communications
  • Information Systems
  • Signal Processing

Cite this

Braun, M. L., Buhmann, J., & Muller, K. (2007). Denoising and dimension reduction in feature space. In Advances in Neural Information Processing Systems (pp. 185-192)

Denoising and dimension reduction in feature space. / Braun, Mikio L.; Buhmann, Joachim; Muller, Klaus.

Advances in Neural Information Processing Systems. 2007. p. 185-192.

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

Braun, ML, Buhmann, J & Muller, K 2007, Denoising and dimension reduction in feature space. in Advances in Neural Information Processing Systems. pp. 185-192, 20th Annual Conference on Neural Information Processing Systems, NIPS 2006, Vancouver, BC, Canada, 06/12/4.
Braun ML, Buhmann J, Muller K. Denoising and dimension reduction in feature space. In Advances in Neural Information Processing Systems. 2007. p. 185-192
Braun, Mikio L. ; Buhmann, Joachim ; Muller, Klaus. / Denoising and dimension reduction in feature space. Advances in Neural Information Processing Systems. 2007. pp. 185-192
@inproceedings{9ab459a2be2c42e3ae7c7b7078ac50b5,
title = "Denoising and dimension reduction in feature space",
abstract = "We show that the relevant information about a classification problem in feature space is contained up to negligible error in a finite number of leading kernel PCA components if the kernel matches the underlying learning problem. Thus, kernels not only transform data sets such that good generalization can be achieved even by linear discriminant functions, but this transformation is also performed in a manner which makes economic use of feature space dimensions. In the best case, kernels provide efficient implicit representations of the data to perform classification. Practically, we propose an algorithm which enables us to recover the subspace and dimensionality relevant for good classification. Our algorithm can therefore be applied (1) to analyze the interplay of data set and kernel in a geometric fashion, (2) to help in model selection, and to (3) de-noise in feature space in order to yield better classification results.",
author = "Braun, {Mikio L.} and Joachim Buhmann and Klaus Muller",
year = "2007",
month = "12",
day = "1",
language = "English",
isbn = "9780262195683",
pages = "185--192",
booktitle = "Advances in Neural Information Processing Systems",

}

TY - GEN

T1 - Denoising and dimension reduction in feature space

AU - Braun, Mikio L.

AU - Buhmann, Joachim

AU - Muller, Klaus

PY - 2007/12/1

Y1 - 2007/12/1

N2 - We show that the relevant information about a classification problem in feature space is contained up to negligible error in a finite number of leading kernel PCA components if the kernel matches the underlying learning problem. Thus, kernels not only transform data sets such that good generalization can be achieved even by linear discriminant functions, but this transformation is also performed in a manner which makes economic use of feature space dimensions. In the best case, kernels provide efficient implicit representations of the data to perform classification. Practically, we propose an algorithm which enables us to recover the subspace and dimensionality relevant for good classification. Our algorithm can therefore be applied (1) to analyze the interplay of data set and kernel in a geometric fashion, (2) to help in model selection, and to (3) de-noise in feature space in order to yield better classification results.

AB - We show that the relevant information about a classification problem in feature space is contained up to negligible error in a finite number of leading kernel PCA components if the kernel matches the underlying learning problem. Thus, kernels not only transform data sets such that good generalization can be achieved even by linear discriminant functions, but this transformation is also performed in a manner which makes economic use of feature space dimensions. In the best case, kernels provide efficient implicit representations of the data to perform classification. Practically, we propose an algorithm which enables us to recover the subspace and dimensionality relevant for good classification. Our algorithm can therefore be applied (1) to analyze the interplay of data set and kernel in a geometric fashion, (2) to help in model selection, and to (3) de-noise in feature space in order to yield better classification results.

UR - http://www.scopus.com/inward/record.url?scp=84862272240&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84862272240&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:84862272240

SN - 9780262195683

SP - 185

EP - 192

BT - Advances in Neural Information Processing Systems

ER -