An empirical study on the performance of spectral manifold learning techniques

Peter Mysling; Søren Hauberg; Kim Steenstrup Pedersen

doi:10.1007/978-3-642-21735-7_43

An empirical study on the performance of spectral manifold learning techniques

Peter Mysling, Søren Hauberg, Kim Steenstrup Pedersen

Department of Computer Science

3 Citations (Scopus)

Abstract

In recent years, there has been a surge of interest in spectral manifold learning techniques. Despite the interest, only little work has focused on the empirical behavior of these techniques. We construct synthetic data of variable complexity and observe the performance of the techniques as they are subjected to increasingly difficult problems. We evaluate performance in terms of both a classification and a regression task. Our study includes Isomap, LLE, Laplacian eigenmaps, and diffusion maps. Among others, our results indicate that the techniques are highly dependent on data density, sensitive to scaling, and greatly influenced by intrinsic dimensionality.

Original language	English
Title of host publication	Artificial Neural Networks and Machine Learning – ICANN 2011 : 21st International Conference on Artificial Neural Networks, Espoo, Finland, June 14-17, 2011, Proceedings, Part I
Editors	Timo Honkela, Włodzisław Duch, Mark Girolami, Samuel Kaski
Number of pages	8
Publisher	Springer
Publication date	2011
Pages	347-354
ISBN (Print)	978-3-642-21734-0
ISBN (Electronic)	978-3-642-21735
DOIs	https://doi.org/10.1007/978-3-642-21735-7_43
Publication status	Published - 2011
Event	21st International Conference on Artificial Neural Networks - Espoo, Finland Duration: 14 Jun 2011 → 17 Jun 2011 Conference number: 21

Conference

Conference	21st International Conference on Artificial Neural Networks
Number	21
Country/Territory	Finland
City	Espoo
Period	14/06/2011 → 17/06/2011

Series	Lecture notes in computer science
Volume	6791
ISSN	0302-9743

Access to Document

10.1007/978-3-642-21735-7_43

Cite this

Mysling, P., Hauberg, S., & Pedersen, K. S. (2011). An empirical study on the performance of spectral manifold learning techniques. In T. Honkela, W. Duch, M. Girolami, & S. Kaski (Eds.), Artificial Neural Networks and Machine Learning – ICANN 2011: 21st International Conference on Artificial Neural Networks, Espoo, Finland, June 14-17, 2011, Proceedings, Part I (pp. 347-354). Springer. Lecture notes in computer science Vol. 6791 https://doi.org/10.1007/978-3-642-21735-7_43

An empirical study on the performance of spectral manifold learning techniques. / Mysling, Peter; Hauberg, Søren; Pedersen, Kim Steenstrup.

Artificial Neural Networks and Machine Learning – ICANN 2011: 21st International Conference on Artificial Neural Networks, Espoo, Finland, June 14-17, 2011, Proceedings, Part I. ed. / Timo Honkela; Włodzisław Duch; Mark Girolami; Samuel Kaski. Springer, 2011. p. 347-354 (Lecture notes in computer science, Vol. 6791).

Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review

Mysling, P, Hauberg, S & Pedersen, KS 2011, An empirical study on the performance of spectral manifold learning techniques. in T Honkela, W Duch, M Girolami & S Kaski (eds), Artificial Neural Networks and Machine Learning – ICANN 2011: 21st International Conference on Artificial Neural Networks, Espoo, Finland, June 14-17, 2011, Proceedings, Part I. Springer, Lecture notes in computer science, vol. 6791, pp. 347-354, 21st International Conference on Artificial Neural Networks, Espoo, Finland, 14/06/2011. https://doi.org/10.1007/978-3-642-21735-7_43

Mysling P, Hauberg S, Pedersen KS. An empirical study on the performance of spectral manifold learning techniques. In Honkela T, Duch W, Girolami M, Kaski S, editors, Artificial Neural Networks and Machine Learning – ICANN 2011: 21st International Conference on Artificial Neural Networks, Espoo, Finland, June 14-17, 2011, Proceedings, Part I. Springer. 2011. p. 347-354. (Lecture notes in computer science, Vol. 6791). doi: 10.1007/978-3-642-21735-7_43

Mysling, Peter ; Hauberg, Søren ; Pedersen, Kim Steenstrup. / An empirical study on the performance of spectral manifold learning techniques. Artificial Neural Networks and Machine Learning – ICANN 2011: 21st International Conference on Artificial Neural Networks, Espoo, Finland, June 14-17, 2011, Proceedings, Part I. editor / Timo Honkela ; Włodzisław Duch ; Mark Girolami ; Samuel Kaski. Springer, 2011. pp. 347-354 (Lecture notes in computer science, Vol. 6791).

@inproceedings{9773d89a74d04acf9d6731832273a0bb,

title = "An empirical study on the performance of spectral manifold learning techniques",

abstract = "In recent years, there has been a surge of interest in spectral manifold learning techniques. Despite the interest, only little work has focused on the empirical behavior of these techniques. We construct synthetic data of variable complexity and observe the performance of the techniques as they are subjected to increasingly difficult problems. We evaluate performance in terms of both a classification and a regression task. Our study includes Isomap, LLE, Laplacian eigenmaps, and diffusion maps. Among others, our results indicate that the techniques are highly dependent on data density, sensitive to scaling, and greatly influenced by intrinsic dimensionality.",

author = "Peter Mysling and S{\o}ren Hauberg and Pedersen, {Kim Steenstrup}",

year = "2011",

doi = "10.1007/978-3-642-21735-7_43",

language = "English",

isbn = "978-3-642-21734-0",

series = "Lecture notes in computer science",

publisher = "Springer",

pages = "347--354",

editor = "Timo Honkela and Duch, { W{\l}odzis{\l}aw} and Mark Girolami and Samuel Kaski",

booktitle = "Artificial Neural Networks and Machine Learning – ICANN 2011",

note = "21st International Conference on Artificial Neural Networks ; Conference date: 14-06-2011 Through 17-06-2011",

}

TY - GEN

T1 - An empirical study on the performance of spectral manifold learning techniques

AU - Mysling, Peter

AU - Hauberg, Søren

AU - Pedersen, Kim Steenstrup

N1 - Conference code: 21

PY - 2011

Y1 - 2011

N2 - In recent years, there has been a surge of interest in spectral manifold learning techniques. Despite the interest, only little work has focused on the empirical behavior of these techniques. We construct synthetic data of variable complexity and observe the performance of the techniques as they are subjected to increasingly difficult problems. We evaluate performance in terms of both a classification and a regression task. Our study includes Isomap, LLE, Laplacian eigenmaps, and diffusion maps. Among others, our results indicate that the techniques are highly dependent on data density, sensitive to scaling, and greatly influenced by intrinsic dimensionality.

AB - In recent years, there has been a surge of interest in spectral manifold learning techniques. Despite the interest, only little work has focused on the empirical behavior of these techniques. We construct synthetic data of variable complexity and observe the performance of the techniques as they are subjected to increasingly difficult problems. We evaluate performance in terms of both a classification and a regression task. Our study includes Isomap, LLE, Laplacian eigenmaps, and diffusion maps. Among others, our results indicate that the techniques are highly dependent on data density, sensitive to scaling, and greatly influenced by intrinsic dimensionality.

U2 - 10.1007/978-3-642-21735-7_43

DO - 10.1007/978-3-642-21735-7_43

M3 - Article in proceedings

SN - 978-3-642-21734-0

T3 - Lecture notes in computer science

SP - 347

EP - 354

BT - Artificial Neural Networks and Machine Learning – ICANN 2011

A2 - Honkela, Timo

A2 - Duch, Włodzisław

A2 - Girolami, Mark

A2 - Kaski, Samuel

PB - Springer

T2 - 21st International Conference on Artificial Neural Networks

Y2 - 14 June 2011 through 17 June 2011

ER -