Spatially adaptive sparse grids for high-dimensional data-driven problems

Dirk Pflger, Benjamin Peherstorfer, Hans Joachim Bungartz

Research output: Contribution to journalArticle

Abstract

Sparse grids allow one to employ grid-based discretization methods in data-driven problems. We present an extension of the classical sparse grid approach that allows us to tackle high-dimensional problems by spatially adaptive refinement, modified ansatz functions, and efficient regularization techniques. The competitiveness of this method is shown for typical benchmark problems with up to 166 dimensions for classification in data mining, pointing out properties of sparse grids in this context. To gain insight into the adaptive refinement and to examine the scope for further improvements, the approximation of non-smooth indicator functions with adaptive sparse grids has been studied as a model problem. As an example for an improved adaptive grid refinement, we present results for an edge-detection strategy.

Original languageEnglish (US)
Pages (from-to)508-522
Number of pages15
JournalJournal of Complexity
Volume26
Issue number5
DOIs
StatePublished - Jan 1 2010

Fingerprint

Sparse Grids
Adaptive Grid
High-dimensional Data
Data-driven
Adaptive Refinement
Edge detection
Data mining
Grid Refinement
Indicator function
Nonsmooth Function
Regularization Technique
Discretization Method
Competitiveness
Edge Detection
Data Mining
High-dimensional
Benchmark
Grid
Approximation

Keywords

  • Classification
  • High-dimensional approximation
  • Non-smooth functions
  • Regularization
  • Spatially adaptive sparse grids

ASJC Scopus subject areas

  • Algebra and Number Theory
  • Statistics and Probability
  • Numerical Analysis
  • Control and Optimization
  • Applied Mathematics

Cite this

Spatially adaptive sparse grids for high-dimensional data-driven problems. / Pflger, Dirk; Peherstorfer, Benjamin; Bungartz, Hans Joachim.

In: Journal of Complexity, Vol. 26, No. 5, 01.01.2010, p. 508-522.

Research output: Contribution to journalArticle

Pflger, Dirk ; Peherstorfer, Benjamin ; Bungartz, Hans Joachim. / Spatially adaptive sparse grids for high-dimensional data-driven problems. In: Journal of Complexity. 2010 ; Vol. 26, No. 5. pp. 508-522.
@article{c9e8ef7bc1304982b650f1d9aa5f142f,
title = "Spatially adaptive sparse grids for high-dimensional data-driven problems",
abstract = "Sparse grids allow one to employ grid-based discretization methods in data-driven problems. We present an extension of the classical sparse grid approach that allows us to tackle high-dimensional problems by spatially adaptive refinement, modified ansatz functions, and efficient regularization techniques. The competitiveness of this method is shown for typical benchmark problems with up to 166 dimensions for classification in data mining, pointing out properties of sparse grids in this context. To gain insight into the adaptive refinement and to examine the scope for further improvements, the approximation of non-smooth indicator functions with adaptive sparse grids has been studied as a model problem. As an example for an improved adaptive grid refinement, we present results for an edge-detection strategy.",
keywords = "Classification, High-dimensional approximation, Non-smooth functions, Regularization, Spatially adaptive sparse grids",
author = "Dirk Pflger and Benjamin Peherstorfer and Bungartz, {Hans Joachim}",
year = "2010",
month = "1",
day = "1",
doi = "10.1016/j.jco.2010.04.001",
language = "English (US)",
volume = "26",
pages = "508--522",
journal = "Journal of Complexity",
issn = "0885-064X",
publisher = "Academic Press Inc.",
number = "5",

}

TY - JOUR

T1 - Spatially adaptive sparse grids for high-dimensional data-driven problems

AU - Pflger, Dirk

AU - Peherstorfer, Benjamin

AU - Bungartz, Hans Joachim

PY - 2010/1/1

Y1 - 2010/1/1

N2 - Sparse grids allow one to employ grid-based discretization methods in data-driven problems. We present an extension of the classical sparse grid approach that allows us to tackle high-dimensional problems by spatially adaptive refinement, modified ansatz functions, and efficient regularization techniques. The competitiveness of this method is shown for typical benchmark problems with up to 166 dimensions for classification in data mining, pointing out properties of sparse grids in this context. To gain insight into the adaptive refinement and to examine the scope for further improvements, the approximation of non-smooth indicator functions with adaptive sparse grids has been studied as a model problem. As an example for an improved adaptive grid refinement, we present results for an edge-detection strategy.

AB - Sparse grids allow one to employ grid-based discretization methods in data-driven problems. We present an extension of the classical sparse grid approach that allows us to tackle high-dimensional problems by spatially adaptive refinement, modified ansatz functions, and efficient regularization techniques. The competitiveness of this method is shown for typical benchmark problems with up to 166 dimensions for classification in data mining, pointing out properties of sparse grids in this context. To gain insight into the adaptive refinement and to examine the scope for further improvements, the approximation of non-smooth indicator functions with adaptive sparse grids has been studied as a model problem. As an example for an improved adaptive grid refinement, we present results for an edge-detection strategy.

KW - Classification

KW - High-dimensional approximation

KW - Non-smooth functions

KW - Regularization

KW - Spatially adaptive sparse grids

UR - http://www.scopus.com/inward/record.url?scp=83755201307&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=83755201307&partnerID=8YFLogxK

U2 - 10.1016/j.jco.2010.04.001

DO - 10.1016/j.jco.2010.04.001

M3 - Article

AN - SCOPUS:83755201307

VL - 26

SP - 508

EP - 522

JO - Journal of Complexity

JF - Journal of Complexity

SN - 0885-064X

IS - 5

ER -