Topological Analysis of Scalar Fields with Outliers

Mickaël Buchet, Frédéric Chazal, Tamal K. Dey, Fengtao Fan, Steve Y. Oudot, Yusu Wang

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

Abstract

Given a real-valued function f defined over a manifold M embedded in Rd, we are interested in recovering structural information about f from the sole information of its values on a finite sample P. Existing methods provide approximation to the persistence diagram of f when geometric noise and functional noise are bounded. However, they fail in the presence of aberrant values, also called outliers, both in theory and practice. We propose a new algorithm that deals with outliers. We handle aberrant functional values with a method inspired from the k-nearest neighbors regression and the local median filtering, while the geometric outliers are handled using the distance to a measure. Combined with topological results on nested filtrations, our algorithm performs robust topological analysis of scalar fields in a wider range of noise models than handled by current methods. We provide theoretical guarantees and experimental results on the quality of our approximation of the sampled scalar field.

Original languageEnglish
Title of host publication31st International Symposium on Computational Geometry, SoCG 2015
PublisherSchloss Dagstuhl, Leibniz-Zentrum fü Informatik GmbH
Pages827-841
Number of pages15
Volume34
ISBN (Electronic)9783939897835
DOIs
Publication statusPublished - 1 Jun 2015
Externally publishedYes
Event31st International Symposium on Computational Geometry, SoCG 2015 - Eindhoven, Netherlands
Duration: 22 Jun 201525 Jun 2015

Conference

Conference31st International Symposium on Computational Geometry, SoCG 2015
Abbreviated titleSoCG 2015
CountryNetherlands
CityEindhoven
Period22/06/1525/06/15

Keywords

  • Distance to a Measure
  • Nested Rips Filtration
  • Persistent Homology
  • Scalar Field Analysis
  • Topological Data Analysis

ASJC Scopus subject areas

  • Software

Cite this

Buchet, M., Chazal, F., Dey, T. K., Fan, F., Oudot, S. Y., & Wang, Y. (2015). Topological Analysis of Scalar Fields with Outliers. In 31st International Symposium on Computational Geometry, SoCG 2015 (Vol. 34, pp. 827-841). Schloss Dagstuhl, Leibniz-Zentrum fü Informatik GmbH. https://doi.org/10.4230/LIPIcs.SOCG.2015.827

Topological Analysis of Scalar Fields with Outliers. / Buchet, Mickaël; Chazal, Frédéric; Dey, Tamal K.; Fan, Fengtao; Oudot, Steve Y.; Wang, Yusu.

31st International Symposium on Computational Geometry, SoCG 2015. Vol. 34 Schloss Dagstuhl, Leibniz-Zentrum fü Informatik GmbH, 2015. p. 827-841.

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

Buchet, M, Chazal, F, Dey, TK, Fan, F, Oudot, SY & Wang, Y 2015, Topological Analysis of Scalar Fields with Outliers. in 31st International Symposium on Computational Geometry, SoCG 2015. vol. 34, Schloss Dagstuhl, Leibniz-Zentrum fü Informatik GmbH, pp. 827-841, 31st International Symposium on Computational Geometry, SoCG 2015, Eindhoven, Netherlands, 22/06/15. https://doi.org/10.4230/LIPIcs.SOCG.2015.827
Buchet M, Chazal F, Dey TK, Fan F, Oudot SY, Wang Y. Topological Analysis of Scalar Fields with Outliers. In 31st International Symposium on Computational Geometry, SoCG 2015. Vol. 34. Schloss Dagstuhl, Leibniz-Zentrum fü Informatik GmbH. 2015. p. 827-841 https://doi.org/10.4230/LIPIcs.SOCG.2015.827
Buchet, Mickaël ; Chazal, Frédéric ; Dey, Tamal K. ; Fan, Fengtao ; Oudot, Steve Y. ; Wang, Yusu. / Topological Analysis of Scalar Fields with Outliers. 31st International Symposium on Computational Geometry, SoCG 2015. Vol. 34 Schloss Dagstuhl, Leibniz-Zentrum fü Informatik GmbH, 2015. pp. 827-841
@inproceedings{6da2b5f0641a4fb9a7f5e63011884d10,
title = "Topological Analysis of Scalar Fields with Outliers",
abstract = "Given a real-valued function f defined over a manifold M embedded in Rd, we are interested in recovering structural information about f from the sole information of its values on a finite sample P. Existing methods provide approximation to the persistence diagram of f when geometric noise and functional noise are bounded. However, they fail in the presence of aberrant values, also called outliers, both in theory and practice. We propose a new algorithm that deals with outliers. We handle aberrant functional values with a method inspired from the k-nearest neighbors regression and the local median filtering, while the geometric outliers are handled using the distance to a measure. Combined with topological results on nested filtrations, our algorithm performs robust topological analysis of scalar fields in a wider range of noise models than handled by current methods. We provide theoretical guarantees and experimental results on the quality of our approximation of the sampled scalar field.",
keywords = "Distance to a Measure, Nested Rips Filtration, Persistent Homology, Scalar Field Analysis, Topological Data Analysis",
author = "Micka{\"e}l Buchet and Fr{\'e}d{\'e}ric Chazal and Dey, {Tamal K.} and Fengtao Fan and Oudot, {Steve Y.} and Yusu Wang",
year = "2015",
month = "6",
day = "1",
doi = "10.4230/LIPIcs.SOCG.2015.827",
language = "English",
volume = "34",
pages = "827--841",
booktitle = "31st International Symposium on Computational Geometry, SoCG 2015",
publisher = "Schloss Dagstuhl, Leibniz-Zentrum f{\"u} Informatik GmbH",
address = "Germany",

}

TY - GEN

T1 - Topological Analysis of Scalar Fields with Outliers

AU - Buchet, Mickaël

AU - Chazal, Frédéric

AU - Dey, Tamal K.

AU - Fan, Fengtao

AU - Oudot, Steve Y.

AU - Wang, Yusu

PY - 2015/6/1

Y1 - 2015/6/1

N2 - Given a real-valued function f defined over a manifold M embedded in Rd, we are interested in recovering structural information about f from the sole information of its values on a finite sample P. Existing methods provide approximation to the persistence diagram of f when geometric noise and functional noise are bounded. However, they fail in the presence of aberrant values, also called outliers, both in theory and practice. We propose a new algorithm that deals with outliers. We handle aberrant functional values with a method inspired from the k-nearest neighbors regression and the local median filtering, while the geometric outliers are handled using the distance to a measure. Combined with topological results on nested filtrations, our algorithm performs robust topological analysis of scalar fields in a wider range of noise models than handled by current methods. We provide theoretical guarantees and experimental results on the quality of our approximation of the sampled scalar field.

AB - Given a real-valued function f defined over a manifold M embedded in Rd, we are interested in recovering structural information about f from the sole information of its values on a finite sample P. Existing methods provide approximation to the persistence diagram of f when geometric noise and functional noise are bounded. However, they fail in the presence of aberrant values, also called outliers, both in theory and practice. We propose a new algorithm that deals with outliers. We handle aberrant functional values with a method inspired from the k-nearest neighbors regression and the local median filtering, while the geometric outliers are handled using the distance to a measure. Combined with topological results on nested filtrations, our algorithm performs robust topological analysis of scalar fields in a wider range of noise models than handled by current methods. We provide theoretical guarantees and experimental results on the quality of our approximation of the sampled scalar field.

KW - Distance to a Measure

KW - Nested Rips Filtration

KW - Persistent Homology

KW - Scalar Field Analysis

KW - Topological Data Analysis

UR - http://www.scopus.com/inward/record.url?scp=84958162110&partnerID=8YFLogxK

U2 - 10.4230/LIPIcs.SOCG.2015.827

DO - 10.4230/LIPIcs.SOCG.2015.827

M3 - Conference contribution

VL - 34

SP - 827

EP - 841

BT - 31st International Symposium on Computational Geometry, SoCG 2015

PB - Schloss Dagstuhl, Leibniz-Zentrum fü Informatik GmbH

ER -