SepMe: 2002 New visual separation measures

Michael Aupetit, Michael Sedlmair

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

15 Citations (Scopus)

Abstract

Our goal is to accurately model human class separation judgements in color-coded scatterplots. Towards this goal, we propose a set of 2002 visual separation measures, by systematically combining 17 neighborhood graphs and 14 class purity functions, with different parameterizations. Using a Machine Learning framework, we evaluate these measures based on how well they predict human separation judgements. We found that more than 58% of the 2002 new measures outperform the best state-of-the-art Distance Consistency (DSC) measure. Among the 2002, the best measure is the average proportion of same-class neighbors among the 0.35-Observable Neighbors of each point of the target class (short GONG 0.35 DIR CPT), with a prediction accuracy of 92.9%, which is 11.7% better than DSC. We also discuss alternative, well-performing measures and give guidelines when to use which.

Original languageEnglish
Title of host publication2016 IEEE Pacific Visualization Symposium, PacificVis 2016 - Proceedings
PublisherIEEE Computer Society
Pages1-8
Number of pages8
Volume2016-May
ISBN (Electronic)9781509014514
DOIs
Publication statusPublished - 4 May 2016
Event9th IEEE Pacific Visualization Symposium, PacificVis 2016 - Taipei, Taiwan, Province of China
Duration: 19 Apr 201622 Apr 2016

Other

Other9th IEEE Pacific Visualization Symposium, PacificVis 2016
CountryTaiwan, Province of China
CityTaipei
Period19/4/1622/4/16

Fingerprint

Parameterization
Learning systems
Color

Keywords

  • H.5.2 [Information Interfaces and Presentation]: User Interfaces - Theory and methods

ASJC Scopus subject areas

  • Computer Graphics and Computer-Aided Design
  • Computer Vision and Pattern Recognition
  • Hardware and Architecture
  • Software

Cite this

Aupetit, M., & Sedlmair, M. (2016). SepMe: 2002 New visual separation measures. In 2016 IEEE Pacific Visualization Symposium, PacificVis 2016 - Proceedings (Vol. 2016-May, pp. 1-8). [7465244] IEEE Computer Society. https://doi.org/10.1109/PACIFICVIS.2016.7465244

SepMe : 2002 New visual separation measures. / Aupetit, Michael; Sedlmair, Michael.

2016 IEEE Pacific Visualization Symposium, PacificVis 2016 - Proceedings. Vol. 2016-May IEEE Computer Society, 2016. p. 1-8 7465244.

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

Aupetit, M & Sedlmair, M 2016, SepMe: 2002 New visual separation measures. in 2016 IEEE Pacific Visualization Symposium, PacificVis 2016 - Proceedings. vol. 2016-May, 7465244, IEEE Computer Society, pp. 1-8, 9th IEEE Pacific Visualization Symposium, PacificVis 2016, Taipei, Taiwan, Province of China, 19/4/16. https://doi.org/10.1109/PACIFICVIS.2016.7465244
Aupetit M, Sedlmair M. SepMe: 2002 New visual separation measures. In 2016 IEEE Pacific Visualization Symposium, PacificVis 2016 - Proceedings. Vol. 2016-May. IEEE Computer Society. 2016. p. 1-8. 7465244 https://doi.org/10.1109/PACIFICVIS.2016.7465244
Aupetit, Michael ; Sedlmair, Michael. / SepMe : 2002 New visual separation measures. 2016 IEEE Pacific Visualization Symposium, PacificVis 2016 - Proceedings. Vol. 2016-May IEEE Computer Society, 2016. pp. 1-8
@inproceedings{2d219845ee214df8bac9310c56d8675f,
title = "SepMe: 2002 New visual separation measures",
abstract = "Our goal is to accurately model human class separation judgements in color-coded scatterplots. Towards this goal, we propose a set of 2002 visual separation measures, by systematically combining 17 neighborhood graphs and 14 class purity functions, with different parameterizations. Using a Machine Learning framework, we evaluate these measures based on how well they predict human separation judgements. We found that more than 58{\%} of the 2002 new measures outperform the best state-of-the-art Distance Consistency (DSC) measure. Among the 2002, the best measure is the average proportion of same-class neighbors among the 0.35-Observable Neighbors of each point of the target class (short GONG 0.35 DIR CPT), with a prediction accuracy of 92.9{\%}, which is 11.7{\%} better than DSC. We also discuss alternative, well-performing measures and give guidelines when to use which.",
keywords = "H.5.2 [Information Interfaces and Presentation]: User Interfaces - Theory and methods",
author = "Michael Aupetit and Michael Sedlmair",
year = "2016",
month = "5",
day = "4",
doi = "10.1109/PACIFICVIS.2016.7465244",
language = "English",
volume = "2016-May",
pages = "1--8",
booktitle = "2016 IEEE Pacific Visualization Symposium, PacificVis 2016 - Proceedings",
publisher = "IEEE Computer Society",

}

TY - GEN

T1 - SepMe

T2 - 2002 New visual separation measures

AU - Aupetit, Michael

AU - Sedlmair, Michael

PY - 2016/5/4

Y1 - 2016/5/4

N2 - Our goal is to accurately model human class separation judgements in color-coded scatterplots. Towards this goal, we propose a set of 2002 visual separation measures, by systematically combining 17 neighborhood graphs and 14 class purity functions, with different parameterizations. Using a Machine Learning framework, we evaluate these measures based on how well they predict human separation judgements. We found that more than 58% of the 2002 new measures outperform the best state-of-the-art Distance Consistency (DSC) measure. Among the 2002, the best measure is the average proportion of same-class neighbors among the 0.35-Observable Neighbors of each point of the target class (short GONG 0.35 DIR CPT), with a prediction accuracy of 92.9%, which is 11.7% better than DSC. We also discuss alternative, well-performing measures and give guidelines when to use which.

AB - Our goal is to accurately model human class separation judgements in color-coded scatterplots. Towards this goal, we propose a set of 2002 visual separation measures, by systematically combining 17 neighborhood graphs and 14 class purity functions, with different parameterizations. Using a Machine Learning framework, we evaluate these measures based on how well they predict human separation judgements. We found that more than 58% of the 2002 new measures outperform the best state-of-the-art Distance Consistency (DSC) measure. Among the 2002, the best measure is the average proportion of same-class neighbors among the 0.35-Observable Neighbors of each point of the target class (short GONG 0.35 DIR CPT), with a prediction accuracy of 92.9%, which is 11.7% better than DSC. We also discuss alternative, well-performing measures and give guidelines when to use which.

KW - H.5.2 [Information Interfaces and Presentation]: User Interfaces - Theory and methods

UR - http://www.scopus.com/inward/record.url?scp=84973640346&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84973640346&partnerID=8YFLogxK

U2 - 10.1109/PACIFICVIS.2016.7465244

DO - 10.1109/PACIFICVIS.2016.7465244

M3 - Conference contribution

AN - SCOPUS:84973640346

VL - 2016-May

SP - 1

EP - 8

BT - 2016 IEEE Pacific Visualization Symposium, PacificVis 2016 - Proceedings

PB - IEEE Computer Society

ER -