Multi-level visualisation using Gaussian process latent variable models

Shahzad Mumtaz, Darren R. Flower, Ian Nabney

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

Abstract

Projection of a high-dimensional dataset onto a two-dimensional space is a useful tool to visualise structures and relationships in the dataset. However, a single two-dimensional visualisation may not display all the intrinsic structure. Therefore, hierarchical/multi-level visualisation methods have been used to extract more detailed understanding of the data. Here we propose a multi-level Gaussian process latent variable model (MLGPLVM). MLGPLVM works by segmenting data (with e.g. K-means, Gaussian mixture model or interactive clustering) in the visualisation space and then fitting a visualisation model to each subset. To measure the quality of multi-level visualisation (with respect to parent and child models), metrics such as trustworthiness, continuity, mean relative rank errors, visualisation distance distortion and the negative log-likelihood per point are used. We evaluate the MLGPLVM approach on the ‘Oil Flow’ dataset and a dataset of protein electrostatic potentials for the ‘Major Histocompatibility Complex (MHC) class I’ of humans. In both cases, visual observation and the quantitative quality measures have shown better visualisation at lower levels.
Original languageEnglish
Title of host publicationIVAPP 2014
Subtitle of host publicationproceedings of the 5th international conference on Information Visualization Theory and Applications
EditorsRobert S. Laramee, Andreas Kerren, José Braz
Place of PublicationLisbon (PT)
PublisherSciTePress
Pages122-129
Number of pages8
ISBN (Print)978-989-758-005-5
Publication statusPublished - 2014
Event5th international conference on Information Visualization Theory and Applications - Lisbon, Portugal
Duration: 5 Jan 20148 Jan 2014

Conference

Conference5th international conference on Information Visualization Theory and Applications
Abbreviated titleIVAPP 2014
CountryPortugal
CityLisbon
Period5/01/148/01/14

Fingerprint

Visualization
Electrostatics
Proteins

Keywords

  • continuity
  • Gaussian mixture model
  • K-means
  • major histocompatibility complex
  • mean relative rank errors
  • multi-level Gaussian process latent variable model
  • negative log-likelihood
  • trustworthiness
  • visualisation distance distortion

Cite this

Mumtaz, S., Flower, D. R., & Nabney, I. (2014). Multi-level visualisation using Gaussian process latent variable models. In R. S. Laramee, A. Kerren, & J. Braz (Eds.), IVAPP 2014: proceedings of the 5th international conference on Information Visualization Theory and Applications (pp. 122-129). Lisbon (PT): SciTePress.
Mumtaz, Shahzad ; Flower, Darren R. ; Nabney, Ian. / Multi-level visualisation using Gaussian process latent variable models. IVAPP 2014: proceedings of the 5th international conference on Information Visualization Theory and Applications. editor / Robert S. Laramee ; Andreas Kerren ; José Braz. Lisbon (PT) : SciTePress, 2014. pp. 122-129
@inproceedings{3f3aa124b58e4611bd828af2d1de7f76,
title = "Multi-level visualisation using Gaussian process latent variable models",
abstract = "Projection of a high-dimensional dataset onto a two-dimensional space is a useful tool to visualise structures and relationships in the dataset. However, a single two-dimensional visualisation may not display all the intrinsic structure. Therefore, hierarchical/multi-level visualisation methods have been used to extract more detailed understanding of the data. Here we propose a multi-level Gaussian process latent variable model (MLGPLVM). MLGPLVM works by segmenting data (with e.g. K-means, Gaussian mixture model or interactive clustering) in the visualisation space and then fitting a visualisation model to each subset. To measure the quality of multi-level visualisation (with respect to parent and child models), metrics such as trustworthiness, continuity, mean relative rank errors, visualisation distance distortion and the negative log-likelihood per point are used. We evaluate the MLGPLVM approach on the ‘Oil Flow’ dataset and a dataset of protein electrostatic potentials for the ‘Major Histocompatibility Complex (MHC) class I’ of humans. In both cases, visual observation and the quantitative quality measures have shown better visualisation at lower levels.",
keywords = "continuity, Gaussian mixture model, K-means, major histocompatibility complex, mean relative rank errors, multi-level Gaussian process latent variable model, negative log-likelihood, trustworthiness, visualisation distance distortion",
author = "Shahzad Mumtaz and Flower, {Darren R.} and Ian Nabney",
year = "2014",
language = "English",
isbn = "978-989-758-005-5",
pages = "122--129",
editor = "Laramee, {Robert S.} and Andreas Kerren and Jos{\'e} Braz",
booktitle = "IVAPP 2014",
publisher = "SciTePress",

}

Mumtaz, S, Flower, DR & Nabney, I 2014, Multi-level visualisation using Gaussian process latent variable models. in RS Laramee, A Kerren & J Braz (eds), IVAPP 2014: proceedings of the 5th international conference on Information Visualization Theory and Applications. SciTePress, Lisbon (PT), pp. 122-129, 5th international conference on Information Visualization Theory and Applications, Lisbon, Portugal, 5/01/14.

Multi-level visualisation using Gaussian process latent variable models. / Mumtaz, Shahzad; Flower, Darren R.; Nabney, Ian.

IVAPP 2014: proceedings of the 5th international conference on Information Visualization Theory and Applications. ed. / Robert S. Laramee; Andreas Kerren; José Braz. Lisbon (PT) : SciTePress, 2014. p. 122-129.

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

TY - GEN

T1 - Multi-level visualisation using Gaussian process latent variable models

AU - Mumtaz, Shahzad

AU - Flower, Darren R.

AU - Nabney, Ian

PY - 2014

Y1 - 2014

N2 - Projection of a high-dimensional dataset onto a two-dimensional space is a useful tool to visualise structures and relationships in the dataset. However, a single two-dimensional visualisation may not display all the intrinsic structure. Therefore, hierarchical/multi-level visualisation methods have been used to extract more detailed understanding of the data. Here we propose a multi-level Gaussian process latent variable model (MLGPLVM). MLGPLVM works by segmenting data (with e.g. K-means, Gaussian mixture model or interactive clustering) in the visualisation space and then fitting a visualisation model to each subset. To measure the quality of multi-level visualisation (with respect to parent and child models), metrics such as trustworthiness, continuity, mean relative rank errors, visualisation distance distortion and the negative log-likelihood per point are used. We evaluate the MLGPLVM approach on the ‘Oil Flow’ dataset and a dataset of protein electrostatic potentials for the ‘Major Histocompatibility Complex (MHC) class I’ of humans. In both cases, visual observation and the quantitative quality measures have shown better visualisation at lower levels.

AB - Projection of a high-dimensional dataset onto a two-dimensional space is a useful tool to visualise structures and relationships in the dataset. However, a single two-dimensional visualisation may not display all the intrinsic structure. Therefore, hierarchical/multi-level visualisation methods have been used to extract more detailed understanding of the data. Here we propose a multi-level Gaussian process latent variable model (MLGPLVM). MLGPLVM works by segmenting data (with e.g. K-means, Gaussian mixture model or interactive clustering) in the visualisation space and then fitting a visualisation model to each subset. To measure the quality of multi-level visualisation (with respect to parent and child models), metrics such as trustworthiness, continuity, mean relative rank errors, visualisation distance distortion and the negative log-likelihood per point are used. We evaluate the MLGPLVM approach on the ‘Oil Flow’ dataset and a dataset of protein electrostatic potentials for the ‘Major Histocompatibility Complex (MHC) class I’ of humans. In both cases, visual observation and the quantitative quality measures have shown better visualisation at lower levels.

KW - continuity

KW - Gaussian mixture model

KW - K-means

KW - major histocompatibility complex

KW - mean relative rank errors

KW - multi-level Gaussian process latent variable model

KW - negative log-likelihood

KW - trustworthiness

KW - visualisation distance distortion

UR - http://www.scopus.com/inward/record.url?scp=84907388107&partnerID=8YFLogxK

M3 - Conference contribution

SN - 978-989-758-005-5

SP - 122

EP - 129

BT - IVAPP 2014

A2 - Laramee, Robert S.

A2 - Kerren, Andreas

A2 - Braz, José

PB - SciTePress

CY - Lisbon (PT)

ER -

Mumtaz S, Flower DR, Nabney I. Multi-level visualisation using Gaussian process latent variable models. In Laramee RS, Kerren A, Braz J, editors, IVAPP 2014: proceedings of the 5th international conference on Information Visualization Theory and Applications. Lisbon (PT): SciTePress. 2014. p. 122-129