Structured neural network modelling of multi-valued functions for wind retrieval from scatterometer measurements

David J. Evans, Dan Cornford, Ian T. Nabney

Research output: Contribution to journalArticle

Abstract

A conventional neural network approach to regression problems approximates the conditional mean of the output vector. For mappings which are multi-valued this approach breaks down, since the average of two solutions is not necessarily a valid solution. In this article mixture density networks, a principled method to model conditional probability density functions, are applied to retrieving Cartesian wind vector components from satellite scatterometer data. A hybrid mixture density network is implemented to incorporate prior knowledge of the predominantly bimodal function branches. An advantage of a fully probabilistic model is that more sophisticated and principled methods can be used to resolve ambiguities.
Original languageEnglish
Pages (from-to)23-30
Number of pages8
JournalNeurocomputing
Volume30
Issue number1-4
DOIs
Publication statusPublished - Jan 2000

Fingerprint

Neural networks
Statistical Models
Probability density function
Satellites

Bibliographical note

See http://eprints.aston.ac.uk/1412/

Keywords

  • wind vector retrieval
  • ERS-1 satellite
  • probabilistic models
  • mixture density networks
  • neural networks

Cite this

Evans, David J. ; Cornford, Dan ; Nabney, Ian T. / Structured neural network modelling of multi-valued functions for wind retrieval from scatterometer measurements. In: Neurocomputing. 2000 ; Vol. 30, No. 1-4. pp. 23-30.
@article{d737a2abcaa44a6bba0ffa27a75cac98,
title = "Structured neural network modelling of multi-valued functions for wind retrieval from scatterometer measurements",
abstract = "A conventional neural network approach to regression problems approximates the conditional mean of the output vector. For mappings which are multi-valued this approach breaks down, since the average of two solutions is not necessarily a valid solution. In this article mixture density networks, a principled method to model conditional probability density functions, are applied to retrieving Cartesian wind vector components from satellite scatterometer data. A hybrid mixture density network is implemented to incorporate prior knowledge of the predominantly bimodal function branches. An advantage of a fully probabilistic model is that more sophisticated and principled methods can be used to resolve ambiguities.",
keywords = "wind vector retrieval, ERS-1 satellite, probabilistic models, mixture density networks, neural networks",
author = "Evans, {David J.} and Dan Cornford and Nabney, {Ian T.}",
note = "See http://eprints.aston.ac.uk/1412/",
year = "2000",
month = "1",
doi = "10.1016/S0925-2312(99)00138-1",
language = "English",
volume = "30",
pages = "23--30",
journal = "Neurocomputing",
issn = "0925-2312",
publisher = "Elsevier",
number = "1-4",

}

Structured neural network modelling of multi-valued functions for wind retrieval from scatterometer measurements. / Evans, David J.; Cornford, Dan; Nabney, Ian T.

In: Neurocomputing, Vol. 30, No. 1-4, 01.2000, p. 23-30.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Structured neural network modelling of multi-valued functions for wind retrieval from scatterometer measurements

AU - Evans, David J.

AU - Cornford, Dan

AU - Nabney, Ian T.

N1 - See http://eprints.aston.ac.uk/1412/

PY - 2000/1

Y1 - 2000/1

N2 - A conventional neural network approach to regression problems approximates the conditional mean of the output vector. For mappings which are multi-valued this approach breaks down, since the average of two solutions is not necessarily a valid solution. In this article mixture density networks, a principled method to model conditional probability density functions, are applied to retrieving Cartesian wind vector components from satellite scatterometer data. A hybrid mixture density network is implemented to incorporate prior knowledge of the predominantly bimodal function branches. An advantage of a fully probabilistic model is that more sophisticated and principled methods can be used to resolve ambiguities.

AB - A conventional neural network approach to regression problems approximates the conditional mean of the output vector. For mappings which are multi-valued this approach breaks down, since the average of two solutions is not necessarily a valid solution. In this article mixture density networks, a principled method to model conditional probability density functions, are applied to retrieving Cartesian wind vector components from satellite scatterometer data. A hybrid mixture density network is implemented to incorporate prior knowledge of the predominantly bimodal function branches. An advantage of a fully probabilistic model is that more sophisticated and principled methods can be used to resolve ambiguities.

KW - wind vector retrieval

KW - ERS-1 satellite

KW - probabilistic models

KW - mixture density networks

KW - neural networks

UR - https://www.sciencedirect.com/science/article/pii/S0925231299001381?via%3Dihub

UR - http://www.scopus.com/inward/record.url?scp=0033991558&partnerID=8YFLogxK

U2 - 10.1016/S0925-2312(99)00138-1

DO - 10.1016/S0925-2312(99)00138-1

M3 - Article

VL - 30

SP - 23

EP - 30

JO - Neurocomputing

JF - Neurocomputing

SN - 0925-2312

IS - 1-4

ER -