A scatterometer neural network sensor model with input noise

Dan Cornford, Guillaume Ramage, Ian T. Nabney

Research output: Contribution to journalArticle

Abstract

The ERS-1 satellite carries a scatterometer which measures the amount of radiation scattered back toward the satellite by the ocean's surface. These measurements can be used to infer wind vectors. The implementation of a neural network based forward model which maps wind vectors to radar backscatter is addressed. Input noise cannot be neglected. To account for this noise, a Bayesian framework is adopted. However, Markov Chain Monte Carlo sampling is too computationally expensive. Instead, gradient information is used with a non-linear optimisation algorithm to find the maximum em a posteriori probability values of the unknown variables. The resulting models are shown to compare well with the current operational model when visualised in the target space.
Original languageEnglish
Pages (from-to)13-21
Number of pages9
JournalNeurocomputing
Volume30
Issue number1
DOIs
Publication statusPublished - Jan 2000

Fingerprint

Radar
Markov Chains
Neural Networks (Computer)
Oceans and Seas
Noise
Radiation
Neural networks
Sensors
Satellites
Markov processes
Sampling

Keywords

  • non-linear regression
  • input uncertainty
  • wind retrieval
  • scatterometer

Cite this

Cornford, Dan ; Ramage, Guillaume ; Nabney, Ian T. / A scatterometer neural network sensor model with input noise. In: Neurocomputing. 2000 ; Vol. 30, No. 1. pp. 13-21.
@article{8a7305846c1b4f2b90f4d45c3e74a5c5,
title = "A scatterometer neural network sensor model with input noise",
abstract = "The ERS-1 satellite carries a scatterometer which measures the amount of radiation scattered back toward the satellite by the ocean's surface. These measurements can be used to infer wind vectors. The implementation of a neural network based forward model which maps wind vectors to radar backscatter is addressed. Input noise cannot be neglected. To account for this noise, a Bayesian framework is adopted. However, Markov Chain Monte Carlo sampling is too computationally expensive. Instead, gradient information is used with a non-linear optimisation algorithm to find the maximum em a posteriori probability values of the unknown variables. The resulting models are shown to compare well with the current operational model when visualised in the target space.",
keywords = "non-linear regression, input uncertainty, wind retrieval, scatterometer",
author = "Dan Cornford and Guillaume Ramage and Nabney, {Ian T.}",
year = "2000",
month = "1",
doi = "10.1016/S0925-2312(99)00137-X",
language = "English",
volume = "30",
pages = "13--21",
journal = "Neurocomputing",
issn = "0925-2312",
publisher = "Elsevier",
number = "1",

}

A scatterometer neural network sensor model with input noise. / Cornford, Dan; Ramage, Guillaume; Nabney, Ian T.

In: Neurocomputing, Vol. 30, No. 1, 01.2000, p. 13-21.

Research output: Contribution to journalArticle

TY - JOUR

T1 - A scatterometer neural network sensor model with input noise

AU - Cornford, Dan

AU - Ramage, Guillaume

AU - Nabney, Ian T.

PY - 2000/1

Y1 - 2000/1

N2 - The ERS-1 satellite carries a scatterometer which measures the amount of radiation scattered back toward the satellite by the ocean's surface. These measurements can be used to infer wind vectors. The implementation of a neural network based forward model which maps wind vectors to radar backscatter is addressed. Input noise cannot be neglected. To account for this noise, a Bayesian framework is adopted. However, Markov Chain Monte Carlo sampling is too computationally expensive. Instead, gradient information is used with a non-linear optimisation algorithm to find the maximum em a posteriori probability values of the unknown variables. The resulting models are shown to compare well with the current operational model when visualised in the target space.

AB - The ERS-1 satellite carries a scatterometer which measures the amount of radiation scattered back toward the satellite by the ocean's surface. These measurements can be used to infer wind vectors. The implementation of a neural network based forward model which maps wind vectors to radar backscatter is addressed. Input noise cannot be neglected. To account for this noise, a Bayesian framework is adopted. However, Markov Chain Monte Carlo sampling is too computationally expensive. Instead, gradient information is used with a non-linear optimisation algorithm to find the maximum em a posteriori probability values of the unknown variables. The resulting models are shown to compare well with the current operational model when visualised in the target space.

KW - non-linear regression

KW - input uncertainty

KW - wind retrieval

KW - scatterometer

UR - http://www.scopus.com/inward/record.url?scp=0033991232&partnerID=8YFLogxK

U2 - 10.1016/S0925-2312(99)00137-X

DO - 10.1016/S0925-2312(99)00137-X

M3 - Article

VL - 30

SP - 13

EP - 21

JO - Neurocomputing

JF - Neurocomputing

SN - 0925-2312

IS - 1

ER -