Efficient Domain Decomposition for a Neural Network Learning Algorithm, used for the Dose Evaluation in External Radiotherapy

Marc Sauget; Remy Laurent; Julien Henriet; Michel Salomon; Regine Gschwind; Sylvain Contassot-Vivier; Libor Makovicka; Charles Soussen

hal-00520154, version 1

Efficient Domain Decomposition for a Neural Network Learning Algorithm, used for the Dose Evaluation in External Radiotherapy

Marc Sauget () ¹, Remy Laurent ¹, Julien Henriet () ¹, Michel Salomon ², Regine Gschwind () ¹, Sylvain Contassot-Vivier () ³, Libor Makovicka () ¹, Charles Soussen () ⁴

Artificial Neural Networks - ICANN 2010 20th International Conference Proceedings Springer-Heidelberg (Ed.) (2010) 261-266

Résumé : The purpose of this work is to further study the relevance of accelerating the Monte Carlo calculations for the gamma rays external radiotherapy through feed-forward neural networks. We have previously presented a parallel incremental algorithm that builds neural networks of reduced size, while providing high quality approximations of the dose deposit. Our parallel algorithm consists in a regular decomposition of the initial learning dataset (also called learning domain) in as much subsets as available processors. However, the initial learning set presents heterogeneous signal complexities and consequently, the learning times of regular subsets are very different. This paper presents an efficient learning domain decomposition which balances the signal complexities across the processors. As will be shown, the resulting irregular decomposition allows for important gains in learning time of the global network.

1 : Franche-Comté Électronique Mécanique, Thermique et Optique - Sciences et Technologies (FEMTO-ST)
CNRS : UMR6174 – Université de Franche-Comté – Université de Technologie de Belfort-Montbeliard – Ecole Nationale Supérieure de Mécanique et des Microtechniques
2 : Laboratoire d'Informatique de Franche-Comté (LIFC)
Université de Franche-Comté : EA4269
3 : ALGORILLE (INRIA Lorraine - LORIA)
INRIA – CNRS : UMR7503 – Université Henri Poincaré - Nancy I – Université Nancy II – Institut National Polytechnique de Lorraine (INPL)
4 : Centre de recherche en automatique de Nancy (CRAN)
CNRS : UMR7039 – Université Henri Poincaré - Nancy I – Institut National Polytechnique de Lorraine (INPL)

Collaboration : CIPREX 2009-2011

hal-00520154, version 1
http://hal.archives-ouvertes.fr/hal-00520154
oai:hal.archives-ouvertes.fr:hal-00520154
Contributeur : Sylvain Contassot-Vivier <>
Soumis le : Mercredi 22 Septembre 2010, 13:12:59
Dernière modification le : Vendredi 24 Décembre 2010, 12:23:44

Voir la fiche détaillée

Documents associés

DOI : 10.1007/978-3-642-15819-3_34

Exporter

Bibtex EndNote TEI RefWorks

%% hal-00520154, version 1
%% http://hal.archives-ouvertes.fr/hal-00520154
@incollection{sauget:hal-00520154,
    hal_id = {hal-00520154},
    url = {http://hal.archives-ouvertes.fr/hal-00520154},
    title = {{Efficient Domain Decomposition for a Neural Network Learning Algorithm, used for the Dose Evaluation in External Radiotherapy}},
    author = {Sauget, Marc and Laurent, Remy and Henriet, Julien and Salomon, Michel and Gschwind, Regine and Contassot-Vivier, Sylvain and Makovicka, Libor and Soussen, Charles},
    abstract = {{The purpose of this work is to further study the relevance of accelerating the Monte Carlo calculations for the gamma rays external radiotherapy through feed-forward neural networks. We have previously presented a parallel incremental algorithm that builds neural networks of reduced size, while providing high quality approximations of the dose deposit. Our parallel algorithm consists in a regular decomposition of the initial learning dataset (also called learning domain) in as much subsets as available processors. However, the initial learning set presents heterogeneous signal complexities and consequently, the learning times of regular subsets are very different. This paper presents an efficient learning domain decomposition which balances the signal complexities across the processors. As will be shown, the resulting irregular decomposition allows for important gains in learning time of the global network.}},
    language = {Anglais},
    affiliation = {Franche-Comt{\'e} {\'E}lectronique M{\'e}canique, Thermique et Optique - Sciences et Technologies - FEMTO-ST , Laboratoire d'Informatique de Franche-Comt{\'e} - LIFC , ALGORILLE - INRIA Lorraine - LORIA , Centre de recherche en automatique de Nancy - CRAN},
    booktitle = {{Artificial Neural Networks - ICANN 2010 20th International Conference Proceedings}},
    publisher = {Springer-Heidelberg},
    pages = {261-266},
    editor = {Konstantinos Diamantaras and Wlodek Duch and Lazaros Iliadis },
    series = {Lecture Notes in Computer Science, Volume 6352 },
    note = {ISBN-978-3-642-15818-6 },
    collaboration = {CIPREX 2009-2011 },
    doi = {10.1007/978-3-642-15819-3\_34 },
    year = {2010},
}

<?xml version="1.0" encoding="UTF-8"?>
<listBibl xmlns="http://www.tei-c.org/ns/1.0" xmlns:xlink="http://www.w3.org/1999/xlink" xmlns:mml="http://www.w3.org/1998/Math/MathML">
<listBibl type="inbook"><head>Chapitres d'ouvrages scientifiques</head>
<biblStruct type="inbook" xml:id="hal-00520154"><analytic>
<author><persName><forename>Marc</forename><surname>Sauget</surname></persName>
<affiliation>
<orgName type="laboratory">FEMTO-ST</orgName>
<orgName type="team">Franche-Comté Électronique Mécanique, Thermique et Optique - Sciences et Technologies</orgName>
<country>FR</country>
<orgName type="institution">CNRS : UMR6174</orgName>
<orgName type="institution">Université de Franche-Comté</orgName>
<orgName type="institution">Université de Technologie de Belfort-Montbeliard</orgName>
<orgName type="institution">Ecole Nationale Supérieure de Mécanique et des Microtechniques</orgName>
</affiliation>
</author>
<author><persName><forename>Remy</forename><surname>Laurent</surname></persName>
<affiliation>
<orgName type="laboratory">FEMTO-ST</orgName>
<orgName type="team">Franche-Comté Électronique Mécanique, Thermique et Optique - Sciences et Technologies</orgName>
<country>FR</country>
<orgName type="institution">CNRS : UMR6174</orgName>
<orgName type="institution">Université de Franche-Comté</orgName>
<orgName type="institution">Université de Technologie de Belfort-Montbeliard</orgName>
<orgName type="institution">Ecole Nationale Supérieure de Mécanique et des Microtechniques</orgName>
</affiliation>
</author>
<author><persName><forename>Julien</forename><surname>Henriet</surname></persName>
<affiliation>
<orgName type="laboratory">FEMTO-ST</orgName>
<orgName type="team">Franche-Comté Électronique Mécanique, Thermique et Optique - Sciences et Technologies</orgName>
<country>FR</country>
<orgName type="institution">CNRS : UMR6174</orgName>
<orgName type="institution">Université de Franche-Comté</orgName>
<orgName type="institution">Université de Technologie de Belfort-Montbeliard</orgName>
<orgName type="institution">Ecole Nationale Supérieure de Mécanique et des Microtechniques</orgName>
</affiliation>
</author>
<author><persName><forename>Michel</forename><surname>Salomon</surname></persName>
<affiliation>
<orgName type="laboratory">LIFC</orgName>
<orgName type="team">Laboratoire d'Informatique de Franche-Comté</orgName>
<country>FR</country>
<orgName type="institution">Université de Franche-Comté : EA4269</orgName>
</affiliation>
</author>
<author><persName><forename>Regine</forename><surname>Gschwind</surname></persName>
<affiliation>
<orgName type="laboratory">FEMTO-ST</orgName>
<orgName type="team">Franche-Comté Électronique Mécanique, Thermique et Optique - Sciences et Technologies</orgName>
<country>FR</country>
<orgName type="institution">CNRS : UMR6174</orgName>
<orgName type="institution">Université de Franche-Comté</orgName>
<orgName type="institution">Université de Technologie de Belfort-Montbeliard</orgName>
<orgName type="institution">Ecole Nationale Supérieure de Mécanique et des Microtechniques</orgName>
</affiliation>
</author>
<author><persName><forename>Sylvain</forename><surname>Contassot-Vivier</surname></persName>
<affiliation>
<orgName type="laboratory">INRIA Lorraine - LORIA</orgName>
<orgName type="team">ALGORILLE</orgName>
<country>FR</country>
<orgName type="institution">INRIA</orgName>
<orgName type="institution">CNRS : UMR7503</orgName>
<orgName type="institution">Université Henri Poincaré - Nancy I</orgName>
<orgName type="institution">Université Nancy II</orgName>
<orgName type="institution">Institut National Polytechnique de Lorraine (INPL)</orgName>
</affiliation>
</author>
<author><persName><forename>Libor</forename><surname>Makovicka</surname></persName>
<affiliation>
<orgName type="laboratory">FEMTO-ST</orgName>
<orgName type="team">Franche-Comté Électronique Mécanique, Thermique et Optique - Sciences et Technologies</orgName>
<country>FR</country>
<orgName type="institution">CNRS : UMR6174</orgName>
<orgName type="institution">Université de Franche-Comté</orgName>
<orgName type="institution">Université de Technologie de Belfort-Montbeliard</orgName>
<orgName type="institution">Ecole Nationale Supérieure de Mécanique et des Microtechniques</orgName>
</affiliation>
</author>
<author><persName><forename>Charles</forename><surname>Soussen</surname></persName>
<affiliation>
<orgName type="laboratory">CRAN</orgName>
<orgName type="team">Centre de recherche en automatique de Nancy</orgName>
<country>FR</country>
<orgName type="institution">CNRS : UMR7039</orgName>
<orgName type="institution">Université Henri Poincaré - Nancy I</orgName>
<orgName type="institution">Institut National Polytechnique de Lorraine (INPL)</orgName>
</affiliation>
</author>
<title type="main" level="a">Efficient Domain Decomposition for a Neural Network Learning Algorithm, used for the Dose Evaluation in External Radiotherapy</title></analytic><monogr><title level="m" type="main">Artificial Neural Networks - ICANN 2010 20th International Conference Proceedings</title><imprint><publisher>Springer-Heidelberg</publisher><biblScope type="publicationDate">2010-00-00</biblScope><biblScope type="pp">261-266</biblScope></imprint></monogr><idno type="doi">10.1007/978-3-642-15819-3_34</idno><idno type="HALid">http://hal.archives-ouvertes.fr/hal-00520154</idno></biblStruct>
</listBibl>
</listBibl>