Alvarez Isabelle

Irstea − Lisc
9 avenue Blaise Pascal − CS 20085
63178 Aubière − France
+33 1 44 27 87 38
Contact

ICPEF (Engineer), Researcher at LISC and at LIP6 (Paris VI University)
Chercheur au LISC et au LIP6 (UPMC)
Robustness, viability and resilience in environmental models; geometric methods, decision support.
Etude de la robustesse et de la résilience de modèles de viabilité ; point de vue géométrique vs. point de vue probabiliste en aide à la décision.
http://www-desir.lip6.fr/~ia/

As a researcher at LISC (Irstea, ex-Cemagref), my main present research interest is to study the link between robustness, viability and resilience in environmental models. I mainly use geometric approaches, which are also useful in decision support.

Mes centres d'intérêt en tant que chercheur : étude de la robustesse et de la résilience de modèles de viabilité, extension au cas multi-agent ; point de vue géométrique vs. point de vue probabiliste en aide à la décision.

As an engineer, I try to promote the use of models for Agriculture and Environment
Mon objectif en tant qu'ingénieur est de promovoir l'utilisation des modèles pour l'agriculture et l'environnement.

Please visit my website (Cette page est en construction, visitez ma page) : http://www-desir.lip6.fr/~ia/

Future official site(Le futur lien sur le site d'Irstea) : www.irstea.fr/alvarez

Publications

Journal Articles (11)

J.B. Rouquier, I. Alvarez, R. Reuillon, P.H. Wuillemin
A kd-tree algorithm to discover the boundary of a black box hypervolume
Annals of Mathematics and Artificial Intelligence 75 (3), 2015
abs doi web bib

n the framework of Decision Support Systems, mathematical viability theory can be used to classify the states and the trajectories of a dynamical system evolving in a set of desirable states. Since obtaining this viability theory output is a complex and computationally intensive task, we propose in this article to consider a compact representation of this set and its approximations using kd-trees. Given a subset of $R^n$ of non null measure, defined through a black box function (an oracle), and assuming some regularity properties on this set, we build a kd-tree based data structure representing this set, which can be used as an input to the viability algorithm. This data structure has a complexity close to gaining one dimension, both in terms of space and in number of calls to the oracle, compared to the exhaustive computation on a regular grid. This data structure produces a characteristic function (i.e. a function that can be used in lieu of the oracle), allows to measure the volume of the set, and to compute the distance to the boundary of the set for any point. It offers distance guarantee between the points of the original set and its kd-tree approximation.
close

@article{PUB00046791, author = {Rouquier, J.B. and Alvarez, I. and Reuillon, R. and Wuillemin, P.H.}, title = {A kd-tree algorithm to discover the boundary of a black box hypervolume}, title = {Un algorithme basé sur les kd-tree pour découvrir la frontière d'un hypervolume boîte-noire}, journal = {Annals of Mathematics and Artificial Intelligence}, volume = {75}, number = {3}, pages = {335-350}, url = {http://cemadoc.irstea.fr/cemoa/PUB00046791}, doi = {http://dx.doi.org/10.1007/s10472-015-9456-8}, year = {2015}, affiliation = {INSTITUT DES SYSTEMES COMPLEXES PARIS FRA ; IRSTEA CLERMONT FERRAND UR LISC FRA ; INSTITUT DES SYSTEMES COMPLEXES PARIS FRA ; UNIVERSITE DE PARIS VI UMR 7606 LIP6 LABORATOIRE D'INFORMATIQUE DE PARIS FRA}, abstract = {n the framework of Decision Support Systems, mathematical viability theory can be used to classify the states and the trajectories of a dynamical system evolving in a set of desirable states. Since obtaining this viability theory output is a complex and computationally intensive task, we propose in this article to consider a compact representation of this set and its approximations using kd-trees. Given a subset of $R^n$ of non null measure, defined through a black box function (an oracle), and assuming some regularity properties on this set, we build a kd-tree based data structure representing this set, which can be used as an input to the viability algorithm. This data structure has a complexity close to gaining one dimension, both in terms of space and in number of calls to the oracle, compared to the exhaustive computation on a regular grid. This data structure produces a characteristic function (i.e. a function that can be used in lieu of the oracle), allows to measure the volume of the set, and to compute the distance to the boundary of the set for any point. It offers distance guarantee between the points of the original set and its kd-tree approximation.}, }
close
S. Martin, I. Alvarez, J.D. Kant
Micro/macro viability analysis of individual-based models: Investigation into the viability of a stylized agricultural cooperative
Complexity 21 (2), 2014
abs doi web bib

The mathematical viability theory proposes methods and tools to study at a global level how controlled dynamical systems can be confined in a desirable subset of the state space. Multilevel viability problems are rarely studied since they induce combinatorial explosion (the set of N agents each evolving in a p-dimensional state space, can evolve in a Np dimensional state space). In this article, we propose an original approach which consists in solving first local viability problems and then studying the real viability of the combination of the local strategies, by simulation where necessary. In this article, we consider as multilevel viability problem a stylized agricultural cooperative which has to keep a minimum of members. Members have an economical constraint and some members have a simple model of the functioning of the cooperative and make assumptions on other members' behavior, especially proviable agents which are concerned about their own viability. In this framework, the model assumptions allow us to solve the local viability problem at the agent level. At the cooperative level, considering mixture of agents, simulation results indicate if and when including proviable agents increases the viability of the whole cooperative.
close

@article{PUB00044373, author = {Martin, S. and Alvarez, I. and Kant, J.D.}, title = {Micro/macro viability analysis of individual-based models: Investigation into the viability of a stylized agricultural cooperative}, title = {Analyse de viabilité micro/macro de modèles individus-centrés : étude de la viabilité d'un modèle stylisé de coopérative}, journal = {Complexity}, volume = {21}, number = {2}, url = {http://onlinelibrary.wiley.com/doi/10.1002/cplx.21604/full}, cemadoc = {http://cemadoc.irstea.fr/cemoa/PUB00044373}, doi = {http://dx.doi.org/10.1002/cplx.21604}, year = {2014}, affiliation = {IRSTEA CLERMONT FERRAND UR LISC FRA ; IRSTEA CLERMONT FERRAND UR LISC FRA ; UNIVERSITE PARIS MARIE CURIE LIP6 FRA}, abstract = {The mathematical viability theory proposes methods and tools to study at a global level how controlled dynamical systems can be confined in a desirable subset of the state space. Multilevel viability problems are rarely studied since they induce combinatorial explosion (the set of N agents each evolving in a p-dimensional state space, can evolve in a Np dimensional state space). In this article, we propose an original approach which consists in solving first local viability problems and then studying the real viability of the combination of the local strategies, by simulation where necessary. In this article, we consider as multilevel viability problem a stylized agricultural cooperative which has to keep a minimum of members. Members have an economical constraint and some members have a simple model of the functioning of the cooperative and make assumptions on other members' behavior, especially proviable agents which are concerned about their own viability. In this framework, the model assumptions allow us to solve the local viability problem at the agent level. At the cooperative level, considering mixture of agents, simulation results indicate if and when including proviable agents increases the viability of the whole cooperative.}, }
close
S. Mesmoudi, I. Alvarez, S. Martin, R. Reuillon, M. Sicard, N. Perrot
Coupling geometric analysis and viability theory for system exploration: Application to a living food system
Journal of Process Control 24 (12), 2014
abs pdf doi web bib

This paper addresses the issue of studying a food complex system in a reverse engineering manner with the aim of identifying the set of all possible actions that makes it reach a quality target with respect to manufacturing constraints. Once the set of actions is identified, several criteria can be considered to identify interesting trajectories and control policies. A viability approach, coupling the viability theory and a geometric approach of robustness, is proposed to study complex dynamical systems. It can be implemented for several types of systems, from linear to non linear or hybrid systems. The proposed framework was adapted to a living food system: a ripening model of Camembert cheese to identify the set of states and actions (capture basin) from which it is possible to reach a predefined quality target. Within the set of viable trajectories, particular trajectories that improve the Camembert cheese ripening process are identified using the proposed approach. The results are applied at a pilot scale and are discussed in this paper.
close

@article{PUB00043300, author = {Mesmoudi, S. and Alvarez, I. and Martin, S. and Reuillon, R. and Sicard, M. and Perrot, N.}, title = {Coupling geometric analysis and viability theory for system exploration: Application to a living food system}, title = {Coupler l'analyse géométrique et la théorie de la viabilité pour explorer les dynamiques des systèmes : application à un système alimentaire}, journal = {Journal of Process Control}, volume = {24}, number = {12}, pages = {18-28}, pdf = {http://motive.cemagref.fr/_publication/PUB00043300.pdf}, url = {http://www.sciencedirect.com/science/article/pii/S0959152414002522}, cemadoc = {http://cemadoc.irstea.fr/cemoa/PUB00043300}, doi = {http://dx.doi.org/10.1016/j.jprocont.2014.09.013}, year = {2014}, affiliation = {ISCPIF LIP6 PARIS FRA ; IRSTEA CLERMONT FERRAND UR LISC FRA ; IRSTEA CLERMONT FERRAND UR LISC FRA ; ISCPIF PARIS FRA ; INRA UMR 782 GMPA AGROPARISTECH THIVERVAL GRIGNON FRA ; INRA UMR 782 GMPA AGROPARISTECH THIVERVAL GRIGNON FRA}, abstract = {This paper addresses the issue of studying a food complex system in a reverse engineering manner with the aim of identifying the set of all possible actions that makes it reach a quality target with respect to manufacturing constraints. Once the set of actions is identified, several criteria can be considered to identify interesting trajectories and control policies. A viability approach, coupling the viability theory and a geometric approach of robustness, is proposed to study complex dynamical systems. It can be implemented for several types of systems, from linear to non linear or hybrid systems. The proposed framework was adapted to a living food system: a ripening model of Camembert cheese to identify the set of states and actions (capture basin) from which it is possible to reach a predefined quality target. Within the set of viable trajectories, particular trajectories that improve the Camembert cheese ripening process are identified using the proposed approach. The results are applied at a pilot scale and are discussed in this paper.}, }
close
W. Wei, S. Martin, I. Alvarez
Sustainability analysis: Viability concepts to consider transient and asymptotical dynamics in socio-ecological tourism-based systems
Ecological Modelling 251, 2013
abs doi web bib

With the increasing pressure on natural resources, the sustainability study of socio-ecological systems has become a crucial scientific issue. In this paper we emphasize that transient behaviors have to be taken into account in sustainability analysis. We illustrate their impact with a model of tourism development from the literature, where the sustainability study is based on asymptotic properties. In order to evaluate relevant space and time characteristics of transient dynamics, we propose to use the concepts and tools of the mathematical viability theory. We also extend this analysis to controlled dynamical systems, which are particularly appropriate to model socio-ecological systems when management issues are at stake.
close

@article{PUB00037806, author = {Wei, W. and Martin, S. and Alvarez, I.}, title = {Sustainability analysis: Viability concepts to consider transient and asymptotical dynamics in socio-ecological tourism-based systems}, title = {Analyse de durabilité : prendre en compte les comportements transitoires et asymptotiques par la théorie de la viabilité. Application à un modèle de tourisme.}, journal = {Ecological Modelling}, volume = {251}, pages = {103-113}, url = {http://cemadoc.irstea.fr/cemoa/PUB00037806}, doi = {http://dx.doi.org/10.1016/j.ecolmodel.2012.10.009}, year = {2013}, affiliation = {IRSTEA CLERMONT FERRAND UR LISC FRA ; IRSTEA CLERMONT FERRAND UR LISC FRA ; IRSTEA CLERMONT FERRAND UR LISC FRA}, abstract = {With the increasing pressure on natural resources, the sustainability study of socio-ecological systems has become a crucial scientific issue. In this paper we emphasize that transient behaviors have to be taken into account in sustainability analysis. We illustrate their impact with a model of tourism development from the literature, where the sustainability study is based on asymptotic properties. In order to evaluate relevant space and time characteristics of transient dynamics, we propose to use the concepts and tools of the mathematical viability theory. We also extend this analysis to controlled dynamical systems, which are particularly appropriate to model socio-ecological systems when management issues are at stake.}, }
close
G. Deffuant, I. Alvarez, O. Barreteau, B. De Vries, B. Edmonds, N. Gilbert, N. Gotts, F. Jabot, S. Janssen, M. Hilden, O. Kolditz, D. Murray-Rust, C. Rougé, P. Smits
Data and models for exploring sustainability of human well-being in global environmental change
European Physical Journal-Special Topics 214 (1), 2012
abs doi web bib

This position paper proposes a vision for the research activity about sustainability in global environmental change (GEC) taking place in the FuturICT flagship project. This activity will be organised in an “Exploratory”, gathering a core network of European scientists from ICT, social simulation, complex systems, economics, demographics, Earth system science. These research teams will collaborate in building a self-organising network of data sources and models about GEC and in using new facilities fostering stakeholder participation. We develop examples of concrete directions for this research: world wide virtual population with demographic and some economic descriptors, ecosystem services production and distribution, governance systems at various scales.
close

@article{PUB00036306, author = {Deffuant, G. and Alvarez, I. and Barreteau, O. and De Vries, B. and Edmonds, B. and Gilbert, N. and Gotts, N. and Jabot, F. and Janssen, S. and Hilden, M. and Kolditz, O. and Murray-Rust, D. and Rougé, C. and Smits, P.}, title = {Data and models for exploring sustainability of human well-being in global environmental change}, title = {Données et modèles pour explorer la durabilité du bien-être humain face aux changements environnementaux planétaires}, journal = {European Physical Journal-Special Topics}, volume = {214}, number = {1}, pages = {519-545}, url = {http://cemadoc.irstea.fr/cemoa/PUB00036306}, doi = {http://dx.doi.org/10.1140/epjst/e2012-01704-2}, year = {2012}, affiliation = {IRSTEA CLERMONT FERRAND UR LISC FRA ; IRSTEA CLERMONT FERRAND UR LISC FRA ; IRSTEA MONTPELLIER UR IRMO FRA ; NETHERLANDS ENVIRONMENTAL AGENCY UTRECHT SUSTAINIBILITY INSTITUTE NLD ; CENTER FOR POLICY MODELLING MANCHESTER GBR ; UNIVERSITY OF SURREY CENTRE FOR RESEARCH IN SOCIAL SIMULATION GBR ; JAMES HUTTON INSTITUTE ABERDEEN GBR ; IRSTEA CLERMONT FERRAND UR LISC FRA ; ALTERRA GREEN WORLD RESEARCH WAGENINGEN NLD ; FINNISH ENVIRONMENT INSTITUTE SYKE HELSINKI FIN ; UFZ HELMHOLTZ CENTRE FOR ENVIRONMENTAL RESEARCH LEIPZIG DEU ; UNIVERSITY OF EDINBURGH GBR ; IRSTEA CLERMONT FERRAND UR LISC FRA ; JRC IES JOINT RESEARCH CENTRE INSTITUTE FOR ENVIRONMENT AND SUSTAINABILITY ISPRA ITA}, abstract = {This position paper proposes a vision for the research activity about sustainability in global environmental change (GEC) taking place in the FuturICT flagship project. This activity will be organised in an “Exploratory”, gathering a core network of European scientists from ICT, social simulation, complex systems, economics, demographics, Earth system science. These research teams will collaborate in building a self-organising network of data sources and models about GEC and in using new facilities fostering stakeholder participation. We develop examples of concrete directions for this research: world wide virtual population with demographic and some economic descriptors, ecosystem services production and distribution, governance systems at various scales.}, }
close
N. Sicard, N. Perrot, R. Reuillon, S. Mesmoudi, I. Alvarez, S. Martin
A viability approach to control food processes: Application to a Camembert cheese ripening process
Food Control 23 (2), 2012
abs doi web bib

This paper addresses the issue of studying the viability theory, developed for model exploration purposes and in our example applied to the optimization of a food operation. The aim is to identify the whole set of viable trajectories for a given process. It focuses on the preservation of some specific properties of the system (constraints in the state space). On the basis of this set, a set of actions is identified and robustness is discussed. The proposed framework was adapted to a Camembert ripening model to identify the subset of the space state where almost one evolution starting in the subset remains indefinitely inside of the domain of some viability constraints, that makes it possible to reach a predefined quality target. The results were applied at the pilot scale and are discussed in this paper. The cheese ripening process was shortened by four days without significant changes in the microorganisms kinetics and a good sensory quality of the cheese. / Cet article s'intéresse à l'utilisation de la théorie de la viabilité (Aubin, 1991) pour l'exploration de modèles complexes, en prenant l'exemple de l'affinage du camembert. La démarche consiste à identifier l'ensemble des trajectoires viables du processus : il s'agit de maintenir certaines propriétés du système (des contraintes dans l'espace d'état, avec ici une cible de qualité à atteindre en fin d'affinage). A partir de cet ensemble, des questions d'optimisation et de robustesse peuvent être discutées. Les résultats ont été testés dans un hâloir-pilote et sont présentés dans cet article. Le processus d'affinage a été raccourci de quatre jours (sur 12) sans changement significatif dans la cinétique des microorganismes et en conservant une bonne qualité sensorielle du fromage.
close

@article{PUB00033451, author = {Sicard, N. and Perrot, N. and Reuillon, R. and Mesmoudi, S. and Alvarez, I. and Martin, S.}, title = {A viability approach to control food processes: Application to a Camembert cheese ripening process}, title = {Une approche basée sur la viabilité pour le contrôle de procédés agro-alimentaires : application à l'affichage du camembert}, journal = {Food Control}, volume = {23}, number = {2}, pages = {312-319}, url = {http://cemadoc.irstea.fr/cemoa/PUB00033451}, doi = {http://dx.doi.org/10.1016/j.foodcont.2011.07.007}, year = {2012}, affiliation = {INRA UMR 782 GMPA AGROPARISTECH THIVERVAL GRIGNON FRA ; INRA UMR 782 GMPA AGROPARISTECH THIVERVAL GRIGNON FRA ; ISC PARIS FRA ; UNIVERSITE DE PARIS VI UMR 7606 LIP6 LABORATOIRE D'INFORMATIQUE DE PARIS ; IRSTEA CLERMONT FERRAND UR LISC FRA ; ISC PARIS FRA}, abstract = {This paper addresses the issue of studying the viability theory, developed for model exploration purposes and in our example applied to the optimization of a food operation. The aim is to identify the whole set of viable trajectories for a given process. It focuses on the preservation of some specific properties of the system (constraints in the state space). On the basis of this set, a set of actions is identified and robustness is discussed. The proposed framework was adapted to a Camembert ripening model to identify the subset of the space state where almost one evolution starting in the subset remains indefinitely inside of the domain of some viability constraints, that makes it possible to reach a predefined quality target. The results were applied at the pilot scale and are discussed in this paper. The cheese ripening process was shortened by four days without significant changes in the microorganisms kinetics and a good sensory quality of the cheese. / Cet article s'intéresse à l'utilisation de la théorie de la viabilité (Aubin, 1991) pour l'exploration de modèles complexes, en prenant l'exemple de l'affinage du camembert. La démarche consiste à identifier l'ensemble des trajectoires viables du processus : il s'agit de maintenir certaines propriétés du système (des contraintes dans l'espace d'état, avec ici une cible de qualité à atteindre en fin d'affinage). A partir de cet ensemble, des questions d'optimisation et de robustesse peuvent être discutées. Les résultats ont été testés dans un hâloir-pilote et sont présentés dans cet article. Le processus d'affinage a été raccourci de quatre jours (sur 12) sans changement significatif dans la cinétique des microorganismes et en conservant une bonne qualité sensorielle du fromage.}, }
close
A. Sordoni, J.P. Briot, I. Alvarez, E. Vasconcelos, M.A. Irving, G. Melo
Design of a participatory decision making agent architecture based on argumentation and influence function application to a serious game about biodiversity conservation
RAIRO-Operations Research 44 (4), 2010
abs doi web bib

This paper addresses an ongoing experience in the design of an artificial agent taking decisions and combining them with the decisions taken by human agents. The context is a serious game research project, aimed at computer-based support for participatory management of protected areas (and more specifically national parks) in order to promote biodiversity conservation and social inclusion. Its objective is to help various stakeholders (e.g., environmentalist, tourism operator) to collectively understand conflict dynamics and explore negotiation strategies for the management of parks. In this paper, after introducing the design of our serious game, named SimParc, we will describe the architecture of the decision making agent playing the role of the park manager. In the game, the park manager makes final decisions based on its own analysis and also on the votes of the stakeholders. It includes two modules: 1) individual decision based on a model of argumentation, which also provides a basis to justify and explain the decision; 2) participatory decision to take into account the preferences/votes from the stakeholders. / Cet article présente notre expérience de conception d'un agent artificiel prenant des décisions et capable de prendre également en compte les décisions d'agents humains. Le contexte est un projet de recherche de type jeu sérieux, pour la gestion participative d'espaces protégés (et plus particulièrement de parcs nationaux), de manière à promouvoir conservation de la biodiversité et inclusion sociale. Son objectif est d'aider les différents acteurs sociaux (« stakeholders », ex : environnementaliste, opérateur de tourisme) à comprendre la dynamique collective des conflits et d'explorer des stratégies de négociation pour la gestion de parcs. Dans cet article, après avoir introduit la conception du jeu sérieux, nommé SimParc (pour « simulation de gestion participative de parcs »), nous décrivons l'architecture de l'agent décideur jouant le rôle d'un gestionnaire de parc. Dans le jeu, le gestionnaire du parc prend la décision finale en fonction de sa propre analyse de la situation ainsi que des propositions des acteurs sociaux. Il inclut deux modules : (1) décision individuelle basé sur un modèle d'argumentation, ce qui offre également un point de départ pour justifier et expliquer la décision ; (2) décision participative pour prendre en compte les préférences/votes des acteurs sociaux.
close

@article{PUB00030854, author = {Sordoni, A. and Briot, J.P. and Alvarez, I. and Vasconcelos, E. and Irving, M.A. and Melo, G.}, title = {Design of a participatory decision making agent architecture based on argumentation and influence function application to a serious game about biodiversity conservation}, title = {Conception d'une architecture d'agent de décision participative, basée sur l'argumentation et les fonctions d'influence : application à un jeu sérieux sur la conservation de la biodiversité}, journal = {RAIRO-Operations Research}, volume = {44}, number = {4}, pages = {269-283}, url = {http://cemadoc.irstea.fr/cemoa/PUB00030854}, doi = {http://dx.doi.org/10.1051/ro/2010024}, year = {2010}, affiliation = {UNIVERSITE DE PARIS VI LIP6 FRA ; UNIVERSITE DE PARIS VI LIP6 FRA ; CEMAGREF CLERMONT FERRAND UR LISC FRA ; PONTIFICIA UNVERSIDADE CATOLICA RIO DE JANEIRO BRA ; UNIVERSIDADE FEDERAL DE RIO DE JANEIRO BRA ; UNIVERSIDADE FEDERAL DE RIO DE JANEIRO BRA}, abstract = {This paper addresses an ongoing experience in the design of an artificial agent taking decisions and combining them with the decisions taken by human agents. The context is a serious game research project, aimed at computer-based support for participatory management of protected areas (and more specifically national parks) in order to promote biodiversity conservation and social inclusion. Its objective is to help various stakeholders (e.g., environmentalist, tourism operator) to collectively understand conflict dynamics and explore negotiation strategies for the management of parks. In this paper, after introducing the design of our serious game, named SimParc, we will describe the architecture of the decision making agent playing the role of the park manager. In the game, the park manager makes final decisions based on its own analysis and also on the votes of the stakeholders. It includes two modules: 1) individual decision based on a model of argumentation, which also provides a basis to justify and explain the decision; 2) participatory decision to take into account the preferences/votes from the stakeholders. / Cet article présente notre expérience de conception d'un agent artificiel prenant des décisions et capable de prendre également en compte les décisions d'agents humains. Le contexte est un projet de recherche de type jeu sérieux, pour la gestion participative d'espaces protégés (et plus particulièrement de parcs nationaux), de manière à promouvoir conservation de la biodiversité et inclusion sociale. Son objectif est d'aider les différents acteurs sociaux (« stakeholders », ex : environnementaliste, opérateur de tourisme) à comprendre la dynamique collective des conflits et d'explorer des stratégies de négociation pour la gestion de parcs. Dans cet article, après avoir introduit la conception du jeu sérieux, nommé SimParc (pour « simulation de gestion participative de parcs »), nous décrivons l'architecture de l'agent décideur jouant le rôle d'un gestionnaire de parc. Dans le jeu, le gestionnaire du parc prend la décision finale en fonction de sa propre analyse de la situation ainsi que des propositions des acteurs sociaux. Il inclut deux modules : (1) décision individuelle basé sur un modèle d'argumentation, ce qui offre également un point de départ pour justifier et expliquer la décision ; (2) décision participative pour prendre en compte les préférences/votes des acteurs sociaux. }, }
close
I. Alvarez, S. Huet
Automatic diagnosis of engine of agricultural tractors: the BED experiment
Biosystems Engineering 100 (3), 2008
abs web bib

The test bed for evaluation and diagnosis (BED) is a simple device designed to perform a quick and automatic diagnosis of the engine of agricultural tractors. It compares the present performance of a tractor to the OECD test. Firstly, a test of the tractor is performed which measures the power take-off speed, the engine torque, the fuel consumption and several environmental variables. The performance curves of the tractor are then plotted against their official reference curves. A rule-based system (Scorpio) compares the present performance to the reference, analyses the differences and suggests which parts may be responsible for the lack of performance. A statistical validation of the expert system was carried out using a database of tests collected over 10 years. The output of the system was compared to the expert decision, which was available for more than a thousand records. This validation of the BED was confirmed by an in situ validation protocol conducted by a tractor repairer. The basic principles of the BED and the results of the validation studies are described / Le BED, (banc d'essai et de diagnostic) est un outil d'aide au diagnostic des moteurs de tracteurs. Il compare les performances actuelles d'un tracteur aux performances du test de référence OCDE. En premier lieu il permet d'effectuer un test du tracteur, en mesurant le couple à la prise de force, la consommation de carburant et des variables d'environnement en fonction de la vitesse à la prise de force. Les courbes de performance sont affichées en regard des courbes de référence. Un système à base de règle (Scorpio) compare les performances actuelles à la référence, analyse les différences et suggère les réglages à effectuer. Une validation statistique du système expert a été conduite sur une base de tests couvrant plus de 10 ans d'expérimentation, en comparant les propositions du système de diagnostic aux conclusion d'un expert pour plus de mille cas. Cette validation a été confirmée sur le terrain par un réparateur. Les principes de base du BED ainsi que les résultats des études de validation sont décrites dans cet article.
close

@article{PUB00025620, author = {Alvarez, I. and Huet, S.}, title = {Automatic diagnosis of engine of agricultural tractors: the BED experiment}, title = {Diagnostic automatique des moteurs de tracteurs agricoles : l'expérience du BED}, journal = {Biosystems Engineering}, volume = {100}, number = {3}, pages = {362-369}, url = {http://dx.doi.org/10.1016/j.biosystemseng.2008.04.003}, cemadoc = {http://cemadoc.irstea.fr/cemoa/PUB00025620}, year = {2008}, affiliation = {CEMAGREF CLERMONT FERRAND LISC ; CEMAGREF CLERMONT FERRAND LISC}, abstract = {The test bed for evaluation and diagnosis (BED) is a simple device designed to perform a quick and automatic diagnosis of the engine of agricultural tractors. It compares the present performance of a tractor to the OECD test. Firstly, a test of the tractor is performed which measures the power take-off speed, the engine torque, the fuel consumption and several environmental variables. The performance curves of the tractor are then plotted against their official reference curves. A rule-based system (Scorpio) compares the present performance to the reference, analyses the differences and suggests which parts may be responsible for the lack of performance. A statistical validation of the expert system was carried out using a database of tests collected over 10 years. The output of the system was compared to the expert decision, which was available for more than a thousand records. This validation of the BED was confirmed by an in situ validation protocol conducted by a tractor repairer. The basic principles of the BED and the results of the validation studies are described / Le BED, (banc d'essai et de diagnostic) est un outil d'aide au diagnostic des moteurs de tracteurs. Il compare les performances actuelles d'un tracteur aux performances du test de référence OCDE. En premier lieu il permet d'effectuer un test du tracteur, en mesurant le couple à la prise de force, la consommation de carburant et des variables d'environnement en fonction de la vitesse à la prise de force. Les courbes de performance sont affichées en regard des courbes de référence. Un système à base de règle (Scorpio) compare les performances actuelles à la référence, analyse les différences et suggère les réglages à effectuer. Une validation statistique du système expert a été conduite sur une base de tests couvrant plus de 10 ans d'expérimentation, en comparant les propositions du système de diagnostic aux conclusion d'un expert pour plus de mille cas. Cette validation a été confirmée sur le terrain par un réparateur. Les principes de base du BED ainsi que les résultats des études de validation sont décrites dans cet article.}, }
close
F. Goreaud, I. Alvarez, B. Courbaud, F. De Coligny
Long-Term influence of the spatial structure of an initial state on the dynamics of a forest growth model: a simulation study using the Capsis platform
Simulation 82 (7), 2006
abs web bib

Spatially explicit Individual Based Models are more and more often used in forest modelling, especially because they take into account the influence of the spatial structure on the dynamics. However, they are potentially very sensitive to the initial spatial structure used for a simulation, which can be problematic if the initial state is not known, or is simulated in an unrealistic way. The aim of our paper is to study this sensitivity to initial spatial structure in the case of the Mountain model, an individual based model of irregular spruce stands implemented in the Capsis platform. In order to characterise the influence of the initial spatial structure on the dynamics of the model, we simulated different initial spatial structures and compared the results of long term simulations. We showed that the initial spatial structure can highly influence the dynamics of the model, not only during the first cycle of the evolution, but also at very long term in the evolution of the next generations. We also illustrated how some disturbances, such as a periodic gap opening through storms, can modify both the long term dynamics of the stand and the duration of the influence of the initial spatial structure. / Les modèles individus centrés et spatialement explicites sont de plus en plus souvent utilisés en modélisation forestière, en particulier parce qu'ils prennent en compte l'influence de la structure spatiale sur la dynamique. Cependant, ils peuvent être très sensibles à la structure spatiale initiale utilisée dans les simulations, ce qui peut poser problème si on ne connait pas l'état initial, ou si il n'est pas simulé de façon réaliste. L'objectif de cet article est d'étudier cette sensibilité à la structure spatiale initiale dans le cas du modèle "Mountain", un modèle individus centré de pessières irrégulières de montagne implémenté dans la plate-forme CAPSIS. Pour caractériser l'influence de la structure spatiale initiale sur la dynamique du modèle, nous avons simulé des états initiaux de différentes structures spatiales, et comparé les résultats de simulations à long terme. Nous avons montré que la structure spatiale initiale peut influencer fortement la dynamique du modèle, non seulement pendant le premier cycle de l'évolution, mais également à très long terme dans les générations suivantes. Nous avons également montré comment des perturbations, comme l'ouverture de clairières par des tempêtes, peut modifier à la fois la dynamique à long terme du modèle, et la durée de l'influence de la structure spatiale initiale.
close

@article{PUB00020509, author = {Goreaud, F. and Alvarez, I. and Courbaud, B. and De Coligny, F.}, title = {Long-Term influence of the spatial structure of an initial state on the dynamics of a forest growth model: a simulation study using the Capsis platform}, title = {Influence à long terme de la structure spatiale d'un état initial sur la dynamique d'un modèle de croissance forestière : une étude simulatoire sur la plate-forme CAPSIS}, journal = {Simulation}, volume = {82}, number = {7}, pages = {475-495}, url = {http://cemadoc.irstea.fr/cemoa/PUB00020509}, year = {2006}, affiliation = {CEMAGREF CLERMONT FERRAND LISC ; CEMAGREF CLERMONT FERRAND LISC ; CEMAGREF GRENOBLE EPGR ; INRA MONTPELLIER}, number = {K LI06 1 ; 06/0072}, abstract = {Spatially explicit Individual Based Models are more and more often used in forest modelling, especially because they take into account the influence of the spatial structure on the dynamics. However, they are potentially very sensitive to the initial spatial structure used for a simulation, which can be problematic if the initial state is not known, or is simulated in an unrealistic way. The aim of our paper is to study this sensitivity to initial spatial structure in the case of the Mountain model, an individual based model of irregular spruce stands implemented in the Capsis platform. In order to characterise the influence of the initial spatial structure on the dynamics of the model, we simulated different initial spatial structures and compared the results of long term simulations. We showed that the initial spatial structure can highly influence the dynamics of the model, not only during the first cycle of the evolution, but also at very long term in the evolution of the next generations. We also illustrated how some disturbances, such as a periodic gap opening through storms, can modify both the long term dynamics of the stand and the duration of the influence of the initial spatial structure. / Les modèles individus centrés et spatialement explicites sont de plus en plus souvent utilisés en modélisation forestière, en particulier parce qu'ils prennent en compte l'influence de la structure spatiale sur la dynamique. Cependant, ils peuvent être très sensibles à la structure spatiale initiale utilisée dans les simulations, ce qui peut poser problème si on ne connait pas l'état initial, ou si il n'est pas simulé de façon réaliste. L'objectif de cet article est d'étudier cette sensibilité à la structure spatiale initiale dans le cas du modèle "Mountain", un modèle individus centré de pessières irrégulières de montagne implémenté dans la plate-forme CAPSIS. Pour caractériser l'influence de la structure spatiale initiale sur la dynamique du modèle, nous avons simulé des états initiaux de différentes structures spatiales, et comparé les résultats de simulations à long terme. Nous avons montré que la structure spatiale initiale peut influencer fortement la dynamique du modèle, non seulement pendant le premier cycle de l'évolution, mais également à très long terme dans les générations suivantes. Nous avons également montré comment des perturbations, comme l'ouverture de clairières par des tempêtes, peut modifier à la fois la dynamique à long terme du modèle, et la durée de l'influence de la structure spatiale initiale.}, }
close
I. Alvarez
Explication géométrique du résultat dans les arbres de décision
Revue d'intelligence artificielle 18 (3), 2004
abs web bib

This article is devoted to the problem of the explanation of the result given by a decision tree (DT) when it is used as a decision aid system, to classify input data and to provide the output class as a result. At now the end-user can rely mainly on some estimation of the error-rate or on the trace of the classification, that is the path run through the DT. We propose here two new methods to qualify the result given by the DT for each particular case. These methods are based on a geometric study of the inverse image of the different classes (the decision surface). First we identify the tests in the tree that are the most relevant to explain the final class, by the way of sensitivity analysis: the vector of initial data is projected onto the decision surface. Second, we study the distribution of errors in relation to the decision surface: the distance to this surface is computed for a sample. The distance of a new case is then presented to the end-user in this context. / Cet article s'intéresse au problème de l'explication des résultats fournis par un arbre de décision utilisé en tant que système d'aide à la décision. On cherche à apporter une information supplémentaire à la classe prédite pour chaque vecteur particulier de données d'entrée. Actuellement on dispose surtout de la trace du classement (le chemin parcouru dans l'arbre), et d'une estimation du taux d'erreur ou d'un risque associé à un mauvais classement. Nous proposons ici deux nouvelles méthodes de qualification du résultat, basées sur une étude géométrique de la frontière de l'image inverse des différentes classes (la surface de décision). La première méthode consiste à identifier les séparateurs les plus déterminants pour expliquer le résultat, en effectuant une analyse de sensibilité, par projection des données initiales sur la surface de décision. La deuxième méthode consiste à caractériser la distribution des erreurs relativement à la surface de décision, en calculant la distance à cette surface pour un ensemble de données, et à replacer chaque nouveau cas dans ce contexte.
close

@article{PUB00015174, author = {Alvarez, I.}, title = {Explication géométrique du résultat dans les arbres de décision}, journal = {Revue d'intelligence artificielle}, volume = {18}, number = {3}, pages = {411-441}, url = {http://cemadoc.irstea.fr/cemoa/PUB00015174}, year = {2004}, affiliation = {CEMAGREF CLERMONT FERRAND LISC}, number = {K LI04 4}, abstract = {This article is devoted to the problem of the explanation of the result given by a decision tree (DT) when it is used as a decision aid system, to classify input data and to provide the output class as a result. At now the end-user can rely mainly on some estimation of the error-rate or on the trace of the classification, that is the path run through the DT. We propose here two new methods to qualify the result given by the DT for each particular case. These methods are based on a geometric study of the inverse image of the different classes (the decision surface). First we identify the tests in the tree that are the most relevant to explain the final class, by the way of sensitivity analysis: the vector of initial data is projected onto the decision surface. Second, we study the distribution of errors in relation to the decision surface: the distance to this surface is computed for a sample. The distance of a new case is then presented to the end-user in this context. / Cet article s'intéresse au problème de l'explication des résultats fournis par un arbre de décision utilisé en tant que système d'aide à la décision. On cherche à apporter une information supplémentaire à la classe prédite pour chaque vecteur particulier de données d'entrée. Actuellement on dispose surtout de la trace du classement (le chemin parcouru dans l'arbre), et d'une estimation du taux d'erreur ou d'un risque associé à un mauvais classement. Nous proposons ici deux nouvelles méthodes de qualification du résultat, basées sur une étude géométrique de la frontière de l'image inverse des différentes classes (la surface de décision). La première méthode consiste à identifier les séparateurs les plus déterminants pour expliquer le résultat, en effectuant une analyse de sensibilité, par projection des données initiales sur la surface de décision. La deuxième méthode consiste à caractériser la distribution des erreurs relativement à la surface de décision, en calculant la distance à cette surface pour un ensemble de données, et à replacer chaque nouveau cas dans ce contexte.}, }
close
P. Grenier, I. Alvarez, J.M. Roger, V. Steinmetz, P. Barre, J.M. Sablayrolles
Artificial intelligence in wine-making
Journal International des Sciences de la Vigne et du Vin 34 (2), 2000
abs web bib

In this paper, some terms of Artificial Intelligence are defined. Some present and potential applications of knowledge based systems are presented in the field of wine-making. Areas of concern were: multi sensor fusion, prediction by model cooperation, and diagnosis. Artificial intelligence techniques can indeed be applied for aiding the wine-maker in his choices. They facilitate the combination between experience and recent progress in technology. When associated with statistical processing, they allow knowledge sources to be used more effectively. Beyond wine-making, the prospects of artificial intelligence are promising for research and food industry, especially for improving the robustness of measurement systems (multi-sensors, sensors interpreted or validated by models), and for process diagnosis (risk prediction, action proposal). / Certains termes d'Intelligence Artificielle (IA) sont définis dans cette publication. Quelques applications en cours ou potentielles de systèmes fondés sur la connaissance sont présentés dans le domaine de la vinification. Les domaines d'étude sont : la fusion multi-capteurs, la prédiction par coopération de modèles, et le diagnostic. Les techniques IA peuvent aider le vinificateur dans ses choix par une symbiose entre expérience et progrès technologiques. En association avec des traitements statistiques, ces techniques permettent une utilisation plus efficace des sources de connaissances. Au-delà de la vinification, les perspectives de l'intelligence artificielle sont prometteuses en industrie alimentaire, en particulier pour améliorer la robustesse d'un système de mesure (fusion de capteurs, validation d'un capteur par un modèle) ou pour élaborer un diagnostic sur l'évolution d'un procédé (prédiction de risques, proposition d'intervention).
close

@article{PUB00008877, author = {Grenier, P. and Alvarez, I. and Roger, J.M. and Steinmetz, V. and Barre, P. and Sablayrolles, J.M.}, title = {Artificial intelligence in wine-making}, title = {L'intelligence artificielle en oenologie}, journal = {Journal International des Sciences de la Vigne et du Vin}, volume = {34}, number = {2}, pages = {61-66}, url = {http://cemadoc.irstea.fr/cemoa/PUB00008877}, year = {2000}, affiliation = {CEMAGREF MONTPELLIER GIMO ; INRA IPV MONTPELLIER ; CEMAGREF CLERMONT FERRAND LISC ; CEMAGREF CLERMONT FERRAND LISC ; CEMAGREF CLERMONT FERRAND LISC ; CEMAGREF CLERMONT FERRAND LISC}, number = {00/0664 ; BGI 0826}, abstract = {In this paper, some terms of Artificial Intelligence are defined. Some present and potential applications of knowledge based systems are presented in the field of wine-making. Areas of concern were: multi sensor fusion, prediction by model cooperation, and diagnosis. Artificial intelligence techniques can indeed be applied for aiding the wine-maker in his choices. They facilitate the combination between experience and recent progress in technology. When associated with statistical processing, they allow knowledge sources to be used more effectively. Beyond wine-making, the prospects of artificial intelligence are promising for research and food industry, especially for improving the robustness of measurement systems (multi-sensors, sensors interpreted or validated by models), and for process diagnosis (risk prediction, action proposal). / Certains termes d'Intelligence Artificielle (IA) sont définis dans cette publication. Quelques applications en cours ou potentielles de systèmes fondés sur la connaissance sont présentés dans le domaine de la vinification. Les domaines d'étude sont : la fusion multi-capteurs, la prédiction par coopération de modèles, et le diagnostic. Les techniques IA peuvent aider le vinificateur dans ses choix par une symbiose entre expérience et progrès technologiques. En association avec des traitements statistiques, ces techniques permettent une utilisation plus efficace des sources de connaissances. Au-delà de la vinification, les perspectives de l'intelligence artificielle sont prometteuses en industrie alimentaire, en particulier pour améliorer la robustesse d'un système de mesure (fusion de capteurs, validation d'un capteur par un modèle) ou pour élaborer un diagnostic sur l'évolution d'un procédé (prédiction de risques, proposition d'intervention).}, }
close

Book Chapters (2)

I. Alvarez, S. Martin, O. Dordan, X. Litrico, P. Saint-Pierre
Indicateurs de sécurité et de restauration dynamiques
Agir en situation d'incertitude en agriculture : regards pluridisciplinaires au Nord et au Sud (p. 14), Peter Lang, Bruxelles, BEL, 2013
abs web bib

Les systèmes agricoles au sens large et leur environnement peuvent être décrits comme des systèmes dynamiques : par exemple, des activités agricoles utilisent chaque année des engrais phosphatés. Ces engrais ne se fixent pas dans le sol, mais descendent par infiltration ou écoulement de surface vers les écosystèmes aquatiques – rivières, lacs, baies. Les concentrations excessives de substances nutritives dans ces bassins aquatiques provoquent la prolifération de nappes d’algues. Ces systèmes agricoles sont soumis à des contraintes de rentabilité bien sûr, mais pas uniquement. En effet, les populations riveraines d'un lac souhaitent qu'il conserve un état d'eau claire (oligotrophe), et éviter la prolifération des algues impose des contraintes supplémentaires sur les activités agricoles, en l'occurrence, limiter les apports d'engrais dans le bassin versant. Enfin, ces systèmes agricoles sont soumis à des incertitudes. Viennent immédiatement à l'esprit les aléas climatiques, mais pour rester dans l'exemple du lac, un événement pluvieux intense peut provoquer un ruissellement important qui a pour conséquence immédiate d'augmenter brusquement la concentration de phosphore dissous. Face à des systèmes qui évoluent soumis à des contraintes et des incertitudes, quelles informations fournir au gestionnaire ?
close

@book{PUB00040096, author = {Alvarez, I. and Martin, S. and Dordan, O. and Litrico, X. and Saint-Pierre, P.}, author = {Ancey, V. and Avelange, I. and Dedieu, B.}, title = {Indicateurs de sécurité et de restauration dynamiques}, title = {Security and restoration indicators}, bookTitle = {Agir en situation d'incertitude en agriculture : regards pluridisciplinaires au Nord et au Sud}, series = {EcoPolis - volume 17 }, address = {Bruxelles, BEL}, publisher = {Peter Lang}, pages = {14}, url = {http://cemadoc.irstea.fr/cemoa/PUB00040096}, year = {2013}, affiliation = {IRSTEA CLERMONT FERRAND UR LISC FRA ; IRSTEA CLERMONT FERRAND UR LISC FRA ; UNIVERSITE DE BORDEAUX FRA ; LYONNAISE DES EAUX BORDEAUX FRA ; UNIVERSITE PARIS DAUPHINE PARIS FRA}, abstract = {Les systèmes agricoles au sens large et leur environnement peuvent être décrits comme des systèmes dynamiques : par exemple, des activités agricoles utilisent chaque année des engrais phosphatés. Ces engrais ne se fixent pas dans le sol, mais descendent par infiltration ou écoulement de surface vers les écosystèmes aquatiques – rivières, lacs, baies. Les concentrations excessives de substances nutritives dans ces bassins aquatiques provoquent la prolifération de nappes d’algues. Ces systèmes agricoles sont soumis à des contraintes de rentabilité bien sûr, mais pas uniquement. En effet, les populations riveraines d'un lac souhaitent qu'il conserve un état d'eau claire (oligotrophe), et éviter la prolifération des algues impose des contraintes supplémentaires sur les activités agricoles, en l'occurrence, limiter les apports d'engrais dans le bassin versant. Enfin, ces systèmes agricoles sont soumis à des incertitudes. Viennent immédiatement à l'esprit les aléas climatiques, mais pour rester dans l'exemple du lac, un événement pluvieux intense peut provoquer un ruissellement important qui a pour conséquence immédiate d'augmenter brusquement la concentration de phosphore dissous. Face à des systèmes qui évoluent soumis à des contraintes et des incertitudes, quelles informations fournir au gestionnaire ?}, }
close
I. Alvarez, S. Martin
Geometric robustness of viability kernels and resilience basins
Viability and resilience of complex systems: concepts, methods and case studies from ecology and society (p. 193-218), Springer, 2011
doi web bib

@book{PUB00033189, author = {Alvarez, I. and Martin, S.}, author = {Deffuant, G. and Gilbert, N.}, title = {Geometric robustness of viability kernels and resilience basins}, bookTitle = {Viability and resilience of complex systems: concepts, methods and case studies from ecology and society}, series = {Understanding complex systems}, publisher = {Springer}, pages = {193-218}, url = {http://cemadoc.irstea.fr/cemoa/PUB00033189}, doi = {http://dx.doi.org/10.1007/978-3-642-20423-4}, year = {2011}, affiliation = {CEMAGREF CLERMONT FERRAND UR LISC FRA ; CEMAGREF CLERMONT FERRAND UR LISC FRA}, ISBNISSN = {ISBN-13: 978-3642204227}, abstract = {}, }
close

In Proceedings (19)

I. Alvarez, R. De Aldama, S. Martin, R. Reuillon
Assessing the resilience of socio-ecosystems: coupling viability theory and active learning with kd-Trees. Application to bilingual societies
International Joint Conference on Artificial Intelligence 2013 ( IJCAI), 03/08/2013-09/08/2013, Beijing, CHN (p. 2776-2782), 2013
abs web bib

This paper proposes a new algorithm to compute the resilience of a social system or an ecosystem when it is defined in the framework of the mathematical viability theory. It is applied to the problem of language coexistence: Although bilingual societies do exist, many languages have disappeared and some seem endangered presently. Mathematical models of language competition generally conclude that one language will disappear, except when the relative prestige of the languages can be modified. The viability theory provides concepts and tools that are suitable to study the resilience, but with severe computational limits since it uses extensive search on regular grids. The method we propose considers the computation of the viability output sets as an active learning problem with the objective of restraining the number of calls to the model and information storage. We adapt a kd-tree algorithm to approximate the level sets of the resilience value. We prove that this algorithm converges to the output sets defined by the viability theory (viability kernel and capture basin). The resilience value we compute can then be used to propose a policy of action in risky situations such as migration flows. / Cet article propose un algorithme couplé pour calculer la résilience d'un socio-écosystème dans le cadre de la théorie mathématique de la viabilité, illustré sur le problème des sociétés bilingues. Les modèles mathématiques de telles sociétés montrent que l'extinction d'un langage est en général inéluctable, sauf s'il est possible de modifier le prestige des langues. La méthode de calcul de la résilience que nous proposons repose sur un algorithme adapté des kd-trees. La convergence de l'algorithme de viabilité dans ce cadre est démontrée. Le calcul de la résilience est utilisé pour proposer des politiques d'action en cas de flux migratoires.
close

@inproceedings{PUB00039946, author = {Alvarez, I. and De Aldama, R. and Martin, S. and Reuillon, R.}, title = {Assessing the resilience of socio-ecosystems: coupling viability theory and active learning with kd-Trees. Application to bilingual societies}, title = {Evaluer la résilience des socio-écosystèmes : coupler la théorie de la viabilité avec l'apprentissage actif à partir de kd-trees. Application à la coexistence des langues.}, bookTitle = {International Joint Conference on Artificial Intelligence 2013 ( IJCAI), 03/08/2013-09/08/2013, Beijing, CHN}, pages = {2776-2782}, url = {http://ijcai.org/papers13/Papers/IJCAI13-409.pdf}, cemadoc = {http://cemadoc.irstea.fr/cemoa/PUB00039946}, year = {2013}, affiliation = {UNIVERSITE PIERRE ET MARIE CURIE LIP6 PARIS FRA ; INSTITUT DES SYSTEMES COMPLEXES ISC PIF PARIS FRA ; IRSTEA CLERMONT FERRAND UR LISC FRA ; INSTITUT DES SYSTEMES COMPLEXES ISC PIF PARIS FRA}, abstract = {This paper proposes a new algorithm to compute the resilience of a social system or an ecosystem when it is defined in the framework of the mathematical viability theory. It is applied to the problem of language coexistence: Although bilingual societies do exist, many languages have disappeared and some seem endangered presently. Mathematical models of language competition generally conclude that one language will disappear, except when the relative prestige of the languages can be modified. The viability theory provides concepts and tools that are suitable to study the resilience, but with severe computational limits since it uses extensive search on regular grids. The method we propose considers the computation of the viability output sets as an active learning problem with the objective of restraining the number of calls to the model and information storage. We adapt a kd-tree algorithm to approximate the level sets of the resilience value. We prove that this algorithm converges to the output sets defined by the viability theory (viability kernel and capture basin). The resilience value we compute can then be used to propose a policy of action in risky situations such as migration flows. / Cet article propose un algorithme couplé pour calculer la résilience d'un socio-écosystème dans le cadre de la théorie mathématique de la viabilité, illustré sur le problème des sociétés bilingues. Les modèles mathématiques de telles sociétés montrent que l'extinction d'un langage est en général inéluctable, sauf s'il est possible de modifier le prestige des langues. La méthode de calcul de la résilience que nous proposons repose sur un algorithme adapté des kd-trees. La convergence de l'algorithme de viabilité dans ce cadre est démontrée. Le calcul de la résilience est utilisé pour proposer des politiques d'action en cas de flux migratoires.}, }
close
W. Wei, I. Alvarez, S. Martin, J.P. Briot, M. De Azevedo Irving, G. Mendes de Melo
Integration of viability models in a serious game for the management of protected areas
IADIS International Conference Intelligent Systems and Agents 2012 and IADIS European Conference Data Mining 2012, MCCSIS 2012 , 17/07/2012-23/07/2012, Lisbonne, PRT (p. 8), 2012
abs pdf bib

Participatory management is considered to be a very important challenge, in particular for the management of renewable resources and biodiversity conservation. In the framework of computer support for participatory management of protected areas, we present in this paper a component of a serious game, devoted to the viability study of the decision proposed by the stakeholders. Mathematical viability theory allows to identify the policies that can retain or restore desirable properties of a dynamical system. The objective of the viability expert is to provide the players with information about the efficiency of management policy. In this paper we present briefly the SimParc project, and we introduce the concepts of the mathematical viability theory. We then describe the viability expert agent architecture and possibility of use, in particular the fact that negotiation focuses on constraints rather than on the management decision itself. We also present the type of information given by the expert agent. In particular the fact that negotiation focuses on constraints rather than on the management decision itself.
close

@inproceedings{PUB00037220, author = {Wei, W. and Alvarez, I. and Martin, S. and Briot, J.P. and De Azevedo Irving, M. and Mendes de Melo , G.}, title = {Integration of viability models in a serious game for the management of protected areas}, bookTitle = {IADIS International Conference Intelligent Systems and Agents 2012 and IADIS European Conference Data Mining 2012, MCCSIS 2012 , 17/07/2012-23/07/2012, Lisbonne, PRT}, pages = {8}, pdf = {http://motive.cemagref.fr/_publication/PUB00037220.pdf}, year = {2012}, affiliation = {IRSTEA CLERMONT FERRAND UR LISC FRA ; IRSTEA CLERMONT FERRAND UR LISC FRA ; IRSTEA CLERMONT FERRAND UR LISC FRA ; CNRS LIP6 PARIS FRA ; UFRJ INCT PPED RIO DE JANEIRO BRA ; UFRJ INCT PPED RIO DE JANEIRO BRA}, abstract = {Participatory management is considered to be a very important challenge, in particular for the management of renewable resources and biodiversity conservation. In the framework of computer support for participatory management of protected areas, we present in this paper a component of a serious game, devoted to the viability study of the decision proposed by the stakeholders. Mathematical viability theory allows to identify the policies that can retain or restore desirable properties of a dynamical system. The objective of the viability expert is to provide the players with information about the efficiency of management policy. In this paper we present briefly the SimParc project, and we introduce the concepts of the mathematical viability theory. We then describe the viability expert agent architecture and possibility of use, in particular the fact that negotiation focuses on constraints rather than on the management decision itself. We also present the type of information given by the expert agent. In particular the fact that negotiation focuses on constraints rather than on the management decision itself.}, }
close
W. Wei, I. Alvarez, S. Martin
Algorithme d'approximation du noyau de viabilité avec procédure de classification
RFIA 2012 (Reconnaissance des Formes et Intelligence Artificielle), 24/01/2012-27/01/2012, Lyon, FRA , 2012
abs pdf bib

Viability theory proposes concepts and tools to control a dynamical system such that it can remain inside a viability constraint set. Integrating viability models in a tool support for participatory management is a new application of the viability theory. In this context, we propose an algorithm for viability kernel approximation, which focuses on improving the computing time. This algorithm uses the classification procedure based on the method of Nearest Neighbor Search. The algorithm satisfies the conditions of convergence, and is easy to use with a graphical user interface. We compare its results for models for which the exact viability kernels have been determined. The accuracy of this algorithm is always greater than 90 percent. The computing time is very interesting for decision support tools. / La théorie de la viabilité propose des concepts et méthodes pour contrôler un système dynamique afin de le maintenir dans un ensemble de contraintes de viabilité. Intégrer les modèles de la viabilité dans un outil d’aide à la gestion participative est une nouvelle application de la théorie de la viabilité. Dans ce contexte, nous proposons un algorithme d’approximation du noyau de viabilité qui se concentre sur la réduction du temps de calcul. Cet algorithme utilise une procédure de classification du type "plus proche voisin". L’algorithme satisfait les conditions de convergence, et est facile à manipuler en utilisant une interface graphique utilisateur. Nous comparons les résultats obtenus pour des problèmes de viabilité pour lesquels les noyaux théoriques ont été calculés. La fidélité de cet algorithme est toujours supérieure à 90 pourcents. Le temps de calcul est très intéressant pour les outils d’aide à la décision.
close

@inproceedings{PUB00037217, author = {Wei, W. and Alvarez, I. and Martin, S.}, title = {Algorithme d'approximation du noyau de viabilité avec procédure de classification}, bookTitle = {RFIA 2012 (Reconnaissance des Formes et Intelligence Artificielle), 24/01/2012-27/01/2012, Lyon, FRA}, pdf = {http://motive.cemagref.fr/_publication/PUB00037217.pdf}, year = {2012}, affiliation = {IRSTEA CLERMONT FERRAND UR LISC FRA ; IRSTEA CLERMONT FERRAND UR LISC FRA ; IRSTEA CLERMONT FERRAND UR LISC FRA}, abstract = {Viability theory proposes concepts and tools to control a dynamical system such that it can remain inside a viability constraint set. Integrating viability models in a tool support for participatory management is a new application of the viability theory. In this context, we propose an algorithm for viability kernel approximation, which focuses on improving the computing time. This algorithm uses the classification procedure based on the method of Nearest Neighbor Search. The algorithm satisfies the conditions of convergence, and is easy to use with a graphical user interface. We compare its results for models for which the exact viability kernels have been determined. The accuracy of this algorithm is always greater than 90 percent. The computing time is very interesting for decision support tools. / La théorie de la viabilité propose des concepts et méthodes pour contrôler un système dynamique afin de le maintenir dans un ensemble de contraintes de viabilité. Intégrer les modèles de la viabilité dans un outil d’aide à la gestion participative est une nouvelle application de la théorie de la viabilité. Dans ce contexte, nous proposons un algorithme d’approximation du noyau de viabilité qui se concentre sur la réduction du temps de calcul. Cet algorithme utilise une procédure de classification du type "plus proche voisin". L’algorithme satisfait les conditions de convergence, et est facile à manipuler en utilisant une interface graphique utilisateur. Nous comparons les résultats obtenus pour des problèmes de viabilité pour lesquels les noyaux théoriques ont été calculés. La fidélité de cet algorithme est toujours supérieure à 90 pourcents. Le temps de calcul est très intéressant pour les outils d’aide à la décision. }, }
close
I. Alvarez, C. Bernard, S. Martin, P. Saint-Pierre
Compétition linguistique : stratégies pour favoriser la coexistence durable en présence de bilinguisme
48ème colloque de l'Association de Science Régionale de Langue Française, ASRDLF 2011, 06/07/2011-08/07/2011, Schoelcher, MTQ / CD Rom, Session spéciale Vulnérabilité-Résilience-Viabilité, Actes du 48ème colloque ASRDLF 2011 (p. 11), 2011
abs web bib

Comment assurer durablement la coexistence de plusieurs langues au sein d'une même population? Cette question est au centre des préoccupations des régions ou pays qui cherchent à préserver une culture multilingue considérée comme un patrimoine historique à préserver ou à développer. Malheureusement, même si on constate effectivement l'existence de sociétés dans lesquelles coexistent plusieurs langues, la plupart des modèles montrent que la cohabitation est un état instable. La cohabitation initiale évolue inéluctablement vers l'extinction d'une des langues, en fonction du nombre de locuteurs et du prestige relatif des deux langues en compétition. En revanche, si par des actions politiques, sociales ou d'autre nature il est possible d'influer sur le prestige relatif des langues, une cohabitation prolongée devient théoriquement possible en fonction des scénarios retenus. L'objectif de cet article est de montrer comment la théorie mathématique de la viabilité, et l'analyse géométrique des résultats produits, peut permettre de trouver les stratégies permettant d'assurer la coexistence de groupes monolingues au sein de la population.
close

@inproceedings{PUB00033744, author = {Alvarez, I. and Bernard, C. and Martin, S. and Saint-Pierre, P.}, title = {Compétition linguistique : stratégies pour favoriser la coexistence durable en présence de bilinguisme }, title = {Language competition : strategies to ensure sustainable coexistence in presence of bilingualism}, bookTitle = {48ème colloque de l'Association de Science Régionale de Langue Française, ASRDLF 2011, 06/07/2011-08/07/2011, Schoelcher, MTQ / CD Rom, Session spéciale Vulnérabilité-Résilience-Viabilité, Actes du 48ème colloque ASRDLF 2011}, pages = {11}, url = {http://cemadoc.irstea.fr/cemoa/PUB00033744}, year = {2011}, affiliation = {CEMAGREF CLERMONT FERRAND UR LISC FRA ; CEMAGREF CLERMONT FERRAND UR LISC FRA ; CEMAGREF CLERMONT FERRAND UR LISC FRA ; LABORATOIRE D'APPLICATIONS DES SYSTEMES TYCHASTIQUES REGULES FRA}, abstract = {Comment assurer durablement la coexistence de plusieurs langues au sein d'une même population? Cette question est au centre des préoccupations des régions ou pays qui cherchent à préserver une culture multilingue considérée comme un patrimoine historique à préserver ou à développer. Malheureusement, même si on constate effectivement l'existence de sociétés dans lesquelles coexistent plusieurs langues, la plupart des modèles montrent que la cohabitation est un état instable. La cohabitation initiale évolue inéluctablement vers l'extinction d'une des langues, en fonction du nombre de locuteurs et du prestige relatif des deux langues en compétition. En revanche, si par des actions politiques, sociales ou d'autre nature il est possible d'influer sur le prestige relatif des langues, une cohabitation prolongée devient théoriquement possible en fonction des scénarios retenus. L'objectif de cet article est de montrer comment la théorie mathématique de la viabilité, et l'analyse géométrique des résultats produits, peut permettre de trouver les stratégies permettant d'assurer la coexistence de groupes monolingues au sein de la population.}, }
close
J.P. Briot, M. de Azevedo Irving, G. Mendes de Melo, J.E. Filho Vasconcelos, I. Alvarez, S. Martin, W. Wei
A serious game and artificial agents to support intercultural participatory management of protected areas for biodiversity conservation and social inclusion
Culture and Computing 2011, The Second International Conference on Culture and Computing, Kyoto University, 20/10/2011-22/10/2011, Kyoto, JPN / Culture and Computing, 2011 Second International Conference ; (p. 15-20), Conference Publishing Services (CPS), 2011
abs doi web bib

This paper addresses our experience in the design of a serious game, aimed at computer-based support for intercultural participatory management of protected areas (e.g., parks, marine reserves, biosphere reserves. . . ), in order to promote biodiversity conservation and social inclusion. Its objective is, via a computer assisted role-playing game, to help various stakeholders (e.g., environmentalist, tourism operator, traditional community. . . ) to collectively understand conflict dynamics and explore negotiation strategies for the management of protected areas. Therefore, this helps at mutual understanding and negotiation between different cultures, contexts and practices (traditional community, technical manager, environmentalist. . . ) about strategic issues when aiming at both biodiversity conservation and social inclusion. After introducing the objectives of our serious game, named SimParc, we will describe its design and its current architecture. We will also discuss the introduction of various types of agents in the system: a decision making agent playing the role of the park manager; artificial players replacing some of the human players in the game; assistant agents assisting human players; and expert agents providing human players with technical information about the viability of their proposal (e.g., about the survival of an endangered species), or to analyse relations (e.g., dominance or equity) among players proposals. This last type of agent aims at introducing a technical viewpoint and culture in this intercultural participatory process. Some of these agents have already been implemented and tested and some others are in progress.
close

@inproceedings{PUB00033596, author = {Briot, J.P. and de Azevedo Irving, M. and Mendes de Melo, G. and Filho Vasconcelos, J.E. and Alvarez, I. and Martin, S. and Wei, W.}, title = {A serious game and artificial agents to support intercultural participatory management of protected areas for biodiversity conservation and social inclusion}, title = {Jeu sérieux et agents artificiels comme aide à la gestion participative et interculturelle d'aires protégées pour la conservation de la biodiversité et l'inclusion sociale}, bookTitle = {Culture and Computing 2011, The Second International Conference on Culture and Computing, Kyoto University, 20/10/2011-22/10/2011, Kyoto, JPN / Culture and Computing, 2011 Second International Conference ; }, publisher = {Conference Publishing Services (CPS)}, pages = {15-20}, url = {http://cemadoc.irstea.fr/cemoa/PUB00033596}, doi = {http://dx.doi.org/10.1109/Culture-Computing.2011.12 }, year = {2011}, affiliation = {LIP6 UNIVERSITE PIERRE ET MARIE CURIE CNRS PARIS FRA ; UFRJ RIO DE JANEIRO BRA ; INCT-PPED RIO DE JANEIRO BRA ; UNIFOR NATI FORTALEZA BRA ; CEMAGREF CLERMONT FERRAND UR LISC FRA ; CEMAGREF CLERMONT FERRAND UR LISC FRA ; CEMAGREF CLERMONT FERRAND UR LISC FRA}, ISBNISSN = {ISBN: 978-1-4577-1593-8 }, abstract = {This paper addresses our experience in the design of a serious game, aimed at computer-based support for intercultural participatory management of protected areas (e.g., parks, marine reserves, biosphere reserves. . . ), in order to promote biodiversity conservation and social inclusion. Its objective is, via a computer assisted role-playing game, to help various stakeholders (e.g., environmentalist, tourism operator, traditional community. . . ) to collectively understand conflict dynamics and explore negotiation strategies for the management of protected areas. Therefore, this helps at mutual understanding and negotiation between different cultures, contexts and practices (traditional community, technical manager, environmentalist. . . ) about strategic issues when aiming at both biodiversity conservation and social inclusion. After introducing the objectives of our serious game, named SimParc, we will describe its design and its current architecture. We will also discuss the introduction of various types of agents in the system: a decision making agent playing the role of the park manager; artificial players replacing some of the human players in the game; assistant agents assisting human players; and expert agents providing human players with technical information about the viability of their proposal (e.g., about the survival of an endangered species), or to analyse relations (e.g., dominance or equity) among players proposals. This last type of agent aims at introducing a technical viewpoint and culture in this intercultural participatory process. Some of these agents have already been implemented and tested and some others are in progress.}, }
close
I. Alvarez, S. Martin, S. Mesmoudi
Describing the result of a classifier to the end-user: geometric-based sensitivity
ECAI 2010, 19th European Conference on Artificial Intelligence, 16/08/2010-20/08/2010, Lisbonne, PRT / Frontiers in Artificial Intelligence and Applications.Helder Coelho, Rudi Studer, Michael Wooldridge (Eds.) (p. 835-840), IOS Press, 2010
abs web bib

This paper addresses the issue of supporting the end-user of a classifier, when it is used as a decision support system, to classify new cases. We consider several kinds of classifiers: Statistical or machine learning classifiers, which are built on data, but also direct model-based classifiers that are built to solve a particular problem (like in viability or control problems). The end-user relies mainly on global information (like error rates or global sensitivity analysis) to assess the quality of the result given by the system. Class membership probability, if available, describes certainly the local statistical viewpoint. But it doesn't take into account other contextual information: Cases with high value of class membership probability can also be close to the decision boundary. In the case of numerical state space, we propose to use the decision boundary of the classifier (which always exists, even implicitly), to describe the situation of a particular case: The distance of a case to the decision boundary measures the robustness of the decision to a change in the input data. Other geometric concepts, such as the maximal maximal ball, can present a precise picture of the situation to the end-user. We show the interest of such a geometric study on different examples. / Cet article propose une méthode, basée sur la géométrie, pour aider l'utilisateur d'un système discriminant à apprécier les résultats dans le cadre de l'aide à la décision.
close

@inproceedings{PUB00030835, author = {Alvarez, I. and Martin, S. and Mesmoudi, S.}, title = {Describing the result of a classifier to the end-user: geometric-based sensitivity}, title = {Décrire le résultat d'un système discriminant à l'utilisateur : la sensibilité géométrique}, volume = {215}, bookTitle = {ECAI 2010, 19th European Conference on Artificial Intelligence, 16/08/2010-20/08/2010, Lisbonne, PRT / Frontiers in Artificial Intelligence and Applications.Helder Coelho, Rudi Studer, Michael Wooldridge (Eds.)}, publisher = {IOS Press}, pages = {835-840}, url = {http://www.booksonline.iospress.nl/Content/BookSeriesBook.aspx?piid=17702}, cemadoc = {http://cemadoc.irstea.fr/cemoa/PUB00030835}, year = {2010}, affiliation = {CEMAGREF CLERMONT FERRAND UR LISC FRA ; CEMAGREF CLERMONT FERRAND UR LISC FRA ; LIP6 UNIVERSITE PIERRE ET MARIE CURIE PARIS FRA}, ISBNISSN = {ISBN 978-1-60750-605-8}, abstract = {This paper addresses the issue of supporting the end-user of a classifier, when it is used as a decision support system, to classify new cases. We consider several kinds of classifiers: Statistical or machine learning classifiers, which are built on data, but also direct model-based classifiers that are built to solve a particular problem (like in viability or control problems). The end-user relies mainly on global information (like error rates or global sensitivity analysis) to assess the quality of the result given by the system. Class membership probability, if available, describes certainly the local statistical viewpoint. But it doesn't take into account other contextual information: Cases with high value of class membership probability can also be close to the decision boundary. In the case of numerical state space, we propose to use the decision boundary of the classifier (which always exists, even implicitly), to describe the situation of a particular case: The distance of a case to the decision boundary measures the robustness of the decision to a change in the input data. Other geometric concepts, such as the maximal maximal ball, can present a precise picture of the situation to the end-user. We show the interest of such a geometric study on different examples. / Cet article propose une méthode, basée sur la géométrie, pour aider l'utilisateur d'un système discriminant à apprécier les résultats dans le cadre de l'aide à la décision.}, }
close
I. Alvarez, S. Martin, O. Dordan, X. Litrico, P. Saint Pierre
Indicateurs de sécurité et de restauration dynamiques
Agir en situation d'incertitude : quelles constructions individuelles et collectives des régimes de protection et d'adaptation en agriculture ?, 22/11/2010-24/11/2010, Montpellier, FRA / Agir en situation d'incertitude : La construction individuelle et collective des régimes de protection et d'adaptation en agriculture (p. 12), 2010
abs web bib

La gestion de systèmes agricoles ou environnementaux, dans une perspective de gestion durable, conduit à réfléchir à des indicateurs de l'état actuel du système mais aussi de son évolution. La gestion durable peut être envisagée comme la conservation de propriétés intéressantes du système. Quand ces propriétés s'expriment sous la forme d'un ensemble de contraintes, la théorie de la viabilité permet de définir des indicateurs dynamiques qui prennent en compte la situation actuelle et les possibilités d'action dans le futur.
close

@inproceedings{PUB00030853, author = {Alvarez, I. and Martin, S. and Dordan, O. and Litrico, X. and Saint Pierre, P.}, title = {Indicateurs de sécurité et de restauration dynamiques}, title = {Safety and restauration dynamic indicators}, bookTitle = {Agir en situation d'incertitude : quelles constructions individuelles et collectives des régimes de protection et d'adaptation en agriculture ?, 22/11/2010-24/11/2010, Montpellier, FRA / Agir en situation d'incertitude : La construction individuelle et collective des régimes de protection et d'adaptation en agriculture}, pages = {12}, url = {http://www.agir-incertitude2010.fr/}, cemadoc = {http://cemadoc.irstea.fr/cemoa/PUB00030853}, year = {2010}, affiliation = {CEMAGREF CLERMONT FERRAND UR LISC FRA ; CEMAGREF CLERMONT FERRAND UR LISC FRA ; UNIVERSITE BORDEAUX II FRA ; CEMAGREF MONTPELLIER UR IRMO UMR G-EAU FRA ; LASTRE ET VIMADES PARIS FRA}, abstract = {La gestion de systèmes agricoles ou environnementaux, dans une perspective de gestion durable, conduit à réfléchir à des indicateurs de l'état actuel du système mais aussi de son évolution. La gestion durable peut être envisagée comme la conservation de propriétés intéressantes du système. Quand ces propriétés s'expriment sous la forme d'un ensemble de contraintes, la théorie de la viabilité permet de définir des indicateurs dynamiques qui prennent en compte la situation actuelle et les possibilités d'action dans le futur.}, }
close
S. Martin, I. Alvarez
Indicateurs dynamiques de gestion durable
23èmes Entretiens Jacques Cartier, "Vulnérabilité et résilience des écosystèmes aquatiques : de l'évaluation à un fonctionnement durable - une utopie ?", 22/11/2010-23/11/2010, Lyon, FRA / Vulnérabilité & résilience des écosystèmes: l'utopie d'une gestion durable (p. 6-15), 2010
abs web bib

La gestion de systèmes agricoles ou environnementaux, dans une perspective de gestion durable, conduit à réfléchir à des indicateurs de l'état actuel du système mais aussi de son évolution. La gestion durable peut être envisagée comme la conservation de propriétés intéressantes du système. Dans certains cas, la théorie de la viabilité permet de définir des indicateurs comme la résilience ou la robustesse, qui prennent en compte les évolutions possibles du système.
close

@inproceedings{PUB00030845, author = {Martin, S. and Alvarez, I.}, title = {Indicateurs dynamiques de gestion durable}, title = {Dynamical indicators of sustainable management}, bookTitle = {23èmes Entretiens Jacques Cartier, "Vulnérabilité et résilience des écosystèmes aquatiques : de l'évaluation à un fonctionnement durable - une utopie ?", 22/11/2010-23/11/2010, Lyon, FRA / Vulnérabilité & résilience des écosystèmes: l'utopie d'une gestion durable}, pages = {6-15}, url = {http://cemadoc.irstea.fr/cemoa/PUB00030845}, year = {2010}, affiliation = {CEMAGREF CLERMONT FERRAND UR LISC FRA ; CEMAGREF CLERMONT FERRAND UR LISC FRA}, abstract = {La gestion de systèmes agricoles ou environnementaux, dans une perspective de gestion durable, conduit à réfléchir à des indicateurs de l'état actuel du système mais aussi de son évolution. La gestion durable peut être envisagée comme la conservation de propriétés intéressantes du système. Dans certains cas, la théorie de la viabilité permet de définir des indicateurs comme la résilience ou la robustesse, qui prennent en compte les évolutions possibles du système.}, }
close
I. Alvarez, S. Martin
Explaining a result to the end-user: a geometric approach for classification problems
Exact09, IJCAI 2009 Workshop on explanation aware computing (International Joint Conferences on Artificial Intelligence), 11/07/2009-13/07/2009, Pasadena, USA (p. 102-109), , 2009
abs pdf web bib

This paper addresses the issue of the explanation of the result given to the end-user by a classier, when it is used as a decision support system. We consider machine learning classiers, which provide a class for new cases, but also deterministic classiers that are built to solve a particular problem (like in viability or control problems). The end-user relies mainly on global information (like error rates) to assess the quality of the result given by the system. Even class membership probability, if available, describes only the statistical viewpoint, it doesn't take into account the context of a particular case. In the case of numerical state space, we propose to use the decision boundary of the classier (which always exists, even implicitly), to describe the situation of a particular case: The distance of a case to the decision boundary measures the ro-bustness of the decision to a change in the input data. Other geometric concepts can present a precise picture of the situation to the end-user. This geometric study is applied to different types of classiers.
close

@inproceedings{PUB00027387, author = {Alvarez, I. and Martin, S.}, title = {Explaining a result to the end-user: a geometric approach for classification problems}, title = {Explication d'un résultat à l'utilisateur final : une approche géométrique pour les systèmes discriminants}, publisher = { }, bookTitle = {Exact09, IJCAI 2009 Workshop on explanation aware computing (International Joint Conferences on Artificial Intelligence), 11/07/2009-13/07/2009, Pasadena, USA}, pages = {102-109}, pdf = {http://motive.cemagref.fr/_publication/PUB00027387.pdf}, url = {http://exact2009.workshop.hm/}, cemadoc = {http://cemadoc.irstea.fr/cemoa/PUB00027387}, year = {2009}, affiliation = {CEMAGREF CLERMONT FERRAND UR LISC FRA ; CEMAGREF CLERMONT FERRAND UR LISC FRA}, abstract = {This paper addresses the issue of the explanation of the result given to the end-user by a classier, when it is used as a decision support system. We consider machine learning classiers, which provide a class for new cases, but also deterministic classiers that are built to solve a particular problem (like in viability or control problems). The end-user relies mainly on global information (like error rates) to assess the quality of the result given by the system. Even class membership probability, if available, describes only the statistical viewpoint, it doesn't take into account the context of a particular case. In the case of numerical state space, we propose to use the decision boundary of the classier (which always exists, even implicitly), to describe the situation of a particular case: The distance of a case to the decision boundary measures the ro-bustness of the decision to a change in the input data. Other geometric concepts can present a precise picture of the situation to the end-user. This geometric study is applied to different types of classiers.}, }
close
S. Mesmoudi, I. Alvarez, S. Martin, M. Sicard, P.H. Wuillemin
Geometric analysis of a capture basin: application to cheese ripening process
European Conference on Complex System, 21/09/2009-25/09/2009, Warwick, GBR (p. 14), 2009
abs web bib

This paper addresses the issue of analysing the results given by viability sets, which represent all the viable states of a dynamical system. Viability sets and capture basin are a relatively new method to study complex dynamical systems, focusing on the preservation of some properties of the system (constraints in the state space) rather than on the possible stable states and equilibria. The viability set delimits the area of the state space where there is always a list of controls that allows the system to verify indefinitely the constraints. A viability set represents a huge amount of information about the system it studies. We propose here a geometric study of viability sets. This study is applied to the viability tube (the sequence of viability set over time) of a food process.
close

@inproceedings{PUB00027389, author = {Mesmoudi, S. and Alvarez, I. and Martin, S. and Sicard, M. and Wuillemin, P.H.}, title = {Geometric analysis of a capture basin: application to cheese ripening process}, title = {Analyse géométrique d'un bassin de capture : application à l'affinage du camembert}, bookTitle = {European Conference on Complex System, 21/09/2009-25/09/2009, Warwick, GBR}, pages = {14}, url = {http://cemadoc.irstea.fr/cemoa/PUB00027389}, year = {2009}, affiliation = {UNIVERSITE PIERRE ET MARIE CURIE LIP6 PARIS FRA ; CEMAGREF CLERMONT FERRAND UR LISC FRA ; CEMAGREF CLERMONT FERRAND UR LISC FRA ; INRA UMR782 GMPA THIVERVAL-GRIGNON FRA ; UNIVERSITE PIERRE ET MARIE CURIE LIP6 PARIS FRA}, abstract = {This paper addresses the issue of analysing the results given by viability sets, which represent all the viable states of a dynamical system. Viability sets and capture basin are a relatively new method to study complex dynamical systems, focusing on the preservation of some properties of the system (constraints in the state space) rather than on the possible stable states and equilibria. The viability set delimits the area of the state space where there is always a list of controls that allows the system to verify indefinitely the constraints. A viability set represents a huge amount of information about the system it studies. We propose here a geometric study of viability sets. This study is applied to the viability tube (the sequence of viability set over time) of a food process. }, }
close
M. Sicard, R. Reuillon, N. Perrot, C. Baudrit, I. Alvarez, S. Martin
Control of Camembert mass loss during ripening with viability theory
XIIème Congrès de la Société Française de Génie des Procédés , 14/10/2009-16/10/2009, Marseille, FRA / Récents Progrès en Génie des Procédés, Numéro 98 (p. 6), SFGP, Paris, 2009
abs web bib

The viability theory developed by [Aubin1991] has been adapted to calculate the optimal monitoring of cheese ripening process. This method was used efficiently in ecology or in finance but never in food process. The aim was to find the controls allowing reaching a compromise between the quality of the ripened cheeses and the production costs. A viability kernel and the costs simulations of the viable trajectories were computed. Then, the optimal ripening trajectories were validated during pilot trials. The results were finally compared to those obtained for cheeses ripened at 92% of relative humidity and 12°C of temperature, the conditions usually applied in dairy industries for the Camembert ripening process. The optimized ripening control application in a pilot trial give interesting results with a ripening shortened of 7 days and a product with a quality level for consumer identical at day 15 of ripening as the cheese ripened following classical controls (92% RH, 12°C) at day 20 of ripening.
close

@inproceedings{PUB00033595, author = {Sicard, M. and Reuillon, R. and Perrot, N. and Baudrit, C. and Alvarez, I. and Martin, S.}, title = {Control of Camembert mass loss during ripening with viability theory}, title = {Contrôler la perte de masse du Camembert durant l'affinage à l'aide de la théorie de la viabilité}, volume = {98}, bookTitle = {XIIème Congrès de la Société Française de Génie des Procédés , 14/10/2009-16/10/2009, Marseille, FRA / Récents Progrès en Génie des Procédés, Numéro 98}, address = {Paris}, publisher = {SFGP}, pages = {6}, url = {http://cemadoc.irstea.fr/cemoa/PUB00033595}, year = {2009}, affiliation = {UMR782 GMPA,AGROPARISTECH INRA THIVERVAL GRIGNON FRA ; ISC PARIS FRA ; UMR782 GMPA AGROPARISTECH INRA THIVERVAL GRIGNON FRA ; UMR782 GMPA,AGROPARISTECH INRA THIVERVAL GRIGNON FRA ; CEMAGREF CLERMONT FERRAND UR LISC FRA ; CEMAGREF CLERMONT FERRAND UR LISC FRA}, ISBNISSN = {ISBN 2-910239-72-1}, abstract = {The viability theory developed by [Aubin1991] has been adapted to calculate the optimal monitoring of cheese ripening process. This method was used efficiently in ecology or in finance but never in food process. The aim was to find the controls allowing reaching a compromise between the quality of the ripened cheeses and the production costs. A viability kernel and the costs simulations of the viable trajectories were computed. Then, the optimal ripening trajectories were validated during pilot trials. The results were finally compared to those obtained for cheeses ripened at 92% of relative humidity and 12°C of temperature, the conditions usually applied in dairy industries for the Camembert ripening process. The optimized ripening control application in a pilot trial give interesting results with a ripening shortened of 7 days and a product with a quality level for consumer identical at day 15 of ripening as the cheese ripened following classical controls (92% RH, 12°C) at day 20 of ripening.}, }
close
M. Sicard, N. Perrot, C. Baudrit, I. Alvarez, S. Martin
The viability theory to control food processes
8th World Congress of Chemical Engineering, WCCE8, 23/08/2009-27/08/2009, Montréal, CAN / Proceedings of Process Control and Optimization Symposium, 8th World Congress on Chemical Engineering (WCCE8) (p. 5), 2009
abs web bib

The viability theory developed by Aubin (1991) has been adapted to calculate the optimal monitoring of cheese ripening process. This method was used efficiently in ecology or in finance but never in food process. The aim was to find the controls allowing to reach a compromise between the quality of the ripened cheeses and the production costs. A viability kernel and the costs simulations of the viable trajectories were computed. Then, the optimal ripening trajectories were validated during pilot trials. The results were finally compared to those obtained for cheeses ripened at 92% of relative humidity and 12°C of temperature, the conditions usually applied in dairy industries for Camembert ripening.
close

@inproceedings{PUB00033594, author = {Sicard, M. and Perrot, N. and Baudrit, C. and Alvarez, I. and Martin, S.}, title = {The viability theory to control food processes}, title = {L'utilisation de la théorie de la viabilité pour contrôler les procédés alimentaires}, bookTitle = {8th World Congress of Chemical Engineering, WCCE8, 23/08/2009-27/08/2009, Montréal, CAN / Proceedings of Process Control and Optimization Symposium, 8th World Congress on Chemical Engineering (WCCE8)}, pages = {5}, url = {http://cemadoc.irstea.fr/cemoa/PUB00033594}, year = {2009}, affiliation = {UMR782 GMPA AGROPARISTECH INRA THIVERVAL GRIGNON FRA ; UMR782 GMPA AGROPARISTECH INRA THIVERVAL GRIGNON FRA ; UMR782 GMPA AGROPARISTECH INRA THIVERVAL GRIGNON FRA ; CEMAGREF CLERMONT FERRAND UR LISC FRA ; CEMAGREF CLERMONT FERRAND UR LISC FRA}, abstract = {The viability theory developed by Aubin (1991) has been adapted to calculate the optimal monitoring of cheese ripening process. This method was used efficiently in ecology or in finance but never in food process. The aim was to find the controls allowing to reach a compromise between the quality of the ripened cheeses and the production costs. A viability kernel and the costs simulations of the viable trajectories were computed. Then, the optimal ripening trajectories were validated during pilot trials. The results were finally compared to those obtained for cheeses ripened at 92% of relative humidity and 12°C of temperature, the conditions usually applied in dairy industries for Camembert ripening. }, }
close
M. Sicard, N. Perrot, C. Baudrit, R. Reuillon, P. Bourgine, I. Alvarez, S. Martin
The viability theory to control complex food processes
ECCS'09, European Conference on Complex Systems, 21/09/2009-25/09/2009, Warwick, GBR (p. 6), 2009
abs web bib

The viability theory developed by [Aubin(1991)] has been adapted to calculate the optimal monitoring of cheese ripening process. This method was used efficiently in ecology or in finance but never in food process. The aim was to find the controls allowing reaching a compromise between the quality of the ripened cheeses and the production costs. A viability kernel and the costs simulations of the viable trajectories were computed. Then, the optimal ripening trajectories were validated during pilot trials. The results were finally compared to those obtained for cheeses ripened at 92% of relative humidity and 12°C of temperature, the conditions usually applied in dairy industries for the Camembert ripening process.
close

@inproceedings{PUB00029638, author = {Sicard, M. and Perrot, N. and Baudrit, C. and Reuillon, R. and Bourgine, P. and Alvarez, I. and Martin, S.}, title = {The viability theory to control complex food processes}, title = {La théorie de la viabilité pour contrôler les procédés alimentaires complexes}, bookTitle = {ECCS'09, European Conference on Complex Systems, 21/09/2009-25/09/2009, Warwick, GBR}, pages = {6}, url = {http://cemadoc.irstea.fr/cemoa/PUB00029638}, year = {2009}, affiliation = {INRA UMR782 GMPA THIVERVAL GRIGNON FRA ; INRA UMR782 GMPA THIVERVAL GRIGNON FRA ; INRA UMR782 GMPA THIVERVAL GRIGNON FRA ; ISC PARIS FRA ; ISC PARIS FRA ; CEMAGREF CLERMONT FERRAND UR LISC FRA ; CEMAGREF CLERMONT FERRAND UR LISC FRA}, abstract = {The viability theory developed by [Aubin(1991)] has been adapted to calculate the optimal monitoring of cheese ripening process. This method was used efficiently in ecology or in finance but never in food process. The aim was to find the controls allowing reaching a compromise between the quality of the ripened cheeses and the production costs. A viability kernel and the costs simulations of the viable trajectories were computed. Then, the optimal ripening trajectories were validated during pilot trials. The results were finally compared to those obtained for cheeses ripened at 92% of relative humidity and 12°C of temperature, the conditions usually applied in dairy industries for the Camembert ripening process.}, }
close
A. Sordoni, J.P. Briot, I. Alvarez, E. Vasconcelos, M. Irving, G. Melo, V. Sebba Patto
Design of a participatory decision making agent architecture based on argumentation and influence function: application to a serious game about biodiversity conservation
Symposium on COGnitive systems with Interactive Sensors (COGIS'09), 16/11/2009-18/11/2009, Paris, FRA (p. 6), SEE, Paris, 2009
abs web bib

This paper addresses an ongoing experience in the design of an artificial agent taking decisions and combining them with the decisions taken by human agents. The context is a serious game research project, aimed at computer-based support for participatory management of protected areas (and more specifically national parks) in order to promote biodiversity conservation and social inclusion. Its objective is to help various stakeholders (e.g., environmentalist, tourism operator) to collectively understand conflict dynamics and explore negotiation strategies for the management of parks. In this paper, after introducing the design of our serious game, named Sim- Parc, we will describe the architecture of the decision making agent playing the role of the park manager. In the game, the park manager makes final decisions based on its own analysis and also on the votes of the stakeholders. It includes two modules: 1) individual decision based on a model of argumentation, which also provides basis to justify and explain the decision; 2) participatory decision to take into account the preferences/votes from the stakeholders. Keywords: serious game, role playing game, park management, biodiversity conservation, participatory, decision making, artificial agent, architecture, argumentation, influence
close

@inproceedings{PUB00027392, author = {Sordoni, A. and Briot, J.P. and Alvarez, I. and Vasconcelos, E. and Irving, M. and Melo, G. and Sebba Patto, V.}, title = {Design of a participatory decision making agent architecture based on argumentation and influence function: application to a serious game about biodiversity conservation}, title = {Architecture agent pour la décision participative, basée sur l'argumentation et les fonctions d'influence : application à un jeu sérieux de conservation de la biodiversité}, bookTitle = {Symposium on COGnitive systems with Interactive Sensors (COGIS'09), 16/11/2009-18/11/2009, Paris, FRA}, address = {Paris}, publisher = {SEE}, pages = {6}, url = {http://cemadoc.irstea.fr/cemoa/PUB00027392}, year = {2009}, affiliation = {UNIVERSITE PIERRE ET MARIE CURIE LIP6 PARIS FRA ; UNIVERSITE PIERRE ET MARIE CURIE LIP6 PARIS FRA ; CEMAGREF CLERMONT FERRAND UR LISC FRA ; PONTIFICIA UNVERSIDADE CATOLICA RIO DE JANEIRO BRA ; UNIVERSIDADE FEDERAL DO RIO DE JANEIRO BRA ; UNIVERSIDADE FEDERAL DO RIO DE JANEIRO BRA ; UNIVERSITE PIERRE ET MARIE CURIE LIP6 PARIS FRA}, abstract = {This paper addresses an ongoing experience in the design of an artificial agent taking decisions and combining them with the decisions taken by human agents. The context is a serious game research project, aimed at computer-based support for participatory management of protected areas (and more specifically national parks) in order to promote biodiversity conservation and social inclusion. Its objective is to help various stakeholders (e.g., environmentalist, tourism operator) to collectively understand conflict dynamics and explore negotiation strategies for the management of parks. In this paper, after introducing the design of our serious game, named Sim- Parc, we will describe the architecture of the decision making agent playing the role of the park manager. In the game, the park manager makes final decisions based on its own analysis and also on the votes of the stakeholders. It includes two modules: 1) individual decision based on a model of argumentation, which also provides basis to justify and explain the decision; 2) participatory decision to take into account the preferences/votes from the stakeholders. Keywords: serious game, role playing game, park management, biodiversity conservation, participatory, decision making, artificial agent, architecture, argumentation, influence}, }
close
I. Alvarez, S. Bernard, G. Deffuant
Keep the decision tree and estimate the class probabilities using its decision boundary
International Joint Conference of Artificial Intelligence, Hyderabad, IND, 7-12 janvier 2007 (p. 6), 2007
abs web bib

This paper proposes a new method to estimate the class membership probability of the cases classified by a Decision Tree. This method provides smooth class probabilities estimate, without any modification of the tree, when the data are numerical. It applies a posteriori and doesn't use additional training cases. It relies on the distance to the decision boundary induced by the decision tree. The distance is computed on the training sample. It is then used as an input for a very simple one dimension kernel-based density estimator, which provides an estimate of the class membership probability. This geometric method gives good results even with pruned trees, so the intelligibility of the tree is fully preserved. / Cet article propose une nouvelle méthode pour estimer les probabilités d'appartenance aux classes des cas classés par un arbre de décision. Cette méthode produit des estimateurs de probabilités qui varient avec les exemples, et sans modification de l'arbre, dès lors que les données sont numériques. La méthode s'applique a posteriori et ne requiert pas d'exemples d'apprentissage supplémentaires. Elle repose sur la frontière de décision induite par l'arbre de décision. Cette distance est calculée sur la base d'apprentissage. Elle est ensuite fournie à un estimateur de densité à noyau qui calcule la probabilité d'appartenance à chaque classe. Cette méthode géométrique donne de bons résultats même après élagage, l'intelligibilité des arbres est donc complètement préservée.
close

@inproceedings{PUB00020610, author = {Alvarez, I. and Bernard, S. and Deffuant, G.}, title = {Keep the decision tree and estimate the class probabilities using its decision boundary}, title = {Garder l'arbre de décision et estimer la probabilité d'appartenance aux classes en utilisant sa frontière de décision}, bookTitle = {International Joint Conference of Artificial Intelligence, Hyderabad, IND, 7-12 janvier 2007}, pages = {6}, url = {http://www.ijcai07.org}, cemadoc = {http://cemadoc.irstea.fr/cemoa/PUB00020610}, year = {2007}, affiliation = {CEMAGREF CLERMONT FERRAND LISC ; CEMAGREF CLERMONT FERRAND LISC ; CEMAGREF CLERMONT FERRAND LISC}, abstract = {This paper proposes a new method to estimate the class membership probability of the cases classified by a Decision Tree. This method provides smooth class probabilities estimate, without any modification of the tree, when the data are numerical. It applies a posteriori and doesn't use additional training cases. It relies on the distance to the decision boundary induced by the decision tree. The distance is computed on the training sample. It is then used as an input for a very simple one dimension kernel-based density estimator, which provides an estimate of the class membership probability. This geometric method gives good results even with pruned trees, so the intelligibility of the tree is fully preserved. / Cet article propose une nouvelle méthode pour estimer les probabilités d'appartenance aux classes des cas classés par un arbre de décision. Cette méthode produit des estimateurs de probabilités qui varient avec les exemples, et sans modification de l'arbre, dès lors que les données sont numériques. La méthode s'applique a posteriori et ne requiert pas d'exemples d'apprentissage supplémentaires. Elle repose sur la frontière de décision induite par l'arbre de décision. Cette distance est calculée sur la base d'apprentissage. Elle est ensuite fournie à un estimateur de densité à noyau qui calcule la probabilité d'appartenance à chaque classe. Cette méthode géométrique donne de bons résultats même après élagage, l'intelligibilité des arbres est donc complètement préservée.}, }
close
I. Alvarez, S. Bernard
Ranking Cases with Decision Trees: a Geometric Method that Preserves Intelligibility
Actes de Nineteenth International Joint Conference on Artificial Intelligence, Edinbourgh, GBR,30/07-05/08/2005 (p. 635-641), 2005
abs web bib

This paper proposes a new method to rank the cases classified by a decision tree. The method applies a posteriori without modification of the tree and doesn't use additional training cases. It consists in computing the distance of the cases to the decision boundary induced by the decision tree, and to rank them according to this geometric score. When the data are numeric it is very easy to implement and efficient. The distance-based score is a global assess, contrary to other methods that evaluate the score at the level of the leaf. The distance-based score gives good results even with pruned tree, so if the tree is intelligible this property is preserved with an improved ranking ability. The main reason for the efficacity of the geometric method is that in most cases when the classifier is sufficiently accurate, errors are located near the decision boundary. / Cet article propose une nouvelle méthode pour ordonner des exemples classés par un arbre de décision. Cette méthode s'applique à posteriori sans modification de l'arbre et sans utiliser d'exemples d'apprentissage supplémentaires. Elle consiste à calculer la distance de chaque cas à la frontière de décision crée par l'arbre de décision, puis à ranger les cas suivants ce score géométrique. Quand les données sont numériques, cette méthode est facile à implémenter et très efficace. Le score basé sur la distance est évalué globalement, contrairement à d'autres scores calculés au niveau de chaque feuille. Le score géométrique donne de bons résultats même sur des arbres élagués, donc l'intelligibilité de l'arbre est préservée. La raison principale de l'efficacité de cette méthode géométrique vient du fait que si le classifieur est suffisamment bon, les erreurs sont localisées principalement près de la surface de décision.
close

@inproceedings{PUB00018580, author = {Alvarez, I. and Bernard, S.}, title = {Ranking Cases with Decision Trees: a Geometric Method that Preserves Intelligibility}, title = {Ranger des cas avec des arbres de décision : une méthode géométrique qui préserve l'intelligibilité}, bookTitle = {Actes de Nineteenth International Joint Conference on Artificial Intelligence, Edinbourgh, GBR,30/07-05/08/2005}, pages = {635-641}, url = {http://www.ijcai.org/papers/1502.pdf}, cemadoc = {http://cemadoc.irstea.fr/cemoa/PUB00018580}, year = {2005}, affiliation = {CEMAGREF CLERMONT FERRAND LISC ; CEMAGREF CLERMONT FERRAND LISC}, abstract = {This paper proposes a new method to rank the cases classified by a decision tree. The method applies a posteriori without modification of the tree and doesn't use additional training cases. It consists in computing the distance of the cases to the decision boundary induced by the decision tree, and to rank them according to this geometric score. When the data are numeric it is very easy to implement and efficient. The distance-based score is a global assess, contrary to other methods that evaluate the score at the level of the leaf. The distance-based score gives good results even with pruned tree, so if the tree is intelligible this property is preserved with an improved ranking ability. The main reason for the efficacity of the geometric method is that in most cases when the classifier is sufficiently accurate, errors are located near the decision boundary. / Cet article propose une nouvelle méthode pour ordonner des exemples classés par un arbre de décision. Cette méthode s'applique à posteriori sans modification de l'arbre et sans utiliser d'exemples d'apprentissage supplémentaires. Elle consiste à calculer la distance de chaque cas à la frontière de décision crée par l'arbre de décision, puis à ranger les cas suivants ce score géométrique. Quand les données sont numériques, cette méthode est facile à implémenter et très efficace. Le score basé sur la distance est évalué globalement, contrairement à d'autres scores calculés au niveau de chaque feuille. Le score géométrique donne de bons résultats même sur des arbres élagués, donc l'intelligibilité de l'arbre est préservée. La raison principale de l'efficacité de cette méthode géométrique vient du fait que si le classifieur est suffisamment bon, les erreurs sont localisées principalement près de la surface de décision.}, }
close
I. Alvarez
Sensitivity Analysis of the Result in Binary Decision Trees
15th European Conference on Machine Learning, ECML, Pisa, ITA, 2004 (p. 12), 2004
abs web bib

This paper proposes a new method to qualify the result given by a decision tree when it is used as a decision aid system. When the data are numerical, we compute the distance of a case from the decision surface. This distance measures the sensitivity of the result to a change in the input data. With a different distance it is also possible to measure the sensitivity of the result to small changes in the tree. The distance from the decision surface can also be combined to the error rate in order to provide a context-dependent information to the end-user.
close

@inproceedings{PUB00015178, author = {Alvarez, I.}, title = {Sensitivity Analysis of the Result in Binary Decision Trees}, bookTitle = {15th European Conference on Machine Learning, ECML, Pisa, ITA, 2004}, pages = {12}, url = {http://cemadoc.irstea.fr/cemoa/PUB00015178}, year = {2004}, affiliation = {CEMAGREF CLERMONT FERRAND LISC}, number = {K LI04 5}, abstract = {This paper proposes a new method to qualify the result given by a decision tree when it is used as a decision aid system. When the data are numerical, we compute the distance of a case from the decision surface. This distance measures the sensitivity of the result to a change in the input data. With a different distance it is also possible to measure the sensitivity of the result to small changes in the tree. The distance from the decision surface can also be combined to the error rate in order to provide a context-dependent information to the end-user.}, }
close
I. Alvarez, D. Lacondemine, F. Stermann
Expliquer à l'utilisateur le résultat d'un arbre de décision
Conférence RFIA 2004, Toulouse, 28-30 janvier 2004 , 2004
abs web bib

Decision trees can be used as decision support systems, in order to classify input data and to provide the output class as a result. In order to appraise this result, the end-user can rely mainly on the trace of the classification, or on some estimation of the error-rate. But the trace doesn`t contain all the information that is necessary to understand the case at hand. We propose a new method to qualify the result given by the decision tree for a new case, when the input data are continuous. We first identify in the tree, (and not in the trace), the tests that are the most relevant to explain the output class, by the way of sensitivity analysis. This is performed by a geometric study of the decision surface (that is the frontier of the inverse image of the different classes). The underlying metric is then used as a medium to describe the global situation to the end-user. He can suggest a more relevant metric and the most sensitive tests are computed for the latter. / Les arbres de décision sont des outils de discrimination qui peuvent être utilisés pour prédire la classe de nouveaux cas. Pour apprécier ce résultat, l'utilisateur ne dispose en général que de la trace du classement et d'une estimation du taux d'erreur. Malheureusement la trace ne contient pas les informations nécessaires à une bonne compréhension de la situation. Nous proposons d'apporter à l'utilisateur une information sur la robustesse du résultat et une représentation de la situation dans le cas de données numériques. Pour cela nous recherchons les tests de l'arbre de décision qui sont les plus sensibles à une modification des données d'entrée. Cette analyse de sensibilité est effectuée par une étude géométrique de la surface de décision (la frontière de l`image inverse des différentes classes). La métrique sous-jacente sert de support pour décrire la situation à l`utilisateur. Celui-ci peut suggérer des changements de métriques et les tests les plus sensibles sont recalculés.
close

@inproceedings{PUB00014087, author = {Alvarez, I. and Lacondemine, D. and Stermann, F.}, title = {Expliquer à l'utilisateur le résultat d'un arbre de décision}, title = {Explaining the result of a decision tree to the end-user}, bookTitle = {Conférence RFIA 2004, Toulouse, 28-30 janvier 2004}, url = {http://cemadoc.irstea.fr/cemoa/PUB00014087}, year = {2004}, affiliation = {CEMAGREF CLERMONT FERRAND LISC ; UNIVERSITE REIMS ; CEMAGREF CLERMONT FERRAND LISC}, number = {K LI04 1}, abstract = {Decision trees can be used as decision support systems, in order to classify input data and to provide the output class as a result. In order to appraise this result, the end-user can rely mainly on the trace of the classification, or on some estimation of the error-rate. But the trace doesn`t contain all the information that is necessary to understand the case at hand. We propose a new method to qualify the result given by the decision tree for a new case, when the input data are continuous. We first identify in the tree, (and not in the trace), the tests that are the most relevant to explain the output class, by the way of sensitivity analysis. This is performed by a geometric study of the decision surface (that is the frontier of the inverse image of the different classes). The underlying metric is then used as a medium to describe the global situation to the end-user. He can suggest a more relevant metric and the most sensitive tests are computed for the latter. / Les arbres de décision sont des outils de discrimination qui peuvent être utilisés pour prédire la classe de nouveaux cas. Pour apprécier ce résultat, l'utilisateur ne dispose en général que de la trace du classement et d'une estimation du taux d'erreur. Malheureusement la trace ne contient pas les informations nécessaires à une bonne compréhension de la situation. Nous proposons d'apporter à l'utilisateur une information sur la robustesse du résultat et une représentation de la situation dans le cas de données numériques. Pour cela nous recherchons les tests de l'arbre de décision qui sont les plus sensibles à une modification des données d'entrée. Cette analyse de sensibilité est effectuée par une étude géométrique de la surface de décision (la frontière de l`image inverse des différentes classes). La métrique sous-jacente sert de support pour décrire la situation à l`utilisateur. Celui-ci peut suggérer des changements de métriques et les tests les plus sensibles sont recalculés.}, }
close
I. Alvarez
Explaining the result of a decision tree to the end-user
IOS PressActes de 16ème European Conference on Artificial Intelligence, Valencia, ESP, 22-27 août 2004 (p. 411-415), 2004
abs web bib

This paper addresses the problem of the explanation of the result given by a decision tree, when it is used to predict the class of new cases. In order to evaluate this result, the end-user relies on some estimate of the error rate and on the trace of the classification. Unfortunately the trace doesn't contain the information necessary to understand the case at hand. We propose a new method to qualify the result given by a decision tree when the data are continuous-valued. We perform a geometric study of the decision surface (the boundary of the inverse image of the different classes). This analysis gives the list of the tests of the tree that are the most sensitive to a change in the input data. Unlike the trace, this list can easily be ordered and pruned so that only the most important tests are presented. We also show how the metric can be used to interact with the end-user. / Cet article s'intéresse au problème de l'explication du résultat fourni par un arbre de décision, quand l'arbre est utilisé pour prédire la classe de nouveaux cas. Pour apprécier ce résultat, l'utilisateur dispose d'information sur l'erreur et aussi de la trace de la classification. Malheureusement la trace ne contient pas les éléments nécessaire à une bonne compréhension de la sotuation. Nous proposons une nouvelle méthode pour qualifier le résultat de l'arbre de décision quand les données sont continues. Nous effectuons une étude géométrique de la surface de décision (la frontière de l'image inverse des différentes classes). Cette analyse produit la liste des tests de l'arbre qui sont les plus sensibles à une modification des données d'entrée. Contrairement à la trace, cette liste peut facilement être ordonnée et élaguée, de manière à ne présenter que les tests les plus importants à l'utilisateur. Nous examinons aussi comment la métrique peut être utilisée comme support d'interaction avec l'utilisateur final.
close

@inproceedings{PUB00014088, author = {Alvarez, I.}, title = {Explaining the result of a decision tree to the end-user}, title = {Expliquer le résultat d'un arbre de décision à son utilisateur}, bookTitle = {IOS PressActes de 16ème European Conference on Artificial Intelligence, Valencia, ESP, 22-27 août 2004}, pages = {411-415}, url = {http://cemadoc.irstea.fr/cemoa/PUB00014088}, year = {2004}, affiliation = {CEMAGREF CLERMONT FERRAND LISC}, number = {K LI04 2}, abstract = {This paper addresses the problem of the explanation of the result given by a decision tree, when it is used to predict the class of new cases. In order to evaluate this result, the end-user relies on some estimate of the error rate and on the trace of the classification. Unfortunately the trace doesn't contain the information necessary to understand the case at hand. We propose a new method to qualify the result given by a decision tree when the data are continuous-valued. We perform a geometric study of the decision surface (the boundary of the inverse image of the different classes). This analysis gives the list of the tests of the tree that are the most sensitive to a change in the input data. Unlike the trace, this list can easily be ordered and pruned so that only the most important tests are presented. We also show how the metric can be used to interact with the end-user. / Cet article s'intéresse au problème de l'explication du résultat fourni par un arbre de décision, quand l'arbre est utilisé pour prédire la classe de nouveaux cas. Pour apprécier ce résultat, l'utilisateur dispose d'information sur l'erreur et aussi de la trace de la classification. Malheureusement la trace ne contient pas les éléments nécessaire à une bonne compréhension de la sotuation. Nous proposons une nouvelle méthode pour qualifier le résultat de l'arbre de décision quand les données sont continues. Nous effectuons une étude géométrique de la surface de décision (la frontière de l'image inverse des différentes classes). Cette analyse produit la liste des tests de l'arbre qui sont les plus sensibles à une modification des données d'entrée. Contrairement à la trace, cette liste peut facilement être ordonnée et élaguée, de manière à ne présenter que les tests les plus importants à l'utilisateur. Nous examinons aussi comment la métrique peut être utilisée comme support d'interaction avec l'utilisateur final.}, }
close

Technical Reports (2)

I. Alvarez
Explication géométrique du résultat dans les arbres de décision
2002
abs web bib

This article is devoted to the problem of the explanation of the result given by the decision tree (DT) when it is used as a decision aid system, to classify input data and to provide the output class as a result. At now the end-user can rely mainly on some estimation of the error-rate or on the trace of the classification, that is the path run through the DT. We propose here two new methods to qualify the result given by the DT for each particular case. These methods are based on a gometric study of the inverse image of the different classes (the decision surface). First the tests in the tree that are the most relevant to explain the final class are identifies by the way of sensitivity analysis: the vector of initial data is projected onto the decision surface. The second method relies on a study of the distribution of errors in relation to the decision surface : the distance to this surface is computed for a sample. The distance of a new case is then presented to the end-user in comparaison to the estimation of the distribution of distances. / Cet article s'intéresse au problème de l'explication des résultats fournis par un arbre de décision (AD) quand celui-ci est utilisé en tant que système d'aide à la décision. On s'intéresse au résultat prédit par l'arbre de décision pour chaque vecteur particulier de données d'entrée, c'est à dire à la classe associée à ce vecteur, et on cherche à apporter une information supplémentaire qui permette d'expliquer ou de qualifier le résultat. Deux types principaux d'information sont actuellement disponibles : la trace du classement, c'est à dire le chemin parcouru dans l'arbre, et une estimation du taux d'erreur ou d'un risque associé à un mauvais classement. Nous proposons ici deux nouvelles méthodes de qualification du résultat, basées sur une étude géométrique de la frontière de l'image inverse des différentes classes (la surface de décision). La première méthode consiste à identifier les séparateurs les plus déterminants pour expliquer le résultat, en effectuant une analyse de sensibilité, c'est à dire en projetant les données initiales sur la surface de décision. La deuxième méthode consiste à caractériser la distribution des erreurs relatives à la surface de décision, en calculant la distance à cette surface pour un ensemble de données.
close

@techreport{PUB00011114, author = {Alvarez, I.}, title = {Explication géométrique du résultat dans les arbres de décision}, title = {Geometric explanation of the result in decision trees}, pages = {18}, url = {http://cemadoc.irstea.fr/cemoa/PUB00011114}, year = {2002}, affiliation = {CEMAGREF CLERMONT FERRAND LISC}, number = {02/0718 ; LI02 4}, abstract = {This article is devoted to the problem of the explanation of the result given by the decision tree (DT) when it is used as a decision aid system, to classify input data and to provide the output class as a result. At now the end-user can rely mainly on some estimation of the error-rate or on the trace of the classification, that is the path run through the DT. We propose here two new methods to qualify the result given by the DT for each particular case. These methods are based on a gometric study of the inverse image of the different classes (the decision surface). First the tests in the tree that are the most relevant to explain the final class are identifies by the way of sensitivity analysis: the vector of initial data is projected onto the decision surface. The second method relies on a study of the distribution of errors in relation to the decision surface : the distance to this surface is computed for a sample. The distance of a new case is then presented to the end-user in comparaison to the estimation of the distribution of distances. / Cet article s'intéresse au problème de l'explication des résultats fournis par un arbre de décision (AD) quand celui-ci est utilisé en tant que système d'aide à la décision. On s'intéresse au résultat prédit par l'arbre de décision pour chaque vecteur particulier de données d'entrée, c'est à dire à la classe associée à ce vecteur, et on cherche à apporter une information supplémentaire qui permette d'expliquer ou de qualifier le résultat. Deux types principaux d'information sont actuellement disponibles : la trace du classement, c'est à dire le chemin parcouru dans l'arbre, et une estimation du taux d'erreur ou d'un risque associé à un mauvais classement. Nous proposons ici deux nouvelles méthodes de qualification du résultat, basées sur une étude géométrique de la frontière de l'image inverse des différentes classes (la surface de décision). La première méthode consiste à identifier les séparateurs les plus déterminants pour expliquer le résultat, en effectuant une analyse de sensibilité, c'est à dire en projetant les données initiales sur la surface de décision. La deuxième méthode consiste à caractériser la distribution des erreurs relatives à la surface de décision, en calculant la distance à cette surface pour un ensemble de données.}, }
close
P. Barbe, M. Berducat, B. Bonicelli, E. Hugo, P. Zwaenepoel, I. Alvarez, J.F. Billot, F. Gaillard, V. Polvèche, M. Rombaut, P. Roux
Evaluation du programme MECAPLUS "Prestations d'ingénieries et d'essais de terrain pour les PME/PMI et les organismes interprofessionnels". Rapport interne
1997
abs web bib

Presentation of the MECAPLUS scientific programme to the specialist Commission of the CEMAGREF's Department for Agricultural and Food Equipment on 24/06/1997. It sets out MECAPLUS's activities from January 1995 to April 1997. Conclusions and prospects at 30/04/1997. / Présentation du programme scientifique MECAPLUS lors de la Commission spécialisée du département Equipements Agricoles et Alimentaires du Cemagref, le 24/06/1997. Exposés des activités MECAPLUS de janvier 1995 à avril 1997. Conclusions et perspectives au 30/04/1997.
close

@techreport{PUB00006571, author = {Barbe, P. and Berducat, M. and Bonicelli, B. and Hugo, E. and Zwaenepoel, P. and Alvarez, I. and Billot, J.F. and Gaillard, F. and Polvèche, V. and Rombaut, M. and Roux, P.}, title = {Evaluation du programme MECAPLUS "Prestations d'ingénieries et d'essais de terrain pour les PME/PMI et les organismes interprofessionnels". Rapport interne}, title = {Evaluation of the MECAPLUS programme "Project study and development services and field tests for SME/SMIs and interprofessional organisms". Internal report.}, pages = {60}, url = {http://cemadoc.irstea.fr/cemoa/PUB00006571}, year = {1997}, affiliation = {CEMAGREF MONTPELLIER GEMO ; CEMAGREF CLERMONT FERRAND CPCF ; CEMAGREF ANTONY TMAN ; CEMAGREF CLERMONT FERRAND QACF ; CEMAGREF CLERMONT FERRAND LISC}, number = {97/0198}, abstract = {Presentation of the MECAPLUS scientific programme to the specialist Commission of the CEMAGREF's Department for Agricultural and Food Equipment on 24/06/1997. It sets out MECAPLUS's activities from January 1995 to April 1997. Conclusions and prospects at 30/04/1997. / Présentation du programme scientifique MECAPLUS lors de la Commission spécialisée du département Equipements Agricoles et Alimentaires du Cemagref, le 24/06/1997. Exposés des activités MECAPLUS de janvier 1995 à avril 1997. Conclusions et perspectives au 30/04/1997.}, }
close