Les travaux de recherche presentes dans cette these rentrent dans le cadre de la problematique de... more Les travaux de recherche presentes dans cette these rentrent dans le cadre de la problematique de l'Extraction de Connaissances a partir des Donnees (ECD) que nous avons illustree a travers le traitement des donnees du Systeme d'Information Geographique etabli par le Bureau de Recherches Geologiques et Minieres (BRGM). Les objectifs de cette these sont d'etablir des cartes predictives de substances, d'avoir une interpretation sur les criteres permettant d'etablir cette decision, et de fournir une hierarchisation des attributs descriptifs. Pour atteindre ces objectifs, nous avons propose etape par etape un ensemble d'ameliorations que nous allons presenter comme suit : Dans un premier temps, nous avons propose d'utiliser l'approche ensemble de classifieurs de reseaux de neurones offrant des reponses probabilistes afin de construire des cartes predictives complementaires pour une meilleure interpretation. Nous avons egalement propose d'utiliser les techniques de determination des " outliers " pour detecter les zones qui ne presentent pas les conditions de mineralisation. Dans un deuxieme temps, nous avons propose des ameliorations de techniques pour la selection de variables, en particulier pour les deux familles de selection Optimal Brain Damage et Optimal Brain Surgeon, et pour des criteres d'evaluation des methodes de selection de variables basees sur l'information mutuelle. Nous avons propose d'utiliser l'approche ensembliste en selection de variables pour determiner les variables les plus pertinentes, hierarchiser de maniere globale l'ensemble des variables, permettre aux experts de choisir la ou les solutions qui apparaissent comme etant les plus comprehensibles, et enfin pour se focaliser uniquement sur une combinaison de regles decrivant le concept-cible issue d'un sous-ensemble pertinent de variables. Les deux premiers objectifs vises ci-dessus ont ete abordes en utilisant les treillis de Galois et l'analyse statistique de distribution.
... INRIA-Lorraine, Campus Scientifique - BP 239 - 54506 Vandœuvre-l`es-Nancy Cedex, France Moham... more ... INRIA-Lorraine, Campus Scientifique - BP 239 - 54506 Vandœuvre-l`es-Nancy Cedex, France Mohammed.Attik@loria.frhttp://www.loria.fr/~attik ... where P eaki represent the height frequencies of preserved features, Mini is the minimum number of preserved features, Maxi is the ...
Les travaux de recherche présentés dans cette thèse rentrent dans le cadre de la problématique de... more Les travaux de recherche présentés dans cette thèse rentrent dans le cadre de la problématique de l'Extraction de Connaissances à partir des Données (ECD) que nous avons illustrée à travers le traitement des données du Système d'Information Géographique établi par le Bureau de Recherches Géologiques et Minières (BRGM). Les objectifs de cette thèse sont d'établir des cartes prédictives de substances, d'avoir une interprétation sur les critères permettant d'établir cette décision, et de fournir une hiérarchisation des attributs descriptifs. Pour atteindre ces objectifs, nous avons proposé étape par étape un ensemble d'améliorations que nous allons présenter comme suit : Dans un premier temps, nous avons proposé d'utiliser l'approche ensemble de classifieurs de réseaux de neurones offrant des réponses probabilistes afin de construire des cartes prédictives complémentaires pour une meilleure interprétation. Nous avons également proposé d'utiliser les ...
Les travaux de recherche presentes dans cette these rentrent dans le cadre de la problematique de... more Les travaux de recherche presentes dans cette these rentrent dans le cadre de la problematique de l'Extraction de Connaissances a partir des Donnees (ECD) que nous avons illustree a travers le traitement des donnees du Systeme d'Information Geographique etabli par le Bureau de Recherches Geologiques et Minieres (BRGM). Les objectifs de cette these sont d'etablir des cartes predictives de substances, d'avoir une interpretation sur les criteres permettant d'etablir cette decision, et de fournir une hierarchisation des attributs descriptifs. Pour atteindre ces objectifs, nous avons propose etape par etape un ensemble d'ameliorations que nous allons presenter comme suit : Dans un premier temps, nous avons propose d'utiliser l'approche ensemble de classifieurs de reseaux de neurones offrant des reponses probabilistes afin de construire des cartes predictives complementaires pour une meilleure interpretation. Nous avons egalement propose d'utiliser les ...
The determination of the optimal architecture of a supervised neural network is an important and ... more The determination of the optimal architecture of a supervised neural network is an important and a difficult task. The classical neural network topology optimization methods select weight(s) or unit(s) from the architecture in order to give a high performance of a learning algorithm. However, all existing topology optimization methods do not guarantee to obtain the optimal solution. In this work, we propose a hybrid approach which combines variable selection method and classical optimization method in order to improve optimization topology solution. The proposed approach suggests to identify the relevant subset of variables which gives a good classification performance in the first step and then to apply a classical topology optimization method to eliminate unnecessary hidden units or weights. A comparison of our approach to classical techniques for architecture optimization is given.
It is the age of the social web, where people express themselves by giving their opinions about v... more It is the age of the social web, where people express themselves by giving their opinions about various issues, from their personal life to the world’s political issues. This process generates a lot of opinion data on the web that can be processed for valuable information, and therefore, semantic annotation of opinions becomes an important task. Unfortunately, existing opinion annotation schemes have failed to satisfy annotation challenges and cannot even adhere to the basic definition of opinion. Opinion holders, topical features and temporal expressions are major components of an opinion that remain ignored in existing annotation schemes. In this work, we propose OpinionML, a new Markup Language, that aims to compensate for the issues that existing typical opinion markup languages fail to resolve. We present a detailed discussion about existing annotation schemes and their associated problems. We argue that OpinionML is more robust, flexible and easier for annotating opinion data....
In this paper, we show how neural networks can be applied to Geographic Information System (GIS) ... more In this paper, we show how neural networks can be applied to Geographic Information System (GIS) for feature selection. Precisely, we want to select the most relevant variables in a classification task between gold deposits and deposits without gold by applying pruning neural methods to a GIS Andes dataset. Two families of pruning methods based on an analysis of the weight saliencies are presented, their strengths and weaknesses are discussed. This work could help in the long term to better understand the formation of gold deposits.
The Semantic Web: ESWC 2015 Satellite Events, 2015
In this paper, we propose SentiML++, an extension of Sen-tiML with a focus on annotating opinions... more In this paper, we propose SentiML++, an extension of Sen-tiML with a focus on annotating opinions answering aspects of the general question "who has what opinion about whom in which context?". A detailed comparison with SentiML and other existing annotation schemes is also presented. The data collection annotated with SentiML has also been annotated with SentiML++ and is available for download for research purpose.
European Conference on Artificial Intelligence, 2004
The determination of the optimal architecture of a multilayer perceptron (MLP) to solve a specifi... more The determination of the optimal architecture of a multilayer perceptron (MLP) to solve a specific problem is an important and a difficult task. Several approaches based on saliency analysis, such as the Optimal Brain Surgeon method (OBS), have been developed in this field. Nevertheless, we show in this paper that OBS does not gives an optimal architecture. We also present the advantages of applying OBS on the architecture obtained by a variable selection method. New hybrid methods are proposed. A comparison of our approaches to standard techniques for architecture optimization is presented.Simulation results obtained on the Monk's problem illustrate the specificities of each method described in this paper.
2004 IEEE International Joint Conference on Neural Networks (IEEE Cat. No.04CH37541)
HAL is a multi-disciplinary open access archive for the deposit and dissemination of scientific r... more HAL is a multi-disciplinary open access archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L'archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d'enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Language models play an important role in large vocabulary speech recognition and statistical mac... more Language models play an important role in large vocabulary speech recognition and statistical machine translation systems. The dominant approach since several decades are back-off language models. Some years ago, there was a clear tendency to build huge language models trained on hundreds of billions of words. Lately, this tendency has changed and recent works concentrate on data selection. Continuous space methods are a very competitive approach, but they have a high computational complexity and are not yet in widespread use. This paper presents an experimental comparison of all these approaches on a large statistical machine translation task. We also describe an open-source implementation to train and use continuous space language models (CSLM) for such large tasks. We describe an efficient implementation of the CSLM using graphical processing units from Nvidia. By these means, we are able to train an CSLM on more than 500 million words in 20 hours. This CSLM provides an improvement of up to 1.8 BLEU points with respect to the best back-off language model that we were able to build.
The solution obtained by Self-Organizing Map (SOM) strongly depends on the initial cluster center... more The solution obtained by Self-Organizing Map (SOM) strongly depends on the initial cluster centers. However, all existing SOM initialization methods do not guarantee to obtain a better minimal solution. Generally, we can group these methods in two classes: random initialization and data analysis based initialization classes. This work proposes an improvement of linear projection initialization method. This method belongs to
... Clustering quality measures for data samples with multiple labels. Authors: Mohammed Attik, L... more ... Clustering quality measures for data samples with multiple labels. Authors: Mohammed Attik, LORIA, Vandœuvre-lès-Nancy, France. Shadi Al Shehabi, LORIA, Vandœuvre-lès-Nancy, France. ... top of page AUTHORS. Mohammed Attik No contact information provided yet. ...
... INRIA-Lorraine, Campus Scientifique - BP 239 - 54506 Vandœuvre-l`es-Nancy Cedex, France Moham... more ... INRIA-Lorraine, Campus Scientifique - BP 239 - 54506 Vandœuvre-l`es-Nancy Cedex, France Mohammed.Attik@loria.frhttp://www.loria.fr/~attik ... where P eaki represent the height frequencies of preserved features, Mini is the minimum number of preserved features, Maxi is the ...
The determination of the optimal architecture of a multilayer perceptron (MLP) to solve a specifi... more The determination of the optimal architecture of a multilayer perceptron (MLP) to solve a specific problem is an important and a difficult task. Several approaches based on saliency analysis, such as the Optimal Brain Surgeon method (OBS), have been developed in this field. Nevertheless, we show in this paper that OBS does not gives an optimal architecture. We also present the advantages of applying OBS on the architecture obtained by a variable selection method. New hybrid methods are proposed. A comparison of our approaches to standard techniques for architecture optimization is presented.Simulation results obtained on the Monk's problem illustrate the specificities of each method described in this paper.
Les travaux de recherche presentes dans cette these rentrent dans le cadre de la problematique de... more Les travaux de recherche presentes dans cette these rentrent dans le cadre de la problematique de l'Extraction de Connaissances a partir des Donnees (ECD) que nous avons illustree a travers le traitement des donnees du Systeme d'Information Geographique etabli par le Bureau de Recherches Geologiques et Minieres (BRGM). Les objectifs de cette these sont d'etablir des cartes predictives de substances, d'avoir une interpretation sur les criteres permettant d'etablir cette decision, et de fournir une hierarchisation des attributs descriptifs. Pour atteindre ces objectifs, nous avons propose etape par etape un ensemble d'ameliorations que nous allons presenter comme suit : Dans un premier temps, nous avons propose d'utiliser l'approche ensemble de classifieurs de reseaux de neurones offrant des reponses probabilistes afin de construire des cartes predictives complementaires pour une meilleure interpretation. Nous avons egalement propose d'utiliser les techniques de determination des " outliers " pour detecter les zones qui ne presentent pas les conditions de mineralisation. Dans un deuxieme temps, nous avons propose des ameliorations de techniques pour la selection de variables, en particulier pour les deux familles de selection Optimal Brain Damage et Optimal Brain Surgeon, et pour des criteres d'evaluation des methodes de selection de variables basees sur l'information mutuelle. Nous avons propose d'utiliser l'approche ensembliste en selection de variables pour determiner les variables les plus pertinentes, hierarchiser de maniere globale l'ensemble des variables, permettre aux experts de choisir la ou les solutions qui apparaissent comme etant les plus comprehensibles, et enfin pour se focaliser uniquement sur une combinaison de regles decrivant le concept-cible issue d'un sous-ensemble pertinent de variables. Les deux premiers objectifs vises ci-dessus ont ete abordes en utilisant les treillis de Galois et l'analyse statistique de distribution.
... INRIA-Lorraine, Campus Scientifique - BP 239 - 54506 Vandœuvre-l`es-Nancy Cedex, France Moham... more ... INRIA-Lorraine, Campus Scientifique - BP 239 - 54506 Vandœuvre-l`es-Nancy Cedex, France Mohammed.Attik@loria.frhttp://www.loria.fr/~attik ... where P eaki represent the height frequencies of preserved features, Mini is the minimum number of preserved features, Maxi is the ...
Les travaux de recherche présentés dans cette thèse rentrent dans le cadre de la problématique de... more Les travaux de recherche présentés dans cette thèse rentrent dans le cadre de la problématique de l'Extraction de Connaissances à partir des Données (ECD) que nous avons illustrée à travers le traitement des données du Système d'Information Géographique établi par le Bureau de Recherches Géologiques et Minières (BRGM). Les objectifs de cette thèse sont d'établir des cartes prédictives de substances, d'avoir une interprétation sur les critères permettant d'établir cette décision, et de fournir une hiérarchisation des attributs descriptifs. Pour atteindre ces objectifs, nous avons proposé étape par étape un ensemble d'améliorations que nous allons présenter comme suit : Dans un premier temps, nous avons proposé d'utiliser l'approche ensemble de classifieurs de réseaux de neurones offrant des réponses probabilistes afin de construire des cartes prédictives complémentaires pour une meilleure interprétation. Nous avons également proposé d'utiliser les ...
Les travaux de recherche presentes dans cette these rentrent dans le cadre de la problematique de... more Les travaux de recherche presentes dans cette these rentrent dans le cadre de la problematique de l'Extraction de Connaissances a partir des Donnees (ECD) que nous avons illustree a travers le traitement des donnees du Systeme d'Information Geographique etabli par le Bureau de Recherches Geologiques et Minieres (BRGM). Les objectifs de cette these sont d'etablir des cartes predictives de substances, d'avoir une interpretation sur les criteres permettant d'etablir cette decision, et de fournir une hierarchisation des attributs descriptifs. Pour atteindre ces objectifs, nous avons propose etape par etape un ensemble d'ameliorations que nous allons presenter comme suit : Dans un premier temps, nous avons propose d'utiliser l'approche ensemble de classifieurs de reseaux de neurones offrant des reponses probabilistes afin de construire des cartes predictives complementaires pour une meilleure interpretation. Nous avons egalement propose d'utiliser les ...
The determination of the optimal architecture of a supervised neural network is an important and ... more The determination of the optimal architecture of a supervised neural network is an important and a difficult task. The classical neural network topology optimization methods select weight(s) or unit(s) from the architecture in order to give a high performance of a learning algorithm. However, all existing topology optimization methods do not guarantee to obtain the optimal solution. In this work, we propose a hybrid approach which combines variable selection method and classical optimization method in order to improve optimization topology solution. The proposed approach suggests to identify the relevant subset of variables which gives a good classification performance in the first step and then to apply a classical topology optimization method to eliminate unnecessary hidden units or weights. A comparison of our approach to classical techniques for architecture optimization is given.
It is the age of the social web, where people express themselves by giving their opinions about v... more It is the age of the social web, where people express themselves by giving their opinions about various issues, from their personal life to the world’s political issues. This process generates a lot of opinion data on the web that can be processed for valuable information, and therefore, semantic annotation of opinions becomes an important task. Unfortunately, existing opinion annotation schemes have failed to satisfy annotation challenges and cannot even adhere to the basic definition of opinion. Opinion holders, topical features and temporal expressions are major components of an opinion that remain ignored in existing annotation schemes. In this work, we propose OpinionML, a new Markup Language, that aims to compensate for the issues that existing typical opinion markup languages fail to resolve. We present a detailed discussion about existing annotation schemes and their associated problems. We argue that OpinionML is more robust, flexible and easier for annotating opinion data....
In this paper, we show how neural networks can be applied to Geographic Information System (GIS) ... more In this paper, we show how neural networks can be applied to Geographic Information System (GIS) for feature selection. Precisely, we want to select the most relevant variables in a classification task between gold deposits and deposits without gold by applying pruning neural methods to a GIS Andes dataset. Two families of pruning methods based on an analysis of the weight saliencies are presented, their strengths and weaknesses are discussed. This work could help in the long term to better understand the formation of gold deposits.
The Semantic Web: ESWC 2015 Satellite Events, 2015
In this paper, we propose SentiML++, an extension of Sen-tiML with a focus on annotating opinions... more In this paper, we propose SentiML++, an extension of Sen-tiML with a focus on annotating opinions answering aspects of the general question "who has what opinion about whom in which context?". A detailed comparison with SentiML and other existing annotation schemes is also presented. The data collection annotated with SentiML has also been annotated with SentiML++ and is available for download for research purpose.
European Conference on Artificial Intelligence, 2004
The determination of the optimal architecture of a multilayer perceptron (MLP) to solve a specifi... more The determination of the optimal architecture of a multilayer perceptron (MLP) to solve a specific problem is an important and a difficult task. Several approaches based on saliency analysis, such as the Optimal Brain Surgeon method (OBS), have been developed in this field. Nevertheless, we show in this paper that OBS does not gives an optimal architecture. We also present the advantages of applying OBS on the architecture obtained by a variable selection method. New hybrid methods are proposed. A comparison of our approaches to standard techniques for architecture optimization is presented.Simulation results obtained on the Monk's problem illustrate the specificities of each method described in this paper.
2004 IEEE International Joint Conference on Neural Networks (IEEE Cat. No.04CH37541)
HAL is a multi-disciplinary open access archive for the deposit and dissemination of scientific r... more HAL is a multi-disciplinary open access archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L'archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d'enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Language models play an important role in large vocabulary speech recognition and statistical mac... more Language models play an important role in large vocabulary speech recognition and statistical machine translation systems. The dominant approach since several decades are back-off language models. Some years ago, there was a clear tendency to build huge language models trained on hundreds of billions of words. Lately, this tendency has changed and recent works concentrate on data selection. Continuous space methods are a very competitive approach, but they have a high computational complexity and are not yet in widespread use. This paper presents an experimental comparison of all these approaches on a large statistical machine translation task. We also describe an open-source implementation to train and use continuous space language models (CSLM) for such large tasks. We describe an efficient implementation of the CSLM using graphical processing units from Nvidia. By these means, we are able to train an CSLM on more than 500 million words in 20 hours. This CSLM provides an improvement of up to 1.8 BLEU points with respect to the best back-off language model that we were able to build.
The solution obtained by Self-Organizing Map (SOM) strongly depends on the initial cluster center... more The solution obtained by Self-Organizing Map (SOM) strongly depends on the initial cluster centers. However, all existing SOM initialization methods do not guarantee to obtain a better minimal solution. Generally, we can group these methods in two classes: random initialization and data analysis based initialization classes. This work proposes an improvement of linear projection initialization method. This method belongs to
... Clustering quality measures for data samples with multiple labels. Authors: Mohammed Attik, L... more ... Clustering quality measures for data samples with multiple labels. Authors: Mohammed Attik, LORIA, Vandœuvre-lès-Nancy, France. Shadi Al Shehabi, LORIA, Vandœuvre-lès-Nancy, France. ... top of page AUTHORS. Mohammed Attik No contact information provided yet. ...
... INRIA-Lorraine, Campus Scientifique - BP 239 - 54506 Vandœuvre-l`es-Nancy Cedex, France Moham... more ... INRIA-Lorraine, Campus Scientifique - BP 239 - 54506 Vandœuvre-l`es-Nancy Cedex, France Mohammed.Attik@loria.frhttp://www.loria.fr/~attik ... where P eaki represent the height frequencies of preserved features, Mini is the minimum number of preserved features, Maxi is the ...
The determination of the optimal architecture of a multilayer perceptron (MLP) to solve a specifi... more The determination of the optimal architecture of a multilayer perceptron (MLP) to solve a specific problem is an important and a difficult task. Several approaches based on saliency analysis, such as the Optimal Brain Surgeon method (OBS), have been developed in this field. Nevertheless, we show in this paper that OBS does not gives an optimal architecture. We also present the advantages of applying OBS on the architecture obtained by a variable selection method. New hybrid methods are proposed. A comparison of our approaches to standard techniques for architecture optimization is presented.Simulation results obtained on the Monk's problem illustrate the specificities of each method described in this paper.
Uploads
Papers by Mohammed Attik