HAL (Le Centre pour la Communication Scientifique Directe), 1998
This paper proposes a neural architecture for a robot to learn how to imitate a sequence of movem... more This paper proposes a neural architecture for a robot to learn how to imitate a sequence of movements performed by another robot or by a human. The main idea is that the imitation process does not need to be given to the system but can emerge from a misinterpretation of the perceived situation at the level of a simple sensorimotor system. We discuss the central position of imitation processes for the understanding of our high level cognitive habilities linked to selfrecognition and to the recognition of the other as something similar to me. Another interesting aspect of this work is that the neural network used for sequences learning is directly inspired from a brain structure called the hippocampus and mainly involved in our memorization capabilities (Banquet & Gaussier 1997).
This paper proposes a neural architecture for a robot to learn how to imitate a sequence of movem... more This paper proposes a neural architecture for a robot to learn how to imitate a sequence of movements performed by another robot or by a human. The main idea is that the imitation process does not need to be given to the system but can emerge from a misinterpretation of the perceived situation at the level of a simple sensorimotor system. We discuss the central position of imitation processes for the understanding of our high level cognitive habilities linked to selfrecognition and to the recognition of the other as something similar to me. Another interesting aspect of this work is that the neural network used for sequences learning is directly inspired from a brain structure called the hippocampus and mainly involved in our memorization capabilities (Banquet & Gaussier 1997).
Parallel Virtual Machine (PVM) and Message Passing Interface (MPI) are the most frequently used t... more Parallel Virtual Machine (PVM) and Message Passing Interface (MPI) are the most frequently used tools for programming according to the message passing paradigm, which is considered one of the best ways to develop parallel applications.
Constantly, the assumption is made that there is an independent contribution of the individual fe... more Constantly, the assumption is made that there is an independent contribution of the individual feature extraction and classifier parameters to the recognition performance. In our approach, the problems of feature extraction and classifier design are viewed together as a single matter of estimating the optimal parameters from limited data. We propose, for the problem of facial recognition, a combination between an Interest Operator based feature extraction technique and a k-NN statistical classifier having the parameters determined using a pattern search based optimization technique. This approach enables us to achieve both higher classification accuracy and faster processing time.
HAL (Le Centre pour la Communication Scientifique Directe), Jan 23, 2007
L'extraction de connaissances à partir de données a pour objet la découverte de connaissances à p... more L'extraction de connaissances à partir de données a pour objet la découverte de connaissances à partir de grandes quantités de données, par des méthodes d'apprentissage automatiques ou semi-automatiques, et l'utilisation industrielle ou opérationnelle de ces connaissances. Nous explorons ici une application concrète de la fouille de données, sur un système de serveur vocal. Nous nous intéressons en particulier à l'étape cruciale de validation de motifs extraits via l'utilisation de mesures de qualité.
This paper presents the experimental results on group dynamics raised in a population of animats ... more This paper presents the experimental results on group dynamics raised in a population of animats generated by the introduction of a learning by imitation mechanism. Our interests consist in validating our animat architecture, discovering the emerging dynamics at the population (or social) level, and trying to understand the implication of such dynamics. Ultimately, we would like to know if some dynamic introduced by a simple imitating behaviour based on Maslow theory could illustrate more complex phenomena that of the classical imitation mechanism.
This paper presents a forecasting method for time series. This method combines the wavelet analys... more This paper presents a forecasting method for time series. This method combines the wavelet analysis and several forecasting techniques such as Artificial Neural Networks (ANN), linear regression and random walk. The proposed method is tested using three real time series: the first contains historical data recorded during eight weeks from a WiMAX network and the other two are based on financial series. It is shown that AI with wavelet analysis can be an efficient and versatile approach in time series prediction for small periods time interval (up to 1 month). For long time interval, the best method used is Linear Regression technique. Also we compared the results obtained using various types of wavelets. The results show that Daubechies 1 (db1) and Reverse biorthogonal 1 (rbio1.1) give the best results.
- Nous présentons ici une application des champs neuronaux pour le contrôle d'un robot que no... more - Nous présentons ici une application des champs neuronaux pour le contrôle d'un robot que nous avons utilisé dans le cadre d'un problème d'apprentissage par imitation [5]. Nous avons utilisé les champs neuronaux pour les commandes motrices, la représentation interne de la perception du mouvement dans l'environnement et pour le choix de cibles à suivre. L'utilisation des champs neuronaux nous a permis d'avoir un comportement de suivi avec une dynamique temporelle continue allié à une capacité de bifurcation.
Ce papier est le fruit d'une collaboration entre roboticiens, economistes et informaticiens. ... more Ce papier est le fruit d'une collaboration entre roboticiens, economistes et informaticiens. Sur la base d'une hypothese commune selon laquelle l'espace est central pour comprendre les strategies individuelles, nous montrons comment instrumentaliser le concept de rationalite limitee. De la robotique et de l'informatique, nous mobilisons les outils de simulation et de modelisation des cartes cognitives. Dotes de reseaux de neurones a deux niveaux, l'un destine a apprendre et reconnaitre des lieux, l'autre a memoriser l'ensemble des chemins suivis le plus souvent pour atteindre un objectif, les agents economiques se trouvent en situation de rationalite limitee : ils voient, decouvrent et apprennent leur environnement, planifient des trajets et resolvent des problemes multi-objectifs en fonction de leur pratique et de leur decouverte de l'espace. On souligne ainsi l'interet de travailler sur des agents cognitifs (dotes d'une carte cognitive) par rapport a des agents purement reactifs.
The performance of image denoising algorithms using the Double Tree Complex Wavelet Transform, DT... more The performance of image denoising algorithms using the Double Tree Complex Wavelet Transform, DT CWT, followed by a local adaptive bishrink filter can be improved by reducing the sensitivity of that filter with the local marginal variance of the wavelet coefficients. In this paper is proposed a solution for the sensitivity reduction based on enhanced diversity. First the advantages and disadvantages of a state-of-the-art denoising solution, based on the association DT CWT -bishrink filter are highlighted. Second a blind noise-suppression method correcting the disadvantages of the bishrink filter, performing a non-linear operation on the data is obtained. Finally, some simulation examples prove the performances of the proposed denoising method.
Les deux images de la première ligne de la figure 4 ont été prises de [2] (il s'agit de la figure... more Les deux images de la première ligne de la figure 4 ont été prises de [2] (il s'agit de la figure 1.1). Elles se trouvent aussi dans [3], formant la première ligne de la figure 10.14, mais l'image de la PSFh manque de cette figure. Les deux images de la deuxième ligne de la figure 4 sont prises de la figure 10.14 de [3]. C'est l'explication pour la différence de contraste entre les images de la première ligne et celles de la deuxième ligne de la figure 4. The Fourier basis achieves such a diagonalization, but the energy of the signal is not concentrated over a small number of coefficients (the basis vectors are not spatially localized), so the Fourier transform is not suitable for any thresholding method. Decimated real wavelet transforms are efficient for satellite image deconvolution but produce artefacts since the transform is not translation invariant. C'est le cas de la TODDE. To avoid these artefacts, the resulting image has to be averaged over all the possible integer translations, which slows down the algorithm. En fait c'est la transformée appelée dans nos anciens rapports la TODORCS.
The goal of this paper is to highlight the particularities of Worldwide Interoperability for Micr... more The goal of this paper is to highlight the particularities of Worldwide Interoperability for Microwave Access (WiMAX) traffic from a long‐range dependence perspective. The application presented in this paper consists in the estimation of the quality of base stations (BS) positioning in the architecture of a WiMAX network. Analysing the traffic of the considered WiMAX network, with an original data mining methodology based on long‐range dependence, we determine which BS have a correct position in the topology of the network and which BS have an incorrect position. These BS must be repositioned when the next session of the network maintenance will take place.
HAL (Le Centre pour la Communication Scientifique Directe), 1998
This paper proposes a neural architecture for a robot to learn how to imitate a sequence of movem... more This paper proposes a neural architecture for a robot to learn how to imitate a sequence of movements performed by another robot or by a human. The main idea is that the imitation process does not need to be given to the system but can emerge from a misinterpretation of the perceived situation at the level of a simple sensorimotor system. We discuss the central position of imitation processes for the understanding of our high level cognitive habilities linked to selfrecognition and to the recognition of the other as something similar to me. Another interesting aspect of this work is that the neural network used for sequences learning is directly inspired from a brain structure called the hippocampus and mainly involved in our memorization capabilities (Banquet & Gaussier 1997).
This paper proposes a neural architecture for a robot to learn how to imitate a sequence of movem... more This paper proposes a neural architecture for a robot to learn how to imitate a sequence of movements performed by another robot or by a human. The main idea is that the imitation process does not need to be given to the system but can emerge from a misinterpretation of the perceived situation at the level of a simple sensorimotor system. We discuss the central position of imitation processes for the understanding of our high level cognitive habilities linked to selfrecognition and to the recognition of the other as something similar to me. Another interesting aspect of this work is that the neural network used for sequences learning is directly inspired from a brain structure called the hippocampus and mainly involved in our memorization capabilities (Banquet & Gaussier 1997).
Parallel Virtual Machine (PVM) and Message Passing Interface (MPI) are the most frequently used t... more Parallel Virtual Machine (PVM) and Message Passing Interface (MPI) are the most frequently used tools for programming according to the message passing paradigm, which is considered one of the best ways to develop parallel applications.
Constantly, the assumption is made that there is an independent contribution of the individual fe... more Constantly, the assumption is made that there is an independent contribution of the individual feature extraction and classifier parameters to the recognition performance. In our approach, the problems of feature extraction and classifier design are viewed together as a single matter of estimating the optimal parameters from limited data. We propose, for the problem of facial recognition, a combination between an Interest Operator based feature extraction technique and a k-NN statistical classifier having the parameters determined using a pattern search based optimization technique. This approach enables us to achieve both higher classification accuracy and faster processing time.
HAL (Le Centre pour la Communication Scientifique Directe), Jan 23, 2007
L'extraction de connaissances à partir de données a pour objet la découverte de connaissances à p... more L'extraction de connaissances à partir de données a pour objet la découverte de connaissances à partir de grandes quantités de données, par des méthodes d'apprentissage automatiques ou semi-automatiques, et l'utilisation industrielle ou opérationnelle de ces connaissances. Nous explorons ici une application concrète de la fouille de données, sur un système de serveur vocal. Nous nous intéressons en particulier à l'étape cruciale de validation de motifs extraits via l'utilisation de mesures de qualité.
This paper presents the experimental results on group dynamics raised in a population of animats ... more This paper presents the experimental results on group dynamics raised in a population of animats generated by the introduction of a learning by imitation mechanism. Our interests consist in validating our animat architecture, discovering the emerging dynamics at the population (or social) level, and trying to understand the implication of such dynamics. Ultimately, we would like to know if some dynamic introduced by a simple imitating behaviour based on Maslow theory could illustrate more complex phenomena that of the classical imitation mechanism.
This paper presents a forecasting method for time series. This method combines the wavelet analys... more This paper presents a forecasting method for time series. This method combines the wavelet analysis and several forecasting techniques such as Artificial Neural Networks (ANN), linear regression and random walk. The proposed method is tested using three real time series: the first contains historical data recorded during eight weeks from a WiMAX network and the other two are based on financial series. It is shown that AI with wavelet analysis can be an efficient and versatile approach in time series prediction for small periods time interval (up to 1 month). For long time interval, the best method used is Linear Regression technique. Also we compared the results obtained using various types of wavelets. The results show that Daubechies 1 (db1) and Reverse biorthogonal 1 (rbio1.1) give the best results.
- Nous présentons ici une application des champs neuronaux pour le contrôle d'un robot que no... more - Nous présentons ici une application des champs neuronaux pour le contrôle d'un robot que nous avons utilisé dans le cadre d'un problème d'apprentissage par imitation [5]. Nous avons utilisé les champs neuronaux pour les commandes motrices, la représentation interne de la perception du mouvement dans l'environnement et pour le choix de cibles à suivre. L'utilisation des champs neuronaux nous a permis d'avoir un comportement de suivi avec une dynamique temporelle continue allié à une capacité de bifurcation.
Ce papier est le fruit d'une collaboration entre roboticiens, economistes et informaticiens. ... more Ce papier est le fruit d'une collaboration entre roboticiens, economistes et informaticiens. Sur la base d'une hypothese commune selon laquelle l'espace est central pour comprendre les strategies individuelles, nous montrons comment instrumentaliser le concept de rationalite limitee. De la robotique et de l'informatique, nous mobilisons les outils de simulation et de modelisation des cartes cognitives. Dotes de reseaux de neurones a deux niveaux, l'un destine a apprendre et reconnaitre des lieux, l'autre a memoriser l'ensemble des chemins suivis le plus souvent pour atteindre un objectif, les agents economiques se trouvent en situation de rationalite limitee : ils voient, decouvrent et apprennent leur environnement, planifient des trajets et resolvent des problemes multi-objectifs en fonction de leur pratique et de leur decouverte de l'espace. On souligne ainsi l'interet de travailler sur des agents cognitifs (dotes d'une carte cognitive) par rapport a des agents purement reactifs.
The performance of image denoising algorithms using the Double Tree Complex Wavelet Transform, DT... more The performance of image denoising algorithms using the Double Tree Complex Wavelet Transform, DT CWT, followed by a local adaptive bishrink filter can be improved by reducing the sensitivity of that filter with the local marginal variance of the wavelet coefficients. In this paper is proposed a solution for the sensitivity reduction based on enhanced diversity. First the advantages and disadvantages of a state-of-the-art denoising solution, based on the association DT CWT -bishrink filter are highlighted. Second a blind noise-suppression method correcting the disadvantages of the bishrink filter, performing a non-linear operation on the data is obtained. Finally, some simulation examples prove the performances of the proposed denoising method.
Les deux images de la première ligne de la figure 4 ont été prises de [2] (il s'agit de la figure... more Les deux images de la première ligne de la figure 4 ont été prises de [2] (il s'agit de la figure 1.1). Elles se trouvent aussi dans [3], formant la première ligne de la figure 10.14, mais l'image de la PSFh manque de cette figure. Les deux images de la deuxième ligne de la figure 4 sont prises de la figure 10.14 de [3]. C'est l'explication pour la différence de contraste entre les images de la première ligne et celles de la deuxième ligne de la figure 4. The Fourier basis achieves such a diagonalization, but the energy of the signal is not concentrated over a small number of coefficients (the basis vectors are not spatially localized), so the Fourier transform is not suitable for any thresholding method. Decimated real wavelet transforms are efficient for satellite image deconvolution but produce artefacts since the transform is not translation invariant. C'est le cas de la TODDE. To avoid these artefacts, the resulting image has to be averaged over all the possible integer translations, which slows down the algorithm. En fait c'est la transformée appelée dans nos anciens rapports la TODORCS.
The goal of this paper is to highlight the particularities of Worldwide Interoperability for Micr... more The goal of this paper is to highlight the particularities of Worldwide Interoperability for Microwave Access (WiMAX) traffic from a long‐range dependence perspective. The application presented in this paper consists in the estimation of the quality of base stations (BS) positioning in the architecture of a WiMAX network. Analysing the traffic of the considered WiMAX network, with an original data mining methodology based on long‐range dependence, we determine which BS have a correct position in the topology of the network and which BS have an incorrect position. These BS must be repositioned when the next session of the network maintenance will take place.
Uploads
Papers by Sorin Moga