Works on semiotic dynamics and language as a complex adaptive system in general has been an important lane of research over the last decade. In this study, the mean-field naming game model developed in the course of the pioneering research programme of Luc Steels and colleagues is modified to include probabilistic word choice based on weighted lists of words, instead of either deterministic or totally random word choice based on (ordered) sets of words. The parameters&rsquo / interaction and this interaction&rsquo / s effect on time of convergence of the system and size of individual lexicons over time are investigated. The classical model is found to be a special case of this proposed model. Additionally, this model has more parameters and a larger state space which provides additional room for tweaking for time- or space-optimization of the convergence process.

O entendimento dos mecanismos de representação e processamento de odores pelo sistema olfatório é uma das questões centrais da neurociência moderna. Os odores são codificados pela circuitaria interna do bulbo olfatório em padrões espaço-temporais refletidos pela atividade de suas células de saída, as células mitrais e tufosas, que transmitem os resultados das computações dessa estrutura inicial de processamento a regiões corticais superiores. A arquitetura das conexões existentes no bulbo olfatório apresenta inibição lateral em duas camadas diferentes de sua estrutura laminar, intermediadas por dois tipos distintos de interneurônios. Na camada glomerular, mais externa, a inibição lateral é mediada pelas células periglomerulares e na camada plexiforme externa, mais interna, a inibição lateral é mediada pelas células granulares. O papel desses dois níveis distintos de inibição lateral e os mecanismos segundo os quais eles atuam moldando os padrões espaço-temporais de resposta do bulbo olfatório a odores diferentes são ainda pouco conhecidos. O objetivo deste trabalho foi construir um modelo de rede neural biologicamente plausível do bulbo olfatório para investigar como dois tipos diferentes de interneurônios, atuando em estágios distintos de processamento, podem contribuir para a discriminação de odores e a coordenação dos padrões de disparo das células mitrais. O modelo de rede construído, com representação de odores pela atividade das células mitrais e baseado nas interações recíprocas entre essas células e os interneurônios inibitórios, mostrou que a inibição gerada pelas células periglomerulares pode melhorar o contraste entre odores similares, facilitando a discriminação de odores, enquanto que a inibição das células granulares atua no refinamento da resposta de saída da informação olfatória. / The understanding of odor representation and processing mechanisms by the olfactory system is one of the central questions of modern neuroscience. Odors are encoded by the olfactory bulb circuitry in terms of spatiotemporal spiking patterns. These are reflected in the activity of the mitral cells, which are the output cells of the olfactory bulb that transmit the information processed in this early structure to higher cortical regions. The architecture of the olfactory bulb connections presents lateral inhibition at two different layers of its laminar structure, mediated by two distinct types of interneurons. In the glomerular layer, lateral inhibition is mediated by periglomerular cells. In the external plexiform layer, lateral inhibition is mediated by granule cells. The role of these two different lateral inhibition levels and the mechanisms whereby they shape the spatial and temporal patterns of the olfactory bulb response to different odors is not well known. The aim of this work was to build a biologically plausible neural network model of the olfactory bulb to investigate how two different types of interneurons, acting at different processing stages, could contribute to odor discrimination and the coordination of the mitral cells spiking patterns. The results of simulations of the network model shown that the inhibition generated by periglomerular cells can provide contrast enhancement and odors discrimination, while the granule cell inhibition can refine the output response of the olfactory information.

O entendimento dos mecanismos de representação e processamento de odores pelo sistema olfatório é uma das questões centrais da neurociência moderna. Os odores são codificados pela circuitaria interna do bulbo olfatório em padrões espaço-temporais refletidos pela atividade de suas células de saída, as células mitrais e tufosas, que transmitem os resultados das computações dessa estrutura inicial de processamento a regiões corticais superiores. A arquitetura das conexões existentes no bulbo olfatório apresenta inibição lateral em duas camadas diferentes de sua estrutura laminar, intermediadas por dois tipos distintos de interneurônios. Na camada glomerular, mais externa, a inibição lateral é mediada pelas células periglomerulares e na camada plexiforme externa, mais interna, a inibição lateral é mediada pelas células granulares. O papel desses dois níveis distintos de inibição lateral e os mecanismos segundo os quais eles atuam moldando os padrões espaço-temporais de resposta do bulbo olfatório a odores diferentes são ainda pouco conhecidos. O objetivo deste trabalho foi construir um modelo de rede neural biologicamente plausível do bulbo olfatório para investigar como dois tipos diferentes de interneurônios, atuando em estágios distintos de processamento, podem contribuir para a discriminação de odores e a coordenação dos padrões de disparo das células mitrais. O modelo de rede construído, com representação de odores pela atividade das células mitrais e baseado nas interações recíprocas entre essas células e os interneurônios inibitórios, mostrou que a inibição gerada pelas células periglomerulares pode melhorar o contraste entre odores similares, facilitando a discriminação de odores, enquanto que a inibição das células granulares atua no refinamento da resposta de saída da informação olfatória. / The understanding of odor representation and processing mechanisms by the olfactory system is one of the central questions of modern neuroscience. Odors are encoded by the olfactory bulb circuitry in terms of spatiotemporal spiking patterns. These are reflected in the activity of the mitral cells, which are the output cells of the olfactory bulb that transmit the information processed in this early structure to higher cortical regions. The architecture of the olfactory bulb connections presents lateral inhibition at two different layers of its laminar structure, mediated by two distinct types of interneurons. In the glomerular layer, lateral inhibition is mediated by periglomerular cells. In the external plexiform layer, lateral inhibition is mediated by granule cells. The role of these two different lateral inhibition levels and the mechanisms whereby they shape the spatial and temporal patterns of the olfactory bulb response to different odors is not well known. The aim of this work was to build a biologically plausible neural network model of the olfactory bulb to investigate how two different types of interneurons, acting at different processing stages, could contribute to odor discrimination and the coordination of the mitral cells spiking patterns. The results of simulations of the network model shown that the inhibition generated by periglomerular cells can provide contrast enhancement and odors discrimination, while the granule cell inhibition can refine the output response of the olfactory information.

Dans le cadre des travaux sur le rôle des connaissances phonologiques dans le processus de reconnaissance de mot écrit, l’objectif de notre thèse est de contribuer à mieux décrire la nature du code impliqué et de comprendre les mécanismes par lesquels les connaissances phonologiques interviennent en lecture. Nous voudrions avant tout défendre l’idée selon laquelle les premières étapes du processus de reconnaissance de mot écrit impliquent un code phonologique suffisamment fin pour être décrit en termes de traits phonologiques, en montrant que les lecteurs sont sensibles à des aspects infra-phonémiques et qu’il existerait une organisation hiérarchique de ces traits. D’autre part, cette thèse vise à apporter des arguments en faveur de l’existence de troubles phonologiques chez les enfants dyslexiques, à préciser certaines anomalies de leurs représentations au niveau infra-phonémique, tout en tenant compte de disparités possibles dans ces anomalies selon le type de dyslexie.Nous présenterons en premier lieu le cadre théorique et les travaux expérimentaux en psychologie cognitive et psycholinguistique nous conduisant à formuler des questions précises à ce propos. La présentation de travaux réalisés en linguistique et en neuropsychologie étayera les hypothèses que nous proposerons ensuite. Dans la partie théorique, nous présenterons notamment une proposition de modèle de lecture articulant deux mécanismes basés sur les traits phonologiques et l’orientation de nos hypothèses sera guidée par les contraintes de ce modèle.Une démarche expérimentale sera développée en plusieurs étapes pour apporter des données comportementales permettant de discuter nos hypothèses à propos d’adultes bons lecteurs, d’enfants présentant un développement ordinaire et d’enfants dyslexiques. / Some studies in literature show the involvement of phonological knowledge in printed words processing. The goal of this work is to contribute to describe the nature of the code involved in printed word recognition, and to understand the mechanisms of phonological knowledge involvement in reading. We firstly want to show that the early steps of printed words recognition involve a phonological code described in terms of phonological features, by showing that skilled readers are sensitive to infra-phonemic elements and that a hierarchical organisation of these features could exist. Otherwise, this work aims to provide clues in favour of the existence of phonological impairments in dyslexic children, to define some deficits of their infra-phonemic organisation, and to show that these deficits could be different according to the type of developmental dyslexia. On the one hand, we present the theory and the experimental studies in cognitive psychology and in psycholinguistics that lead us to some questions. On the other hand, the presentation of studies in linguistics and in neuropsychology provide clues in favour of our hypothesis. In the theorical part, we present a reading model that involve two reading mechanisms based on phonological features, and our hypothesis allow us to test this model. Our experimental word is developped in several steps to provide behavior data that lead us to discuss our hypothesis about adult skilled readers, normal reading children, and dyslexic chilmdren.

A substantial number of works have aimed at modeling the receptive field properties of the primary visual cortex (V1). Their evaluation criterion is usually the similarity of the model response properties to the recorded responses from biological organisms. However, as several algorithms were able to demonstrate some degree of similarity to biological data based on the existing criteria, we focus on the robustness against loss of information in the form of occlusions as an additional constraint for better understanding the algorithmic level of early vision in the brain. We try to investigate the influence of competition mechanisms on the robustness. Therefore, we compared four methods employing different competition mechanisms, namely, independent component analysis, non-negative matrix factorization with sparseness constraint, predictive coding/biased competition, and a Hebbian neural network with lateral inhibitory connections. Each of those methods is known to be capable of developing receptive fields comparable to those of V1 simple-cells. Since measuring the robustness of methods having simple-cell like receptive fields against occlusion is difficult, we measure the robustness using the classification accuracy on the MNIST hand written digit dataset. For this we trained all methods on the training set of the MNIST hand written digits dataset and tested them on a MNIST test set with different levels of occlusions. We observe that methods which employ competitive mechanisms have higher robustness against loss of information. Also the kind of the competition mechanisms plays an important role in robustness. Global feedback inhibition as employed in predictive coding/biased competition has an advantage compared to local lateral inhibition learned by an anti-Hebb rule.

A substantial number of works have aimed at modeling the receptive field properties of the primary visual cortex (V1). Their evaluation criterion is usually the similarity of the model response properties to the recorded responses from biological organisms. However, as several algorithms were able to demonstrate some degree of similarity to biological data based on the existing criteria, we focus on the robustness against loss of information in the form of occlusions as an additional constraint for better understanding the algorithmic level of early vision in the brain. We try to investigate the influence of competition mechanisms on the robustness. Therefore, we compared four methods employing different competition mechanisms, namely, independent component analysis, non-negative matrix factorization with sparseness constraint, predictive coding/biased competition, and a Hebbian neural network with lateral inhibitory connections. Each of those methods is known to be capable of developing receptive fields comparable to those of V1 simple-cells. Since measuring the robustness of methods having simple-cell like receptive fields against occlusion is difficult, we measure the robustness using the classification accuracy on the MNIST hand written digit dataset. For this we trained all methods on the training set of the MNIST hand written digits dataset and tested them on a MNIST test set with different levels of occlusions. We observe that methods which employ competitive mechanisms have higher robustness against loss of information. Also the kind of the competition mechanisms plays an important role in robustness. Global feedback inhibition as employed in predictive coding/biased competition has an advantage compared to local lateral inhibition learned by an anti-Hebb rule.