In a scientific study of dream content, artificial intelligence has been utilized to automatically score dream content. An initial attempt focused on scoring for emotional tone of dream reports. The contribution of this thesis demonstrates methods by which accuracy of such a system can be improved beyond text-mining. It was hypothesized that data extraction based on psychological processes will provide significant information that would produce an accurate model. In our first article, the significance of words expressed in dream reports, along with their associated words was explored. Extraction and inclusion of these associations provided detailed information that improved automatic scoring of positive and negative affect even though these associations exhibited skewed distribution. The second article demonstrated how normalization of the data was possible and how it could result in a more accurate model. Our last article was able to demonstrate that the model can differentiate between male and female dreams.

In a scientific study of dream content, artificial intelligence has been utilized to automatically score dream content. An initial attempt focused on scoring for emotional tone of dream reports. The contribution of this thesis demonstrates methods by which accuracy of such a system can be improved beyond text-mining. It was hypothesized that data extraction based on psychological processes will provide significant information that would produce an accurate model. In our first article, the significance of words expressed in dream reports, along with their associated words was explored. Extraction and inclusion of these associations provided detailed information that improved automatic scoring of positive and negative affect even though these associations exhibited skewed distribution. The second article demonstrated how normalization of the data was possible and how it could result in a more accurate model. Our last article was able to demonstrate that the model can differentiate between male and female dreams.

In a scientific study of dream content, artificial intelligence has been utilized to automatically score dream content. An initial attempt focused on scoring for emotional tone of dream reports. The contribution of this thesis demonstrates methods by which accuracy of such a system can be improved beyond text-mining. It was hypothesized that data extraction based on psychological processes will provide significant information that would produce an accurate model. In our first article, the significance of words expressed in dream reports, along with their associated words was explored. Extraction and inclusion of these associations provided detailed information that improved automatic scoring of positive and negative affect even though these associations exhibited skewed distribution. The second article demonstrated how normalization of the data was possible and how it could result in a more accurate model. Our last article was able to demonstrate that the model can differentiate between male and female dreams.

In a scientific study of dream content, artificial intelligence has been utilized to automatically score dream content. An initial attempt focused on scoring for emotional tone of dream reports. The contribution of this thesis demonstrates methods by which accuracy of such a system can be improved beyond text-mining. It was hypothesized that data extraction based on psychological processes will provide significant information that would produce an accurate model. In our first article, the significance of words expressed in dream reports, along with their associated words was explored. Extraction and inclusion of these associations provided detailed information that improved automatic scoring of positive and negative affect even though these associations exhibited skewed distribution. The second article demonstrated how normalization of the data was possible and how it could result in a more accurate model. Our last article was able to demonstrate that the model can differentiate between male and female dreams.

CI (Continuous Integration) is a software development practice which became more and more popular in last decade. Ericsson followed the trends and used CI several years. Because of the complexity of RBS (Radio Base Station) software few levels of CI have been implemented there. In RCS (RBS Control System) module CI there are many automatic JCAT (Java Common Auto Tester) test loops running every day and some of them failed. This thesis tries to find a way to classify these test failures automatically, so efficiency and lead time can be improved. Two methods are presented and investigated in this report, rule matching and machine learning. After analysis and comparisons rule matching approach is selected because it does not require huge effort in the initial phase and rule matched data can be used as labeled data for machine learning. This approach requires manual work to add new rules continuously but with correctly defined rules the accuracy is 100%, if the rule is general it can classify one type of issue including the ones which never happen before. One analysis system is designed and implemented, and only small update is required to the result report block of the CI flow. One matching example is showed and according to estimation this method could save many man hours every year.

Made available in DSpace on 2014-06-12T15:56:44Z (GMT). No. of bitstreams: 2 arquivo2985_1.pdf: 5101466 bytes, checksum: 86a375e15b77076e9eb9adffbe664c52 (MD5) license.txt: 1748 bytes, checksum: 8a4605be74aa9ea9d79846c1fba20a33 (MD5) Previous issue date: 2010 / Conselho Nacional de Desenvolvimento Científico e Tecnológico / Uma linha de produtos de software (LPS) é definida como um conjunto de sistemas de software que compartilham características em comum, mas que são suficientemente distintos entre si, desenvolvidos a partir de um conjunto de artefatos reusáveis. Modelos de features e configuração são usados para possibilitar a geração automática de produtos a partir destes artefatos. Um modelo de features representa o conjunto de possíveis configurações de produto de uma LPS, enquanto o modelo de configuração estabelece o mapeamento entre features e implementação. Por exemplo, associando expressões de features, na forma de proposições lógicas, a artefatos. Os benefícios de produtividade que a abordagem de LPS fornece tornam possível que uma LPS seja capaz de gerar milhares de produtos. Neste contexto, erros cometidos ao especificar o modelo de configuração podem resultar em produtos inválidos - o problema da composição segura. Este problema pode ser difícil de ser detectado manualmente, já que os modelos de features e configuração podem tornar-se muito complexos. Gerar todos os produtos de uma LPS pode não ser prático, dado que existem LPS em que é possível gerar milhares de produtos. No entanto, mesmo modelos de configuração que não permitem a geração de produtos inválidos podem ter problemas na sua estrutura interna, como complexidade e duplicação, especialmente no contexto de LPS grandes, onde sua manutenção pode se tornar difícil. Precisamos nos certificar de que não introduzimos erros ao corrigir estes problemas. Neste trabalho, é proposta uma abordagem automática de verificação de composição segura para LPS baseadas em modelos de configuração. Esta abordagem é baseada na tradução de instâncias específicas de modelos de features e configuração em lógica proposicional, usando uma teoria codificada com Alloy. O suporte ferramental fornecido pelo Alloy Analyzer auxilia a verificação. Também é proposto um catálogo de refatoramentos simples para modelos de configuração, como uma maneira de evitar erros ao corrigir problemas na estrutura interna de tais modelos. Este catálogo é formalizado usando uma teoria geral para modelos de configuração especificada com o Prototype Verification System (PVS). Nós avaliamos a abordagem de verificação usando sete versões de uma LPS, com modelos de features que possibilitam a geração de até 272 produtos. Os resultados demonstram a vantagem de usar esta abordagem ao invés de gerar todos os produtos da LPS, já que o tempo médio para compilar um único produto da LPS é maior que o tempo para analisá-la na maior das versões analisadas. Também avaliamos o catálogo de refatoramento provando consistência (soundness) dos refatoramentos propostos no provador de teoremas de PVS

La fibrillation atriale est la plus fréquente des arythmies cardiaques et sa physiopathologie reste mal comprise. Il n’existe pas actuellement de modèle animal dans lequel cette arythmie survienne de manière spontanée. Comme l’âge et l’hypertension artérielle sont deux facteurs de risque majeurs dela fibrillation atriale chez l’Homme, nous avons recherché si le rat spontanément hypertendu (SHR)vieillissant (11-13 mois d’âge) développe cette pathologie. Des enregistrements télémétriques de 24 h de l’électrocardiogramme (ECG) ont été réalisés dans 4 groupes de rats vigils. Ils ont révélé chez le rat SHR vieillissant la présence de nombreux épisodes de tachyarythmie atriale non soutenus, épisodes qui ne sont pas observés chez le rat SHR jeune ni chez lerat normotendu Wistar Kyoto jeune ou vieillissant. Les résultats rapportés ci-dessus ont été obtenus grâce à l’analyse visuelle de l’intégralité de chaque enregistrement ECG, ce qui représente une durée d’analyse incompatible avec des expérimentations de routine. Nous avons donc choisi de développer une méthode automatique de reconnaissance des arythmies atriales chez le rat vigil. Cette méthode utilise exclusivement les intervalles RR car les ondes P ne peuvent pas toujours être identifiées avec fiabilité dans les enregistrements télémétriques de l’ECG. Elle repère les motifs caractéristiques de tachyarythmie grâce à une structure d’automate, et présente une bonne performance sur les enregistrements télémétriques déjà réalisés. Le nouveau modèle animal, conjointement à la nouvelle méthode automatique d’analyse de l’ECG, devrait permettre de réaliser les analyses phénotypiques de qualité indispensables aux hypothèses physiopathologiques moléculaires actuellement avancées dans la fibrillation atriale. / Atrial fibrillation is the most common cardiac arrhythmia and its pathophysiology is still poorly understood. To date, there is no animal model in which the arrhythmia occurs spontaneously. As ageand hypertension are major risk factors of atrial fibrillation in humans, we examined whether the aging (11-13 months of age) spontaneously hypertensive rat (SHR) develops this pathology. Telemetric recordings of electrocardiogram (ECG) were obtained in four groups of conscious rats. They revealed the presence in the aging SHR of many non-sustained atrial tachyarrhythmic episodes that were observed neither in the young SHR nor in the young or aging normotensive Wistar Kyotorat. These results were obtained by visual examination of the whole ECG recordings, which is extremely time consuming and thus, not compatible with routine experiments. Therefore, we chose to develop an automated method to detect atrial arrhythmias in the conscious rat. This method only uses RR intervalsbecause P waves cannot always be accurately identified in telemetric recordings. It recognizes characteristicpatterns of tachyarrhythmia by using an automaton structure, and shows a good performance when applied to previously collected telemetric data. The new animal model, together with the new, automated method of ECG analysis, are expected to be extremely useful for producing high quality phenotypic data, which are necessary to test present molecular hypotheses of atrial fibrillation pathophysiology.

Purpose: To model the time evolution of active caspase-3 protein expression in a healthy lens, and in a lens exposed to UVR-300 nm (UVR-B). To develop an automated method to classify the fluorescent signal of biomarkers in the lens epithelial cells. Methods: Six-week old Sprague-Dawley rats were used. Firstly, expression of active caspase-3 was studied in the lens epithelium of healthy rats. Secondly, rats were unilaterally exposed in vivo to 1 kJ/m2 UVR-B for 15 minutes. At 0.5, 8, 16, and 24 hours after the UVR-B exposure, the exposed and the contralateral non-exposed lenses were removed. Immunohistochemistry was done on three mid-sagittal sections from each lens. The florescent labelling for active caspase-3 in each lens section was counted three times. The time evolution of active caspase-3 expression in response to UVR-B exposure was modelled as a function of cell position in the lens epithelium. An automated objective method was developed to quantify the lens epithelial cells and to classify the fluorescent signal of active caspase-3. Active caspase-3 was selected as a model signal. Results: Active caspase-3 was abundant in the anterior pole of the normal lenses. Spatial distribution of active caspase-3 labelling in the lens epithelium was fitted to a logistic model. The probability of active caspase-3 expression was higher in the UVR-B exposed lenses (95% CI = 0.12 ± 0.01). There was no difference in the expression of active caspase-3 between the 0.5 and the 24 hours groups or between the 8 and the 16 hours groups. A difference was noted, when comparing the 0.5 and 24 hours groups with the 8 and 16 hours groups (Test statistic 7.01, F1;36;0.95= 4.11). Exposure to UVR-B has an impact on the average probability of labelling for active caspase-3 as a function of cell position. The probability of labelling as a function of cell number also varied as a function of time after UVR-B exposure. The automated method counted the lens epithelial cells and estimated the proportion of active caspase-3 labelling in the lens epithelium. Conclusions: Active caspase-3 is present in the healthy lens epithelial cells. Active caspase-3 exhibits higher expression at the anterior pole of the lens and the expression decreases towards the periphery. After UVR-B exposure, the expression of active caspase-3 in the lens epithelium increases with a peak of expression occurring around 16 hours after exposure. The average probability of labelling in the lens epithelium is dependent on both the UVR-B exposure and the time period elapsed after the exposure. The automated method enables objective and fast quantification of lens epithelial cells and the expression of fluorescent signal in the lens cells.

Cette thèse traite du domaine de la cryptographie basée sur des dynamiques chaotiques hybrides.Afin de robustifier la transmission sécurisée de données à l'égard de l'attaque à texte-claire connue, ce travail a été particulièrement axée sur deux directions, l'approche statistique, et l'approche automatique.Les principales contributions de ce travail sont organisées dans ces deux directions.Le choix de la variable mesurée et son influence sur l'émetteur d'un message secret et la possibilité de récupérer la dynamique à la réception.Ceci a été étudié dans le contexte des systèmes chaotiques discrets et continus.L'indépendance statistique des variables d'état des systèmes chaotiques est étudié en relation avec la non-corrélation spatiale de ces états.Ainsi une méthode pour cacher le message secret en fonction de l'évolution de l'émetteur chaotique, et ceci avant son inclusion dans cette dynamique, est proposée.La faisabilité d'un système retardée hybride qui est utilisée pour la transmission sécurisée des données est analysée dans une mise en œuvre analogique.Des simulations et les analyses des résultats obtenus sont faits, afin de prouver l'efficacité des études et des méthodes proposées.La thèse est organisée comme suit: le Chapitre I reprend les notions théoriques et les algorithmes utilisés pour atteindre l'objectif de ce travail.Le chapitre II est consacré à l'étude des exposants de Lyapunov.Les systèmes chaotiques utilisés dans le présent document sont ensuite décrits.Le chapitre III présente une étude de certaines propriétés structurales des systèmes du chapitre II.L'étude se concentre sur le calcul des indices d'observabilité et la détermination des hypersurfaces de la singularité d'observabilité.Le chapitre IV analyse l'indépendance statistique dans le contexte des systèmes chaotiques considérés:la taille de la distance d'échantillonnage (combien d'itérations ou de manière équivalente, combien de temps) pour assurer l'indépendance statistique entre les variables extraites des systèmes chaotiques.Un test original pour l'indépendance statistique (le test Badea-Vlad) a été utilisée; la procédure est applicable à tous les types de variables aléatoires continues, même reparties selon une loi de probabilité inconnue au besoin ici.Le chapitre V illustre le point de vue physique. Le temps transitoire correspond au temps passé par le système chaotique dans le bassin d'attraction avant de rejoindre l'attracteur étrange.De même il est important de savoir après combien de temps les points localisés dans une certaine région de l'attracteur étrange devient non-corrélés.Dans le chapitre VI, sachant l'identifiabilité des paramètres des systèmes chaotiques décrits par des équations polynomiales, une amélioration des inclusions du message dans ce type de cryptographie, est proposé.Le message clair est chiffré en utilisant une substitution classique avec boîtes de transposition, avant son inclusion dans l'émetteur chaotique.Les résultats de l'algorithme proposé sont évalués sur le texte et sur l'image.Le chapitre VII pose quelques questions, et essaie de trouver quelques-unes des réponses à ces questions, dans le cadre du schéma hybride.Comme par exemple, est-il possible de récupérer le message secret en utilisant un observateur, lorsque la dynamique qui lui inclut est retardée?La réponse est positive, et cela est montrée dans le cas d'une transmission intégrale de la sortie du système.Il est important de mentionner que ce travail est pluridisciplinaire, allant de la théorie du contrôle aux statistiques en passant par les domaines de l'électronique, de la mathématique et de l'informatique. / This Thesis deals with the domain of cryptography based on hybrid chaotic dynamics.In order to increase the robustness of the security in data transmission with respect to known text attack, this work was particularly focused on two directions: the statistical approach and the automation control.The main contributions of this work are organized in the mentioned two directions.The choice of the measured variable and its influence on the transmitter of plain messages, alongside the possibility to recover the dynamics at the reception.These are studied in the context of discrete and continuous-time chaotic systems.Statistical independence of the state variables of chaotic systems is investigated in relation with the spatial non-correlation of the states.A method of hiding the secret message, depending on the evolution of the chaotic transmitter and prior to its inclusion in this dynamics is proposed.The feasibility of a delayed time hybrid scheme that is used for secure data transmission is shown in an analog implementation.Simulations and analysis of the obtained results are done in order to prove the efficiency of the proposed studies and methods.The Thesis is organized as follows: Chapter I resumes theoretical notions and algorithms used to achieve the goal of this work.Chapter II is dedicated to the study of the Lyapunov exponents. The chaotic systems used in this report are described.Chapter III presents a study of some structural properties of the chaotic systems from Chapter II.The investigation is focused on the calculation of the observability indexes and the determination of the manifolds of observability singularity.Chapter IV analyses the statistical independence in the context of the considered chaotic systems:how large should be the sampling distance (how many iterations or, equivalently, time) to ensure statical independence between variables extracted from the chaotic systems.An original test for statistical independence (the Badea-Vlad test) was used; the procedure is applicable to all kind of continuous random variables, even of unknown probability law as needed here.Chapter V illustrates the physical point of view.The transient time corresponds to the time spent by the chaotic system in the basin of attraction before rejoining the strange attractor.It is also important to know after how long the points localized in a certain region of the strange attractor become uncorrelated.In Chapter VI, knowing the identifiability of the parameters of chaotic systems described by polynomial equations, an improvement of the inclusion of messages in this type of enciphering is proposed.The plain-message is enciphered using classical substitution and transposition boxes, prior to its inclusion in the chaotic transmitter.The results of the proposed algorithm are evaluated on text and image.Chapter VII rises some questions, and tries to find some answers to these questions, in the context of hybrid dynamical schemes.As for example if it is possible to recover the secret message by using an observer, when the dynamics that includes it is time-delayed.The answer is positive and this is shown in the case of a full transmission of the output of the system.It is important to mention that this work is multidisciplinary, starting from control theory and going to the statistical methods through the fields of electronics, mathematics and computing.

Etudier les lieux et les hommes décrits dans une œuvre littéraire permet de mettre au jour l'intensité de la relation qui lie des habitants à leur espace de vie. A travers le concept de territoire et à l'aide d'un logiciel d'analyse automatique de discours, cette recherche propose une analyse de la Provence que Frédéric Mistral évoque dans Calendal (1867). Une trame conceptuelle du territoire, élaborée à partir des définitions de géographes actuels, nous aide à faire émerger les composantes matérielles et immatérielles du territoire provençal que l'auteur a voulu livrer, en particulier de la fin du 18ème siècle. Dans cette tâche, une attention particulière est accordée aux couples appropriation/territorialisation et identité/territorialité. De cette approche originale de l'œuvre et de ce regard porté sur l'auteur, tous deux choisis pour leur indissociabilité à la Provence, naissent de nouvelles perspectives de compréhension des pratiques territoriales d'une population attachée à un lieu et à une époque donnés. / The study of the places and people described in a literary work enables us to reveal the intensity of the relationship that links people to the place they live in. Through the concept of territory and using software for an automatic analysis of discourse, this research provides a survey of Provence as evoked by Frédéric Mistral in Calendal (1867). A conceptual environment of the country – developed from the definitions of present-time geographers –, helps us to bring out the tangible and intangible components of the provençal territory that the author wished to convey, especially in the late 18th century. In this work, special attention is given to the couples : “appropriation / territorialization” and “identity / territoriality”. From this original approach to the poem and its author, both chosen for their close links to Provence, are born new perspectives for a better understanding of the local pratices of a population attached to a place, at a given period.