Le stress en début de vie (ELS) est associé à une susceptibilité accrue au développement de troubles liés au stress, tels que le trouble dépressif majeur (TDM). L'amygdale latérale (AL), une région du cerveau importante pour la régulation des comportements émotionnels et cognitifs, est vulnérable aux effets du ELS. Cependant, les mécanismes par lesquels l'ELS altère le comportement ne sont pas très bien définis. Auparavant, de nombreuses études se sont concentrées sur les mécanismes neuronaux qui sous-tendent les troubles comportementaux induits par le stress, mais le rôle des cellules gliales dans ce circuit reste indéterminé. Pourtant, les astrocytes, un type de cellule gliale, sont des déterminants clés du comportement. Nous avons donc cherché à identifier le rôle des astrocytes dans les effets de l'ELS sur le comportement dépendant de l'AL. Pour ce faire, nous avons utilisé un modèle de rongeur avec séparation maternelle, limitation de la litière et de la nidification pour reproduire les effets de l'ELS sur le cerveau en développement afin d’évaluer ses effets à long terme sur les astrocytes et le comportement dépendant de l'amygdale latérale. Bien que l'ELS n'ait pas eu d'influence sur le comportement anxieux des souris, ce dernier a altéré de manière significative la détection des menaces, un processus cognitif qui implique la capacité de distinguer avec précision un son menaçant précédemment appris (le stimulus conditionné) d'un son non menaçant dans un contexte nouveau. De plus, la diminution de la sensibilité au stress des astrocytes par la suppression des récepteurs glucocorticoïdes astrocytaires a amélioré de manière significative la fonction cognitive chez les souris ELS et naïves. Globalement, nos résultats suggèrent que les astrocytes jouent un rôle central dans la régulation des effets de l'ELS sur les troubles cognitifs. Ces données soulignent l'importance des astrocytes comme cibles thérapeutiques potentielles pour atténuer le dysfonctionnement cognitif, un symptôme omniprésent de la psychopathologie. / Early Life Stress (ELS) is associated with an enhanced susceptibility to the development of stress-related disorders, such as major depressive disorder (MDD). The lateral amygdala (LA), a brain region important for the regulation of emotive and cognitive behaviours is vulnerable to the effects of ELS. However, the mechanisms by which ELS impairs behaviour are poorly defined. Previously, research has focused on the neuronal mechanisms underlying stress-induced behavioural impairments, however the role of glial cells in this circuitry remains undetermined. Astrocytes, a type of glial cell, are key determinants of behaviour. Hence, we aimed to identify the role of astrocytes in the effects of ELS on LA-dependent behaviour. To accomplish this, we used a rodent model of maternal separation and limited bedding and nesting to replicate the effects of ELS on the developing brain by assessing its long-term effects on astrocytes and lateral-amygdala dependent behaviour. Although ELS did not influence anxiety-like behaviour in mice, ELS significantly impaired threat-detection, a cognitive process involving the ability to accurately distinguish between a previously learned threatening tone (the conditioned stimulus) and a non-threatening tone in a novel context. Additionally, decreasing astrocyte stress sensitivity by deleting astrocyte glucocorticoid receptors significantly enhanced cognitive function in both ELS and naïve mice. Overall, our results suggest that astrocytes are pivotal in the regulation of the effects of ELS on cognitive impairment. This data highlights the importance of astrocytes as potential therapeutic targets for mitigating cognitive dysfunction, a pervasive symptom of psychopathology.

Object tracking can be done in numerous ways, where the goal is to track a target through all frames in a sequence. The ground truth bounding box is used to initialize the object tracking algorithm. Object tracking can be carried out on infrared imagery suitable for military applications to execute tracking even without illumination. Objects, such as aircraft, can deploy countermeasures to impede tracking. The countermeasures most often mainly impact one wavelength band. Therefore, using two different wavelength bands for object tracking can counteract the impact of the countermeasures. The dataset was created from simulations. The countermeasures applied to the dataset are flares and Directional Infrared Countermeasures (DIRCMs). Different object tracking algorithms exist, and many are based on discriminative correlation filters (DCF). The thesis investigated the DCF-based trackers STRCF and ECO on the created dataset. The STRCF and the ECO trackers were analyzed using one and two wavelength bands. The following features were investigated for both trackers: grayscale, Histogram of Oriented Gradients (HOG), and pre-trained deep features. The results indicated that the STRCF and the ECO trackers using two wavelength bands instead of one improved performance on sequences with countermeasures. The use of HOG, deep features, or a combination of both improved the performance of the STRCF tracker using two wavelength bands. Likewise, the performance of the ECO tracker using two wavelength bands was improved by the use of deep features. However, the negative aspect of using two wavelength bands and introducing more features is that it resulted in a lower frame rate.

Les moteurs de recherche font partie de notre vie quotidienne. Actuellement, plus d’un tiers de la population mondiale utilise l’Internet. Les moteurs de recherche leur permettent de trouver rapidement les informations ou les produits qu'ils veulent. La recherche d'information (IR) est le fondement de moteurs de recherche modernes. Les approches traditionnelles de recherche d'information supposent que les termes d'indexation sont indépendants. Pourtant, les termes qui apparaissent dans le même contexte sont souvent dépendants. L’absence de la prise en compte de ces dépendances est une des causes de l’introduction de bruit dans le résultat (résultat non pertinents). Certaines études ont proposé d’intégrer certains types de dépendance, tels que la proximité, la cooccurrence, la contiguïté et de la dépendance grammaticale. Dans la plupart des cas, les modèles de dépendance sont construits séparément et ensuite combinés avec le modèle traditionnel de mots avec une importance constante. Par conséquent, ils ne peuvent pas capturer correctement la dépendance variable et la force de dépendance. Par exemple, la dépendance entre les mots adjacents "Black Friday" est plus importante que celle entre les mots "road constructions". Dans cette thèse, nous étudions différentes approches pour capturer les relations des termes et de leurs forces de dépendance. Nous avons proposé des méthodes suivantes: ─ Nous réexaminons l'approche de combinaison en utilisant différentes unités d'indexation pour la RI monolingue en chinois et la RI translinguistique entre anglais et chinois. En plus d’utiliser des mots, nous étudions la possibilité d'utiliser bi-gramme et uni-gramme comme unité de traduction pour le chinois. Plusieurs modèles de traduction sont construits pour traduire des mots anglais en uni-grammes, bi-grammes et mots chinois avec un corpus parallèle. Une requête en anglais est ensuite traduite de plusieurs façons, et un score classement est produit avec chaque traduction. Le score final de classement combine tous ces types de traduction. Nous considérons la dépendance entre les termes en utilisant la théorie d’évidence de Dempster-Shafer. Une occurrence d'un fragment de texte (de plusieurs mots) dans un document est considérée comme représentant l'ensemble de tous les termes constituants. La probabilité est assignée à un tel ensemble de termes plutôt qu’a chaque terme individuel. Au moment d’évaluation de requête, cette probabilité est redistribuée aux termes de la requête si ces derniers sont différents. Cette approche nous permet d'intégrer les relations de dépendance entre les termes. Nous proposons un modèle discriminant pour intégrer les différentes types de dépendance selon leur force et leur utilité pour la RI. Notamment, nous considérons la dépendance de contiguïté et de cooccurrence à de différentes distances, c’est-à-dire les bi-grammes et les paires de termes dans une fenêtre de 2, 4, 8 et 16 mots. Le poids d’un bi-gramme ou d’une paire de termes dépendants est déterminé selon un ensemble des caractères, en utilisant la régression SVM. Toutes les méthodes proposées sont évaluées sur plusieurs collections en anglais et/ou chinois, et les résultats expérimentaux montrent que ces méthodes produisent des améliorations substantielles sur l'état de l'art. / Search engine has become an integral part of our life. More than one-third of world populations are Internet users. Most users turn to a search engine as the quick way to finding the information or product they want. Information retrieval (IR) is the foundation for modern search engines. Traditional information retrieval approaches assume that indexing terms are independent. However, terms occurring in the same context are often dependent. Failing to recognize the dependencies between terms leads to noise (irrelevant documents) in the result. Some studies have proposed to integrate term dependency of different types, such as proximity, co-occurrence, adjacency and grammatical dependency. In most cases, dependency models are constructed apart and then combined with the traditional word-based (unigram) model on a fixed importance proportion. Consequently, they cannot properly capture variable term dependency and its strength. For example, dependency between adjacent words “black Friday” is more important to consider than those of between “road constructions”. In this thesis, we try to study different approaches to capture term relationships and their dependency strengths. We propose the following methods for monolingual IR and Cross-Language IR (CLIR): We re-examine the combination approach by using different indexing units for Chinese monolingual IR, then propose the similar method for CLIR. In addition to the traditional method based on words, we investigate the possibility of using Chinese bigrams and unigrams as translation units. Several translation models from English words to Chinese unigrams, bigrams and words are created based on a parallel corpus. An English query is then translated in several ways, each producing a ranking score. The final ranking score combines all these types of translations. We incorporate dependencies between terms in our model using Dempster-Shafer theory of evidence. Every occurrence of a text fragment in a document is represented as a set which includes all its implied terms. Probability is assigned to such a set of terms instead of individual terms. During query evaluation phase, the probability of the set can be transferred to those of the related query, allowing us to integrate language-dependent relations to IR. We propose a discriminative language model that integrates different term dependencies according to their strength and usefulness to IR. We consider the dependency of adjacency and co-occurrence within different distances, i.e. bigrams, pairs of terms within text window of size 2, 4, 8 and 16. The weight of bigram or a pair of dependent terms in the final model is learnt according to a set of features. All the proposed methods are evaluated on several English and/or Chinese collections, and experimental results show these methods achieve substantial improvements over state-of-the-art baselines.

Made available in DSpace on 2014-12-17T15:37:03Z (GMT). No. of bitstreams: 1 DiegoSS_DISSERT.pdf: 636795 bytes, checksum: 620ad21d2b47550b5781855775d7f30a (MD5) Previous issue date: 2011-05-17 / Coordena??o de Aperfei?oamento de Pessoal de N?vel Superior / Memory and anxiety are related phenomena. Several evidences suggest that anxiety is fundamental for learnining and may facilitate or impair the memory formation process depending of the context. The majority of animal studies of anxiety and fear use only males as experimental subjects, while studies with females are rare in the literature. However, the prevalence in phobic and anxiety disorders is greater in women than in men. Moreover, it is known that gender maybe influence benzodiazepine effects, the classic drugs used for anxiety disorders treatment. In this respect, to further investigate if fear/anxiety aspects related to learning in female subjects would contribute to the study of phobic and anxiety disorders and their relationship with learning/memory processes, the present work investigates (a) the effects of benzodiazepine diazepam on female rats performance in a aversive memory task that assess concomitantly anxiety/emotionality, as the interaction between both; (b) the influence of estrous cycle phases of female rats on diazepam effects at aversive memory and anxiety/emotionality, and the interaction between both and (c) the role of hormonal fluctuations during estrous cycle phases in absence of diazepam effects in proestrus, because female rats in this phase received or not mifepristone, the antagonist of progesterone receptor, previously to the diazepam treatment. For this purpose, the plus maze discriminative avoidance task, previously validated for studies of anxiety concomitantly to learning/memory, was used. The apparatus employed is an adaptation of a conventional plus maze, with two opens arms and two closed arms, one of which presenting aversive stimulation (noise and light). The parameters used were: time in non-aversive arm compared to time in aversive and percentage of time in aversive arm on several temporal divisions, in order to evaluate memory; percentage of time in open arms, risk assessment, head dipping and end exploring to evaluate anxiety ; and distance traveled for locomotion. In experiment I, we found anxiolytic effect of diazepam only for 4 mg/kg dose, however the amnestic effect appear at a dose of 2 mg/kg. In second experiment, rats were divided in groups according estrous cycle phase (metaestrus/diestrus, proestrus e estrus). In this experiment, when we considered estrous cycle phase or diazepam treatment, the results did not demonstrate any differences in anxiety/emotionality parameters. The amnestic effects of diazepam occur in female rats in metestrus/diestrus and estrus and is absent in proestrous rats. Proestrous female rats that received mifepristone exhibited the amnestic effect of diazepam and also anxiolytic effects, that it was not previously observed in this dose. The results have demonstrated dissociation of anxiolytic and amnestic diazepam effects, not previously observed in males; the absence of amnestic effect of diazepam in proestrous phase; and the possible role of progesterone in aversive memory over diazepam effect, because the mifepristone, associated with diazepam, caused amnestic effect in proestrus / A mem?ria e a ansiedade s?o fen?menos relacionados. Diversas evid?ncias sugerem que a ansiedade ? fundamental para o aprendizado, podendo facilitar ou prejudicar a forma??o de mem?rias dependendo da situa??o, o que se constitui num fator relevante tanto para o funcionamento normal dos processos cognitivos quanto para a compreens?o dos transtornos de ansiedade. A maioria dos estudos com modelos animais que se prop?e a estudar medo e ansiedade usa machos como sujeitos experimentais existindo, assim, escassez no estudo de f?meas na literatura. Entretanto, a preval?ncia para transtornos f?bico-ansiosos ? maior em mulheres do que em homens. Al?m disso, sabe-se que o g?nero pode influenciar o efeito de benzodiazep?nicos, f?rmacos classicamente utilizados no tratamento de transtornos de ansiedade. No intuito de contribuir para o estudo de transtornos f?bico-ansiosos e sua rela??o com processos de mem?ria e aprendizado, o presente trabalho investigou (a) os efeitos do benzodiazep?nico diazepam sobre o desempenho de ratas em uma tarefa de mem?ria aversiva com concomitante avalia??o da ansiedade/emocionalidade; (b) a influ?ncia das fases do ciclo estral de ratas no efeito do diazepam na ansiedade/emocionalidade e mem?ria aversiva, assim como a intera??o entre ambas e (c) o papel de flutua??es hormonais ao longo das fases do ciclo sobre aus?ncia do efeito do diazepam no proestro, pois ratas nessa fase receberam ou n?o o antagonista do receptor da progesterona, mifepristona, previamente ao tratamento com diazepam. Para isso, foi utilizado o modelo da esquiva discriminativa em labirinto cruz elevado, previamente validado para estudos envolvendo ansiedade e aprendizagem. O aparato utilizado ? uma adapta??o do labirinto em cruz elevado convencional, constitu?do de dois bra?os abertos e dois bra?os fechados sendo que um dos fechados tem uma estimula??o aversiva com som e luz. Foram utilizados os par?metros: tempo no bra?o n?o-aversivo comparado ao tempo no aversivo e percentual de tempo no bra?o aversivo em diferentes divis?es temporais para avaliar mem?ria; percentual de tempo nos bra?os abertos, avalia??o de risco, mergulhos de cabe?a e explora??o da ponta do bra?o aberto para ansiedade ; e dist?ncia percorrida para locomo??o. A partir da curva dose-resposta, no primeiro experimento, observamos o efeito ansiol?tico (4mg/kg) e amn?stico (2mg/kg) do diazepam. No segundo experimento, as ratas foram separadas de acordo com as fases do ciclo estral (metaestro/diestro, proestro e estro). N?o foram observadas diferen?as significativas na ansiedade/emocionalidade, nem entre fases do ciclo, nem do tratamento com diazepam (2mg/kg). O efeito amn?stico do diazepam ocorreu nas ratas em metaestro/diestro e estro, estando ausente nas ratas em proestro. Na presen?a da mifepristona as ratas em proestro exibiram o efeito amn?stico do diazepam e tamb?m passaram a apresentar efeito ansiol?tico, o qual n?o havia sido observado previamente nesta dose. Os resultados demonstraram dissocia??o de efeitos amn?sticos e ansiol?ticos em f?meas, n?o previamente observada em machos; aus?ncia do efeito amn?stico do diazepam no proestro, que ocorre nas outras fases e o poss?vel papel da progesterona na mem?ria aversiva sob efeito do diazepam, uma vez que a mifepristona possibilitou o efeito amn?stico no proestro, fase na qual os n?veis de progesterona est?o elevados

De multiples problèmes en apprentissage automatique consistent à minimiser une fonction lisse sur un espace euclidien. Pour l’apprentissage supervisé, cela inclut les régressions par moindres carrés et logistique. Si les problèmes de petite taille sont résolus efficacement avec de nombreux algorithmes d’optimisation, les problèmes de grande échelle nécessitent en revanche des méthodes du premier ordre issues de la descente de gradient. Dans ce manuscrit, nous considérons le cas particulier de la perte quadratique. Dans une première partie, nous nous proposons de la minimiser grâce à un oracle stochastique. Dans une seconde partie, nous considérons deux de ses applications à l’apprentissage automatique : au partitionnement de données et à l’estimation sous contrainte de forme. La première contribution est un cadre unifié pour l’optimisation de fonctions quadratiques non-fortement convexes. Celui-ci comprend la descente de gradient accélérée et la descente de gradient moyennée. Ce nouveau cadre suggère un algorithme alternatif qui combine les aspects positifs du moyennage et de l’accélération. La deuxième contribution est d’obtenir le taux optimal d’erreur de prédiction pour la régression par moindres carrés en fonction de la dépendance au bruit du problème et à l’oubli des conditions initiales. Notre nouvel algorithme est issu de la descente de gradient accélérée et moyennée. La troisième contribution traite de la minimisation de fonctions composites, somme de l’espérance de fonctions quadratiques et d’une régularisation convexe. Nous étendons les résultats existants pour les moindres carrés à toute régularisation et aux différentes géométries induites par une divergence de Bregman. Dans une quatrième contribution, nous considérons le problème du partitionnement discriminatif. Nous proposons sa première analyse théorique, une extension parcimonieuse, son extension au cas multi-labels et un nouvel algorithme ayant une meilleure complexité que les méthodes existantes. La dernière contribution de cette thèse considère le problème de la sériation. Nous adoptons une approche statistique où la matrice est observée avec du bruit et nous étudions les taux d’estimation minimax. Nous proposons aussi un estimateur computationellement efficace. / Many problems in machine learning are naturally cast as the minimization of a smooth function defined on a Euclidean space. For supervised learning, this includes least-squares regression and logistic regression. While small problems are efficiently solved by classical optimization algorithms, large-scale problems are typically solved with first-order techniques based on gradient descent. In this manuscript, we consider the particular case of the quadratic loss. In the first part, we are interestedin its minimization when its gradients are only accessible through a stochastic oracle. In the second part, we consider two applications of the quadratic loss in machine learning: clustering and estimation with shape constraints. In the first main contribution, we provided a unified framework for optimizing non-strongly convex quadratic functions, which encompasses accelerated gradient descent and averaged gradient descent. This new framework suggests an alternative algorithm that exhibits the positive behavior of both averaging and acceleration. The second main contribution aims at obtaining the optimal prediction error rates for least-squares regression, both in terms of dependence on the noise of the problem and of forgetting the initial conditions. Our new algorithm rests upon averaged accelerated gradient descent. The third main contribution deals with minimization of composite objective functions composed of the expectation of quadratic functions and a convex function. Weextend earlier results on least-squares regression to any regularizer and any geometry represented by a Bregman divergence. As a fourth contribution, we consider the the discriminative clustering framework. We propose its first theoretical analysis, a novel sparse extension, a natural extension for the multi-label scenario and an efficient iterative algorithm with better running-time complexity than existing methods. The fifth main contribution deals with the seriation problem. We propose a statistical approach to this problem where the matrix is observed with noise and study the corresponding minimax rate of estimation. We also suggest a computationally efficient estimator whose performance is studied both theoretically and experimentally.

The aim of the thesis is to define the dependency of the classification of patients affected by neurodegenerative diseases on the choice of the parcellation atlas. Part of this thesis is the application of the functional connectivity analysis and the calculation of graph metrics according to the method published by Olaf Sporns and Mikail Rubinov [1] on fMRI data measured at CEITEC MU. The application is preceded by the theoretical research of parcellation atlases for brain segmentation from fMRI frames and the research of mathematical methods for classification as well as classifiers of neurodegenerative diseases. The first chapters of the thesis brings a theoretical basis of knowledge from the field of magnetic and functional magnetic resonance imaging. The physical principles of the method, the conditions and the course of acquisition of image data are defined. The third chapter summarizes the graph metrics used in the diploma thesis for analyzing and classifying graphs. The paper presents a brief overview of the brain segmentation methods, with the focuse on the atlas-based segmentation. After a theoretical research of functional connectivity methods and mathematical classification methods, the findings were used for segmentation, calculation of graph metrics and for classification of fMRI images obtained from 96 subjects into the one of two classes using Binary classifications by support vector machines and linear discriminatory analysis. The data classified in this study was measured on patiens with Parkinson’s disease (PD), Alzheimer’s disease (AD), Mild cognitive impairment (MCI), a combination of PD and MCI and subjects belonging to the control group of healthy individuals. For pre-processing and analysis, the MATLAB environment, the SPM12 toolbox and The Brain Connectivity Toolbox were used.

This work is focused on objects and scenes recognition using machine learning and computer vision tools. Before the solution of this problem has been studied basic phases of the machine learning concept and statistical models with accent on their division into discriminative and generative method. Further, the Bag-of-words method and its modification have been investigated and described. In the practical part of this work, the implementation of the Bag-of-words method with the SVM classifier was created in the Matlab environment and the model was tested on various sets of publicly available images.

Visual tracking is a computer vision problem where the task is to follow a targetthrough a video sequence. Tracking has many important real-world applications in several fields such as autonomous vehicles and robot-vision. Since visual tracking does not assume any prior knowledge about the target, it faces different challenges such occlusion, appearance change, background clutter and scale change. In this thesis we try to improve the capabilities of tracking frameworks using discriminative correlation filters by incorporating scene depth information. We utilize scene depth information on three main levels. First, we use raw depth information to segment the target from its surroundings enabling occlusion detection and scale estimation. Second, we investigate different visual features calculated from depth data to decide which features are good at encoding geometric information available solely in depth data. Third, we investigate handling missing data in the depth maps using a modified version of the normalized convolution framework. Finally, we introduce a novel approach for parameter search using genetic algorithms to find the best hyperparameters for our tracking framework. Experiments show that depth data can be used to estimate scale changes and handle occlusions. In addition, visual features calculated from depth are more representative if they were combined with color features. It is also shown that utilizing normalized convolution improves the overall performance in some cases. Lastly, the usage of genetic algorithms for hyperparameter search leads to accuracy gains as well as some insights on the performance of different components within the framework.