In studies of urban quality of life, the information that can be extracted from satellite images is limited by image resolution and by the standard method of pixel classification. Recently, very high spatial resolution (VHSR) satellite images have allowed the development of new remote sensing application, especially for complex urban areas. Despite of the numerous advantages of the object-oriented approach for VHSR image processing, the parameters used to carry it out, especially at the object creation stage, are not very well documented. Moreover, the evaluation of urban quality of life has never considered the perception of inhabitants of the zones under study. This dissertation therefore addresses these two issues and aims 1) at testing a systematic ways of achieving the best parameters for object-oriented classification with the software Definiens and 2) at quantifying the relation between objective indicators and perceived satisfaction. Hoàn Kiém district, in Hanoi, Vietnam, was chosen as our zone of interest. The image used for this study is a 0,7m spatial resolution Quickbird image.In the first part of the dissertation, we identify eight land occupation classes on the image: lakes, river, parks, groups of trees along streets, isolated trees, large road and residential blocks. Using these classes and additional cartographic information, we calculate nine quality of life indicators that correspond to two central aspects of urban life: commodity (urban services) and amenity (urban landscape). For each group of indicators, we carried out a principal components analysis to obtain non-correlated components. We then conducted a survey with eight city planning experts who live and work in the zone under study to obtain an assessment of the satisfaction of inhabitants towards their area of residence. The weight of each component in the determination of quality of life was achieved through an ordinal regression whose independent variables are the components and the dependent variable is the level of satisfaction as evaluated by the experts. The weights were then used to interpret the importance of our indicators for quality of life. Our results show that it is possible to classify land occupation types with a good accuracy: our average accuracy rate is 80.5%. As for the weight of quality of life indicators, our results allow us to make methodological and interpretative contributions. Contrary to previous work, our method allows us to evaluate the explanatory power of our model. Our regression shows that 22% of variation in satisfaction towards commodity and nearly 54% of variation in satisfaction towards amenity can be attributed to our indicators. As for the nature of the factors playing a role in quality of life, our results show that the relation between indicators and perceived satisfaction is not linear, which had never been shown in previous studies. Satisfaction towards commodity increases when transportation and health care are both sufficient. Satisfaction towards amenity is on the other hand largely determined by residential space, while vegetation plays a minor role, contrary to what was found in the urban zones of developed countries.

Jusqu'à l'avènement de méthodes de neuroimagerie non invasives les connaissances du cerveau sont acquis par l'étude de ses lésions, des analyses post-mortem et expérimentations invasives. De nos jours, les techniques modernes d'imagerie telles que l'IRMf sont capables de révéler plusieurs aspects du cerveau humain à une résolution spatio-temporelle progressivement élevé. Cependant, afin de pouvoir répondre à des questions neuroscientifiques de plus en plus complexes, les améliorations techniques dans l'acquisition doivent être jumelés à de nouvelles méthodes d'analyse des données. Dans cette thèse, je propose différentes applications de l'apprentissage statistique au traitement des données d'IRMf. Souvent, les données acquises par le scanner IRMf suivent une étape de sélection de variables dans lequel les cartes d'activation sont extraites du signal IRMf. La première contribution de cette thèse est l'introduction d'un modèle nommé Rank-1 GLM (R1-GLM) pour l'estimation jointe des cartes d'activation et de la fonction de réponse hémodynamique (HRF). Nous quantifions l'amélioration de cette approche par rapport aux procédures existantes sur différents jeux de données IRMf. La deuxième partie de cette thèse est consacrée au problème de décodage en IRMf, ce est à dire, la tâche de prédire quelques informations sur les stimuli à partir des cartes d'activation du cerveau. D'un point de vue statistique, ce problème est difficile due à la haute dimensionnalité des données, souvent des milliers de variables, tandis que le nombre d'images disponibles pour la formation est faible, typiquement quelques centaines. Nous examinons le cas où la variable cible est composé à partir de valeurs discrets et ordonnées. La deuxième contribution de cette thèse est de proposer les deux mesures suivantes pour évaluer la performance d'un modèle de décodage: l'erreur absolue et de désaccord par paires. Nous présentons plusieurs modèles qui optimisent une approximation convexe de ces fonctions de perte et examinent leur performance sur des ensembles de données IRMf. Motivé par le succès de certains modèles de régression ordinales pour la tâche du décodage basé IRMf, nous nous tournons vers l'étude de certaines propriétés théoriques de ces méthodes. La propriété que nous étudions est connu comme la cohérence de Fisher. La troisième, et la plus théorique, la contribution de cette thèse est d'examiner les propriétés de cohérence d'une riche famille de fonctions de perte qui sont utilisés dans les modèles de régression ordinales. / Until the advent of non-invasive neuroimaging modalities the knowledge of the human brain came from the study of its lesions, post-mortem analyses and invasive experimentations. Nowadays, modern imaging techniques such as fMRI are revealing several aspects of the human brain with progressively high spatio-temporal resolution. However, in order to answer increasingly complex neuroscientific questions the technical improvements in acquisition must be matched with novel data analysis methods. In this thesis we examine different applications of machine learning to the processing of fMRI data. We propose novel extensions and investigate the theoretical properties of different models. % The goal of an fMRI experiments is to answer a neuroscientific question. However, it is usually not possible to perform hypothesis testing directly on the data output by the fMRI scanner. Instead, fMRI data enters a processing pipeline in which it suffers several transformations before conclusions are drawn. Often the data acquired through the fMRI scanner follows a feature extraction step in which time-independent activation coefficients are extracted from the fMRI signal. The first contribution of this thesis is the introduction a model named Rank-1 GLM (R1-GLM) for the joint estimation of time-independent activation coefficients and the hemodynamic response function (HRF). We quantify the improvement of this approach with respect to existing procedures on different fMRI datasets. The second part of this thesis is devoted to the problem of fMRI-based decoding, i.e., the task of predicting some information about the stimuli from brain activation maps. From a statistical standpoint, this problem is challenging due to the high dimensionality of the data, often thousands of variables, while the number of images available for training is small, typically a few hundreds. We examine the case in which the target variable consist of discretely ordered values. The second contribution of this thesis is to propose the following two metrics to assess the performance of a decoding model: the absolute error and pairwise disagreement. We describe several models that optimize a convex surrogate of these loss functions and examine their performance on different fMRI datasets. Motivated by the success of some ordinal regression models for the task of fMRI-based decoding, we turn to study some theoretical properties of these methods. The property that we investigate is known as consistency or Fisher consistency and relates the minimization of a loss to the minimization of its surrogate. The third, and most theoretical, contribution of this thesis is to examine the consistency properties of a rich family of surrogate loss functions that are used in the context of ordinal regression. We give sufficient conditions for the consistency of the surrogate loss functions considered. This allows us to give theoretical reasons for some empirically observed differences in performance between surrogates.

La littérature sur la diversification criminelle s’est particulièrement développée au cours des deux dernières décennies. Parmi les études, la diversité dans les activités illicites a été examinée auprès de divers sous-échantillons. Toutefois, aucune n’a été menée sur la versatilité criminelle de fournisseurs de drogues. Pourtant, ces derniers se distinguent quant à plusieurs aspects de leurs activités criminelles. L’objectif principal de ce mémoire est de déterminer quels sont les facteurs ayant une influence sur la diversification criminelle des distributeurs de drogues au Québec. Pour ce faire, un sous-échantillon de 33 270 fournisseurs de drogues a été circonscrit à l’aide de données d’arrestations policières des années 2003 à 2009. Le résultat émergeant d’une régression logistique ordinale est que les facteurs de diversification criminelle sont essentiellement les mêmes pour notre sous-échantillon et les autres délinquants. Nous constatons également que nous gagnons en rigueur en distinguant les fournisseurs spécialisés dans un marché illicite, des distributeurs dits polydrogues. En effet, les analyses indiquent que ces derniers s’impliquent dans un nombre restreint de types de crimes. Finalement, contrairement à ce qui a été observé dans la littérature sur la diversité criminelle, certaines variables n’ont que peu ou pas d’effet sur la versatilité dans les activités illicites, comme l’intensité criminelle ainsi que l’âge. / The literature on criminal diversification was developed particularly over the past two decades. Among the studies, diversity in illicit activities has been examined in various subsamples. However, no study has been conducted on the criminal versatility of drug providers. However, they differ in several aspects of their criminal activities. The main objective of this work is to determine the factors that influence the criminal diversification of drug distributors in Quebec. To do this, we use a sub-sample of 33,270 drug providers, created using police arrest data from the years 2003 to 2009. The main result of an ordinal logistic regression is that the factors of criminal diversification are essentially the same for our subsample and other offenders. We also note that we are gaining in rigour by distinguishing suppliers specialized in an illicit market from the so-called polydrugs distributors. Indeed, analyses indicate that these polydrugs distributors are involved in a limited number of crime types. Finally, contrary to what has been observed in the literature on criminal diversity, some variables have little or no effect on versatility in illicit activities, such as criminal intensity and age.