Inledning: Den specialiserade ätstörningsvård som bedrivs i Stockholm har olika behandlingskoncept gällande anorexia nervosa (AN) för ungdomar. Familjens roll varierar likväl. Syftet med denna studie är att jämföra två av de största ätstörningsklinikerna i Stockholm; AB Mando samt Stockholms Centrum för Ätstörningar (SCÄ) för att se om och i så fall hur de skiljer sig åt. Frågeställningar: Vilken grundsyn har vårdpersonal vid två olika kliniker på AN och på familjen roll under behandlingsarbetet och hur argumenterar man för detta? Hur går behandlingen till och vad är familjens roll? Hur förhåller man sig överhuvudtaget till familjen och varför? Metod: I studien, som är en kvalitativ undersökning, har vårdpersonal från dessa ätstörningskliniker deltagit i semistrukturerade intervjuer. Studien är bearbetat utifrån en tematisk analys. Resultat: De två klinikerna skiljer sig diametralt åt vilket grundar sig i två helt olika kunskapsparadigm. Med utgångspunkt i den teoretiska bakgrunden kan man se att det systemteoretiska synsättet råder på SCÄ där familjen roll är avgörande under behandlingen, genom att arbeta relationellt med hela familjen så kan patienten tillfriskna. På Mando där det inlärningsteoretiska/biologiska synsättet råder är konceptet det motsatta; genom att särskilja patienten från familjen och hemmiljön kan tillfrisknande ske då det ätstörda beteendet riskerar att vidmakthållas om patienten är kvar i den kontext där det blivit inlärt. Diskussion: Dessa olikheter torde ge skilda kliniska implikationer. Frågan om lika vård lyfts men också att det kan vara en fördel med ett varierat behandlingsutbud utifrån familjens behov. / Introduction: The specialized eating disorder treatments provided in Stockholm have different treatment concepts of anorexia nervosa (AN) for adolescents. The role of the family varies as well. The purpose of this study is to compare two of the biggest eating disorder clinics in Stockholm, AB Mando and the Stockholm Centrum för Ätstörningar (SCÄ) to see whether, and if so, they differ. Questions: What views have health professionals at two different clinics on AN and the family's role in the treatment process, and how do they argue for this? How does one relate to the family and why so? Method: In this study, which has a qualitative approach, health professionals at the two eating disorder clinics participated in semi-structured interviews. The interviews were processed in a thematic analysis. Results: The two clinics differs diametrically, which is based in two completely different paradigms. Based on the theoretical background, one can see that the system theoretical approach prevails at SCA, where the family's role is crucial during treatment, only by working relationally with the whole family the patient can recover. At Mando where learning theory / biological approach prevails, the concept is the opposite; by separating the patient from the family and the home environment recovery can be made. The behavior is likely to be maintained if the patient is left in the context where the eating disorders has developed. Discussion: These differences should result in different clinical implications. The issue of equal treatment is raised, but also that it can be an advantage with a varied range of treatments based on the varied needs of the families.

This study aims to identify tactics used by the Turkish Hezbollah to recruit college students into joining their terrorist organization. This study based on the assumptions that social networks and institutional structures are two main tools that are used effectively by the Turkish Hezbollah to recruit college educated students. In this sense, the researcher claims that Social Learning theory and Social Control Theories can be used to provide theoretical explanation to the Hezbollah’s recruitment strategy. Parallel to these theories assumptions, while having militants within social networks increases the likelihood of being recruited through social learning theory assumptions, college students who are away from their families are more likely to be recruited through social control theory assumptions. The researcher uses individual level secondary data related to members of the Turkish Hezbollah. The data comprised of self reports that each member submitted to the Turkish Hezbollah as part of their recruitment process. The data are derived from the Turkish National Police’s database. Initially, frequency table is used to determine which structure and which theory best explain the Turkish Hezbollah’s recruitment strategies. Then, to decide which demographic factors increase or decrease the likelihood of being recruited through social networks (social learning theory) or institutional structures (social control theory), logistic regression is used. Eight independent variables are used to identify those factors such as having Hezbollah militants within social networks, pursuing college education while being away from family, family’s religious ideology, having online or campus education, family size, income level, college student’s religiosity level, and reason for attending Hezbollah. The findings indicated that social networks and institutional structures are two important tools that are used by the Turkish Hezbollah. Social networks are more effectively used structures comparing to institutional structures. According to the results, there are two important variables have more weight on dependent variable comparing to other variables. While having militants within the social networks increases the likelihood of being recruited through social learning theory assumptions, being away from families during college education increases the likelihood of being recruited through social control theory assumptions.

Cette thèse, réalisée dans le cadre d’un projet de recherche pluridisciplinaire financé par le Ministère de l’Économie, de l’Industrie et de l’Emploi, étudie le fonctionnement de systèmes didactiques scolaires et universitaires dans un contexte d’apprentissage médiatisé par un serious game dédié au génie mécanique appelé Mecagenius®. Les travaux ont pour parti contribué à produire des outils de conception de ce serious game et à étudier les usages didactiques de professeurs et d’élèves dans des classes de lycée, d’IUT et d’université. Le processus de conception et d’évaluation s'est appuyé sur trois études emboitées relevant de programmes épistémologique, didactique et cognitif. Les résultats ont permis de rendre compte de l’usage didactique de Mecagenius® dans des classes ordinaires à travers la manière dont les professeurs l’utilisent et d'identifier les stratégies d’études individualisées mises en œuvre par des élèves contrastés selon leur position d'excellence dans la classe. La recherche met en évidence des usages contrastés d'intégration de l'artefact à la pratique des enseignants allant de modalités transmissives à des formes de genèse instrumentale spécifiques leur permettant, au fil des interactions avec les élèves, d'utiliser le serious game en plus ou moins grande adéquation avec les choix didactiques ayant présidé à sa conception. Du côté des élèves, les résultats pointent des stratégies différentielles pouvant contribuer pour les meilleurs au développement de compétences spécifiques de génie mécanique ; les plus faibles n'étant pas en mesure de dépasser les obstacles à l’apprentissage rencontrés dans le jeu. La conclusion envisage des pistes pour la formation des enseignants relative à l’usage des serious games en classe. / This thesis, as part of a multidisciplinary research project financed by the Ministry of Economy, Industry and Employment, examines the process of educational systems using a serious game dedicated to mechanical engineering called Mecagenius® in schools and universities. The study partly contributed to produce engineering tools of that serious game and explore how students and teachers used it in high school, IUT and university classes. The design and evaluation process relied on three related studies based on epistemological, cognitive and didactic programs. The findings report on the didactic use of Mecagenius® in regular classes and identify individualized learning strategies implemented by students according to their contrasting level of excellence in the classroom. The research highlighted differences in the way teachers integrated the artifact to their teaching methods ranging from transmissive to specific forms of instrumental genesis allowing them over the interactions with students, to use the serious game in ways that are nearly consistent with the educational choices that led to its design. Concerning students, the results showed different strategies that could help high skilled students in acquiring specific skills on Mechanical Engineering but could not help the weakest ones to go through the learning obstacles encountered in the game. The conclusion is about ways to train and support teachers with use of serious games in classroom.

This paper concerns a case study which attempts to underline the importance of culture learning through foreign language acquisition in the transformation of prejudices and the fomentation of cultural awareness. More specifically, it aims to present the way in which adult learners’ critical reflection on culture in combination to foreign language learning can lead to their acknowledgement of the equity between cultures and their proximity to “the other”, a fact that is considered to foster cultural respect and awareness. For doing so, a qualitative approach was used in order to examine the perceptions of a group of adults learning Spanish in an elementary level class in Greece. Data was collected from in-depth interviews where participants were asked to expose their opinion of the importance of culture and its connection to language, to consider the implication of culture learning in foreign language learning and to share their personal views of the Hispanic culture as it derives from their studying the Spanish language and reflecting on the Hispanic culture-Spanish language correlation. Research delivered positive outcomes concerning this specific issue and showed evidence that the transformative effect of learning about a culture through language acquisition is indeed possible. Suggestions for further research on the matter are made so as to investigate the replicability of such a transformative effect in other similar cases and to secure the generalization of results.

Doctor of Philosophy / Department of Family Studies & Human Services / Sonya L. Britt / Briana S. Nelson-Goff / The current three essay dissertation researched the financial behaviors of military service members before and after deployment using primary data collected at a Midwestern U.S. Army installation. The introduction (Chapter 1) reviewed the two financial surveys administered to Soldiers before (N = 701) and after (N = 670) they left for a yearlong deployment to a war zone. The first essay (Chapter 2) explored the financial behaviors, financial knowledge, and financial anxiety as they relate to rank and deployment. The results suggested that financial behaviors after deployment (Time 2) were significantly better than financial behaviors before deployment (Time 1). Rank had a positive effect on increased subjective financial knowledge where all ranks above privates (E1 to E2) had greater financial knowledge. Privates first class, specialists, and corporals (E3 to E4) had significantly lower financial knowledge than their direct supervisors, sergeants and staff sergeants (E5 to E6). Finally, Soldiers reported more financial anxiety before deployment (Time 1) than after deployment (Time 2). Using the framework of social learning theory, the second essay (Chapter 3) expanded the research of military financial behaviors before deployment to more fully understand stress and other factors that influence financial behavior outcomes. Results suggested that past behaviors and some personal factors played a significant role in Soldiers’ financial behaviors. Higher levels of subjective financial knowledge, more internal locus of control, and lower levels of financial anxiety all had a positive effect on financial behavior outcomes. The past behaviors variable had the most explanatory value in Soldiers’ financial behaviors before deployment. Soldiers with any amount of credit card debt had worse financial behaviors compared to Soldiers with no credit card debt, while Soldiers with greater amounts of emergency financial savings were more likely to have better financial behaviors than those who did not have any emergency financial savings. The final essay (Chapter 4) studied the factors that influenced financial behavior outcomes of both Soldiers and college students. This essay used primary data from a college student sample to compare to the before deployment (Time 1) survey data of Soldiers. Findings reported that past behaviors and some personal factors played a significant role in the financial behavior outcomes. Soldiers and college students with higher levels of subjective financial knowledge, more internal locus of control, and lower financial anxiety reported positive financial behaviors The most explanatory concept was that of past behaviors, which revealed that participants with no credit card debt had better financial behaviors compared to respondents who had any level of credit card debt. The conclusion (Chapter 5) highlights the findings of all three essays, which contribute both to the financial behavior literature. These papers also contribute to the research on the personal financial matters of service members. The research has direct implications for policy makers, military leaders, service providers, and financial planners and counselors.

This is a quantitative study of archival data that examines Merger and Acquisition (M&A) activity using currently established healthcare quality and financial performance metrics. The research seeks to explicate the relationship between M&A activity and M&A experience in the healthcare industry as it relates to initiatives aimed at improving the quality and decreasing the cost of healthcare. The Affordable Care Act (ACA) legislation appears to be contributing to a trend toward M&A consolidation; by illuminating how this trend potentially impacts healthcare quality and cost reduction initiatives, this study’s contribution is both useful and practical. The units of analysis are Medicare reporting hospitals, hospital systems, and related healthcare providers that have or have not experienced an M&A or multiple M&As. The study shows a statistically significant improvement in quality each year from 2006–2014, which is reflected in higher scores for the four quality metrics measured. M&A activity, as measured by acquisition status and acquirer experience, did not appear to influence these quality metrics, with the exception of the heart failure measure, which showed a statistically significant positive influence of acquirer experience across all specifications. M&A activity’s possible effects on hospital financial performance was assessed through operating-cost-to-charge and capital-cost-to-charge ratios (CCRs). The operating CCR appears to be positively influenced by both acquisition status and acquirer experience, while the capital CCR was positively influenced only by acquirer experience. A positive influence is reflected in a decreasing ratio. Results on quality improvement over time, both before and after the ACA, suggest that the ACA itself may not be the driver for quality improvement. Similarly, decreases in OCCR occurred consistently and statistically significantly over time, both pre- and post-ACA, while CCCR showed statistically significant decreases in 2006–2008, 2013, and 2014. These results appear to support the notion that the trend was ongoing before the ACA was enacted and gave these measures high-profile exposure. This is a quantitative study of archival data that examines Merger and Acquisition (M&A) activity using currently established healthcare quality and financial performance metrics. The research seeks to explicate the relationship between M&A activity and M&A experience in the healthcare industry as it relates to initiatives aimed at improving the quality and decreasing the cost of healthcare. The Affordable Care Act (ACA) legislation appears to be contributing to a trend toward M&A consolidation; by illuminating how this trend potentially impacts healthcare quality and cost reduction initiatives, this study’s contribution is both useful and practical. The units of analysis are Medicare reporting hospitals, hospital systems, and related healthcare providers that have or have not experienced an M&A or multiple M&As. The study shows a statistically significant improvement in quality each year from 2006–2014, which is reflected in higher scores for the four quality metrics measured. M&A activity, as measured by acquisition status and acquirer experience, did not appear to influence these quality metrics, with the exception of the heart failure measure, which showed a statistically significant positive influence of acquirer experience across all specifications. M&A activity’s possible effects on hospital financial performance was assessed through operating-cost-to-charge and capital-cost-to-charge ratios (CCRs). The operating CCR appears to be positively influenced by both acquisition status and acquirer experience, while the capital CCR was positively influenced only by acquirer experience. A positive influence is reflected in a decreasing ratio. Results on quality improvement over time, both before and after the ACA, suggest that the ACA itself may not be the driver for quality improvement. Similarly, decreases in OCCR occurred consistently and statistically significantly over time, both pre- and post-ACA, while CCCR showed statistically significant decreases in 2006–2008, 2013, and 2014. These results appear to support the notion that the trend was ongoing before the ACA was enacted and gave these measures high-profile exposure.

Syftet med denna studie var att undersöka hur stor erfarenhet svenska studenter har av, och hur stor rädsla de har för, nättroll. Ett annat syfte var att undersöka om erfarenheten har ett samband med deras eventuella rädsla för nättroll. Studien ämnar även undersöka vad svenska studenter har för definition av nättroll. Datainsamlingen utgjordes av en elektronisk enkät bestående av en självkonstruerad Likertskala (Troll-Likert) som distribuerades via webbplattformen survey & report. Skalan har aldrig tidigare använts i annat forskningssyfte. Deltagarantalet var 93 studenter. För att undersöka och analysera resultatet användes såväl kvantitativ som kvalitativ metod. Till den kvantitativa delen användes Spearmans rangkorrelationskoefficient och en enkel regressionsanalys. I fråga om den kvalitativa delen användes tematisk analys enligt Braun och Clarke (2006). Resultatet av studien visade att studenterna både hade en hög erfarenhet och rädsla för nättroll. Regressionsanalysen resulterade i ett värde på r2= 0,120. Spearmans resultat visade på en signifikant korrelation mellan rädsla och erfarenhet och regressionslinjen visade att studenternas rädsla kan till 12% förklaras av deras erfarenhet. Den tematiska analysen resulterade i ett tema som sammanfattar respondenternas definitioner av nättroll: personer som provocerar fram reaktioner för nöjes skull. / The purpose of this study was to investigate whether Swedish students have experience of, and / or fear of, online trolls. Also, if there was a connection between them two. The study also intended to investigate what definition of online trolls Swedish students had. The data collection was made up of an electronic questionnaire consisting of a self-designed Likert scale (Troll-Likert) distributed via the web survey & report. The questionnaire has never been used for any other research purposes. The number of participants was 93 students. To investigate the results both a quantitative and a qualitative method were used. Spearman's rank correlation coefficient and a regression analysis were used as in the quantitative method and a thematic analysis according to Braun and Clarke (2006) was used in the qualitative method. The study showed that the students both had a high experience and fear of online trolls. The regression analysis resulted in a value of r2 = 0.120. Spearman's results showed a significant correlation between fear and experience and the regression line showed that the students' fear can be explained to 12% by their experience. The thematic analysis resulted in one theme which summarizes the respondents' definitions of online trolls: people who provoke reactions for pleasure

Les systèmes de communication modernes sont caractérisés par leur besoin croissant en mécanismes d’auto-configuration. En effet, dans de nombreux cas pratiques, la présence de dispositifs de centralisation tel qu’une station de base n’est ni réaliste ni pratique. Ceci est le cas, par exemple, des situations militaires ou aussi celles d’urgence, ou lorsque le déploiement de plus en plus dense de points d’accès rend la planification humaine irréalisable. Par conséquent, des problèmes tel que la conception de règles de comportement pour les appareils (ou groupes d’appareils) sur la fa¸ con de choisir leurs propres paramètres de transmission, se présentent naturellement. En particulier, les algorithmes d’auto-configuration doivent être en mesure de répondre à la nécessité de détecter, d’éviter ou de réduire les interférences, maintenant ainsi une qualité suffisante de communications quand une centralisation est indisponible, et ceci avec un minimum d’échange d’informations et de coopération. En outre, ces algorithmes doivent être en mesure de faire face aux variations naturelles des conditions d’émission, en raison de l’atténuation, des effets de masque, de la mobilité et de la variation des comportements des autres dispositifs qui peuvent éventuellement créer des interférences supplémentaires. L’objectif de cette thèse est d’étudier le problème conjoint de sélection de canal et de contrôle de puissance dans le contexte de réseaux ad hoc clustérisés à canaux multiples, c’est à dire, des réseaux décentralisés dans lesquels les appareils radio sont disposés en groupes appelés clusters, et de proposer un algorithme d’auto-configuration décentralisé viable pour un tel réseau.Le réseau est étudié et analysé par l’intermédiaire de la théorie des jeux, et les équilibres relatifs sont identifiés. Le premier objectif consiste à utiliser ces équilibres afin de quantifier les performances des différents algorithmes qui proviennent de la théorie de l’apprentissage dans les jeux. Un algorithme basé sur le paradigme “trial and error” est alors sélectionné en tant que solution candidat. Une fonction d’utilité particulière est conçue afin que l’équilibre puisse coïncider avec les solutions d’ un problème d’optimisation, maximisant ainsi la qualité des communications, tout en minimisant les ressources nécessaires. Ces résultats sont présentés sous la forme la plus générale et, par conséquent, ils peuvent ˆ être aussi considérés comme un cadre théorique général pour la conception des jeux, ainsi que des algorithmes d’apprentissage avec lesquels les réseaux décentralisés peuvent fonctionner à des points optimaux globaux, et ceci à l’aide uniquement de leurs connaissances locales disponibles. La pertinence de la conception du jeu ainsi que de l’algorithme d’apprentissage est mis en évidence au moyen de scénarios spécifiques dans des réseaux ad hoc clustérisés et décentralisés. Les résultats numériques confirment la pertinence de l’utilisation des fonctions utilitaires appropriées ainsi que de l’apprentissage ”trial and error” dans l’amélioration de la performance des réseaux décentralisés. / Modern communication systems are characterized by an increasing need for self-configuring networks. In fact, in many practical cases, the presence of centralizing devices such as a base station is neither realistic nor practical. This is the case, for instance, in military or emergency situation, or when the increasingly dense deployment of access points makes a man-made planning unfeasible. As a consequence, problems like designing behavioral rules for devices (or groups of devices) on how to select their own transmit parameters naturally arise. In particular, self-configuring algorithms must be able to respond to the necessity of detecting, avoiding or reducing interference, thus maintaining a sufficient quality of the communications when no centralization is available, and with minimum information exchange and cooperation. Moreover, these algorithms must be able to cope with the variations of the transmission conditions due to fading, shadowing, mobility and to the change in other devices behavioral patterns eventually creating extra interference.The goal of this thesis is to study the joint problem of channel selection and power control in the context of multiple-channel clustered ad-hoc networks, i.e., decentralized networks in which radio devices are arranged into groups known as clusters, and to propose a viable decentralized self-configuring algorithm for such a network.The network is studied and analyzed through game theory, and the relative equilibria are identified. The first purpose is to use these equilibria in order to quantify the performance of different algorithms that originate from the theory of learning in games. An algorithm based on the trial and error paradigm is then selected as a candidate solution. A particular utility function is designed in order for the equilibria to coincide with the solutions of an optimization problem, thus maximizing the quality of the communications while minimizing the resources needed. These results are presented in the most general form and therefore, they can also be seen as a framework for designing both games and learning algorithms with which decentralized networks can operate atglobal optimal points using only their available local knowledge. The pertinence of the game design and the learning algorithm are highlighted using specific scenarios in decentralized clustered ad hoc networks. Numerical results confirm the relevance of using appropriate utility functions and trial and error learning for enhancing the performance of decentralized networks.

The curriculum for ICT in UK schools was discontinued in September 2012 and replaced by a 'rebranded' subject of Computing, divided into three sub domains: Computer Science; Information Technology; and digital literacy. The latter was positioned as basic technical skills. There were concerns in the education community that the new curriculum promoted programming and computer science topics to the detriment of digital literacy and applied uses of technology. Much of the Computing education literature perpetuates the hegemony of the logical and abstract, and implies computational thinking and rationality are synonymous with criticality. During the same period, a maker culture was growing rapidly in the UK, and discourses around these activities promoted an entirely different notion of digital literacy, aligned with the wide body of literacy literature that focuses on notions of empowerment and criticality rather than basic functional skills. A digital maker tool called the Raspberry Pi was released with the intention of supporting the development of computer science and digital making competence, and thus sat at the boundary of the academic and maker communities. This thesis argues that developing 'criticality' is a vital component of Computing education and explores how learning activities with the Raspberry Pi might support development of 'criticality'. In setting the scene for the investigation, I will first explore the notions underpinning discourse around both computational and critical thinking and digital literacy, suggesting that the frictions would be best overcome by abandoning abstract constructs of knowledge and assumptions that it is possible to separate theory and practice. I show how the term 'critical' is itself problematic in the literature and I look to Wenger's social theory of learning to avoid the individualistic limits of Papert's constructionism, a popular learning theory in Computing education. Wenger's constructs of knowledgeability and competence help tell a different story of what it means to be a learner of the practice of Computing, both in learning for academic purposes and with intentions towards becoming a practitioner. In concert with learning citizenship, these constructs offer a more ethical framing of 'criticality'. Informed by this theoretical position, I suggest an original, exploratory implementation of Q methodology to explore learning with technology in school settings. I qualitatively compare 'before' and 'after' Q studies that represent perspectives at the individual and collective level, with reference to observations of classroom learning. The methodology facilitates a nuanced and complex investigation and the findings of the project suggest that where pupils are already predisposed to the subject, working with the Raspberry Pi develops a broader knowledgeability, but where there is no such predisposition, a pedagogy of expressibility influences how participation in Raspberry Pi learning activities may impact knowledgeability.

De nombreux algorithmes en Apprentissage Automatique utilisent une notion de distance ou de similarité entre les exemples pour résoudre divers problèmes tels que la classification, le partitionnement ou l'adaptation de domaine. En fonction des tâches considérées ces métriques devraient avoir des propriétés différentes mais les choisir manuellement peut-être fastidieux et difficile. Une solution naturelle est alors d'adapter automatiquement ces métriques à la tâche considérée. Il s'agit alors d'un problème connu sous le nom d'Apprentissage de Métriques et où le but est principalement de trouver les meilleurs paramètres d'une métrique respectant des contraintes spécifiques. Les approches classiques dans ce domaine se focalisent habituellement sur l'apprentissage de distances de Mahalanobis ou de similarités bilinéaires et l'une des principales limitations est le fait que le contrôle du comportement de ces métriques est souvent limité. De plus, si des travaux théoriques existent pour justifier de la capacité de généralisation des modèles appris, la plupart des approches ne présentent pas de telles garanties. Dans cette thèse nous proposons de nouveaux algorithmes pour apprendre des métriques à comportement contrôlé et nous mettons l'accent sur les propriétés théoriques de ceux-ci. Nous proposons quatre contributions distinctes qui peuvent être séparées en deux parties: (i) contrôler la métrique apprise en utilisant une métrique de référence et (ii) contrôler la transformation induite par la métrique apprise. Notre première contribution est une approche locale d'apprentissage de métriques où le but est de régresser une distance proportionnelle à la perception humaine des couleurs. Notre approche est justifiée théoriquement par des garanties en généralisation sur les métriques apprises. Dans notre deuxième contribution nous nous sommes intéressés à l'analyse théorique de l'intérêt d'utiliser une métrique de référence dans un terme de régularisation biaisé pour aider lors du processus d'apprentissage. Nous proposons d'utiliser trois cadres théoriques différents qui nous permettent de dériver trois mesures différentes de l'apport de la métrique de référence. Ces mesures nous donnent un aperçu de l'impact de la métrique de référence sur celle apprise. Dans notre troisième contribution nous proposons un algorithme d'apprentissage de métriques où la transformation induite est contrôlée. L'idée est que, plutôt que d'utiliser des contraintes de similarité et de dissimilarité, chaque exemple est associé à un point virtuel qui appartient déjà à l'espace induit par la métrique apprise. D'un point de vue théorique nous montrons que les métriques apprises de cette façon généralisent bien mais aussi que notre approche est liée à une méthode plus classique d'apprentissage de métriques basée sur des contraintes de paires. Dans notre quatrième contribution nous essayons aussi de contrôler la transformation induite par une métrique apprise. Cependant, plutôt que considérer un contrôle individuel pour chaque exemple, nous proposons une approche plus globale en forçant la transformation à suivre une transformation géométrique associée à un problème de transport optimal. D'un point de vue théorique nous proposons une discussion sur le lien entre la transformation associée à la métrique apprise et la transformation associée au problème de transport optimal. D'un point de vue plus pratique nous montrons l'intérêt de notre approche pour l'adaptation de domaine mais aussi pour l'édition d'images / Many Machine Learning algorithms make use of a notion of distance or similarity between examples to solve various problems such as classification, clustering or domain adaptation. Depending on the tasks considered these metrics should have different properties but manually choosing an adapted comparison function can be tedious and difficult. A natural trend is then to automatically tailor such metrics to the task at hand. This is known as Metric Learning and the goal is mainly to find the best parameters of a metric under some specific constraints. Standard approaches in this field usually focus on learning Mahalanobis distances or Bilinear similarities and one of the main limitations is that the control over the behaviour of the learned metrics is often limited. Furthermore if some theoretical works exist to justify the generalization ability of the learned models, most of the approaches do not come with such guarantees. In this thesis we propose new algorithms to learn metrics with a controlled behaviour and we put a particular emphasis on the theoretical properties of these algorithms. We propose four distinct contributions which can be separated in two parts, namely (i) controlling the metric with respect to a reference metric and (ii) controlling the underlying transformation corresponding to the learned metric. Our first contribution is a local metric learning method where the goal is to regress a distance proportional to the human perception of colors. Our approach is backed up by theoretical guarantees on the generalization ability of the learned metrics. In our second contribution we are interested in theoretically studying the interest of using a reference metric in a biased regularization term to help during the learning process. We propose to use three different theoretical frameworks allowing us to derive three different measures of goodness for the reference metric. These measures give us some insights on the impact of the reference metric on the learned one. In our third contribution we propose a metric learning algorithm where the underlying transformation is controlled. The idea is that instead of using similarity and dissimilarity constraints we associate each learning example to a so-called virtual point belonging to the output space associated with the learned metric. We theoretically show that metrics learned in this way generalize well but also that our approach is linked to a classic metric learning method based on pairs constraints. In our fourth contribution we also try to control the underlying transformation of a learned metric. However instead of considering a point-wise control we consider a global one by forcing the transformation to follow the geometrical transformation associated to an optimal transport problem. From a theoretical standpoint we propose a discussion on the link between the transformation associated with the learned metric and the transformation associated with the optimal transport problem. On a more practical side we show the interest of our approach for domain adaptation but also for a task of seamless copy in images