The aim of this thesis is to improve our understanding of the neural systems involved in schizophrenia by suggesting possible avenues for future computational modelling in an attempt to make sense of the vast number of studies relating to the symptoms and cognitive deficits relating to the disorder. This multidisciplinary research has covered three different levels of analysis: abnormalities in the microscopic brain structure, dopamine dysfunction at a neurochemical level, and interactions between cortical and subcortical brain areas, connected by cortico-basal ganglia circuit loops; and has culminated in the production of five models that provide useful clarification in this difficult field. My thesis comprises three major relevant modelling themes. Firstly, in Chapter 3 I looked at an existing neural network model addressing the Neurodevelopmental Hypothesis of Schizophrenia by Hoffman and McGlashan (1997). However, it soon became clear that such models were overly simplistic and brittle when it came to replication. While they focused on hallucinations and connectivity in the frontal lobes they ignored other symptoms and the evidence of reductions in volume of the temporal lobes in schizophrenia. No mention was made of the considerable evidence of dysfunction of the dopamine system and associated areas, such as the basal ganglia. This led to my second line of reasoning: dopamine dysfunction. Initially I helped create a novel model of dopamine neuron firing based on the Computational Substrate for Incentive Salience by McClure, Daw and Montague (2003), incorporating temporal difference (TD) reward prediction errors (Chapter 5). I adapted this model in Chapter 6 to address the ongoing debate as to whether or not dopamine encodes uncertainty in the delay period between presentation of a conditioned stimulus and receipt of a reward, as demonstrated by sustained activation seen in single dopamine neuron recordings (Fiorillo, Tobler & Schultz 2003). An answer to this question could result in a better understanding of the nature of dopamine signaling, with implications for the psychopathology of cognitive disorders, like schizophrenia, for which dopamine is commonly regarded as having a primary role. Computational modelling enabled me to suggest that while sustained activation is common in single trials, there is the possibility that it increases with increasing probability, in which case dopamine may not be encoding uncertainty in this manner. Importantly, these predictions can be tested and verified by experimental data. My third modelling theme arose as a result of the limitations to using TD alone to account for a reinforcement learning account of action control in the brain. In Chapter 8 I introduce a dual weighted artificial neural network, originally designed by Hinton and Plaut (1987) to address the problem of catastrophic forgetting in multilayer artificial neural networks. I suggest an alternative use for a model with fast and slow weights to address the problem of arbitration between two systems of control. This novel approach is capable of combining the benefits of model free and model based learning in one simple model, without need for a homunculus and may have important implications in addressing how both goal directed and stimulus response learning may coexist. Modelling cortical-subcortical loops offers the potential of incorporating both the symptoms and cognitive deficits associated with schizophrenia by taking into account the interactions between midbrain/striatum and cortical areas.

The research presented in this thesis is aimed at developing reliable information systems through the application of model-driven and formal techniques. These are techniques in which a precise, formal model of system behaviour is exploited as source code. As such a model may be more abstract, and more concise, than source code written in a conventional programming language, it should be easier and more economical to create, to analyse, and to change. The quality of the model of the system can be ensured through certain kinds of formal analysis and fixed accordingly if necessary. Most valuably, the model serves as the basis for the automated generation or configuration of a working system. This thesis provides four research contributions. The first involves the analysis of a proposed modelling language targeted at the model-driven development of information systems. Logical properties of the language are derived, as are properties of its compiled form---a guarded substitution notation. The second involves the extension of this language, and its semantics, to permit the description of workflows on information systems. Workflows described in this way may be analysed to determine, in advance of execution, the extent to which their concurrent execution may introduce the possibility of deadlock or blocking: a condition that, in this context, is synonymous with a failure to achieve the specified outcome. The third contribution concerns the validation of models written in this language by adapting existing techniques of software testing to the analysis of design models. A methodology is presented for checking model consistency, on the basis of a generated test suite, against the intended requirements. The fourth and final contribution is the presentation of an implementation strategy for the language, targeted at standard, relational databases, and an argument for its correctness, based on a simple, set-theoretic semantics for structure and operations.

Moderne Theorien der Verhaltenskontrolle unterscheiden zwei Systeme, wobei das Handeln gesunder Individuen von beiden geprägt ist: Das retrospektiv agierende habituelle, sog. modellfreie Verhalten zeichnet sich durch Wiederholung zuvor belohnter Entscheidungen aus. Es passt sich nur langsam an möglicherweise veränderte Umweltbedingungen an. Die verstärkte Nutzung des habituellen Systems gilt als assoziiert mit verschiedenen psychischen Erkrankungen. Dem gegenüber steht das zielgerichtete, sog. modellbasierte Verhalten, das sich durch vorausschauende Entscheidungen auszeichnet. Hierbei werden die möglichen Konsequenzen einer Handlung berücksichtigt, um ein gewünschtes Ergebnis zu erreichen. Dazu wird ein „mentales“ Modell der Umwelt- bedingungen erstellt. In einer Verhaltensstudie mit 39 Versuchspersonen wurde untersucht, ob biopsychologischer Stress zu einer Reduktion von modellbasiertem hin zu mehr modellfreiem Verhalten führt. Dazu absolvierten 39 Versuchspersonen eine sequentielle Entscheidungsaufgabe, nachdem sie psychosozialem Stress ausgesetzt wurden. Subjektive und physiologische Stress-Parameter wurden über das Experiment hinweg wiederholt erhoben. Ein direkter Effekt von akutem Stress auf das Gleichgewicht modellfreien vs. modellbasierten Verhaltens konnte nicht beobachtet werden. Allerdings zeigten diejenigen Versuchspersonen, die in den letzten zwei Jahren eine hohe Anzahl an stressbehafteten Lebensereignissen aufwiesen (chronischer Stress), signifikant weniger modellbasiertes Verhalten nach der Exposition von akutem Stress als in der Kontrollbedingung. Darüber hinaus korrelierte physiologische Stressreaktivität (stressbedingte Cortisol- Ausschüttung) negativ mit modellbasiertem Entscheidungsverhalten, während subjektive Stressreaktivität (basierend auf Fragebögen) positiv mit modellbasiertem Verhalten assoziiert war. Der in der Forschung beschriebene Einfluss von akutem und chronischem Stress auf die Entstehung und Aufrechterhaltung psychischer Erkrankungen könnte demnach teilweise über ein solches Ungleichgewicht der beiden Entscheidungsstrategien vermittelt sein.

Ein geringer Treibstoffverbrauch ist aufgrund von ökologischen und ökonomischen Zielen für die zivile Luftfahrt von großer Bedeutung. Daher werden seit Jahrzehnten konventionell hydraulisch oder pneumatisch betriebene Flugzeugsysteme durch elektrisch betriebene Systeme ersetzt. Dieser Trend wird auch als „More Electric Aircraft (MEA)“ bezeichnet. In bisherigen Studien waren MEA-Architekturen zwar effizienter, jedoch deutlich schwerer als die konventionellen Architekturen. Basierend auf ökonomischen Modellen wird in der vorliegenden Arbeit die modellbasierte Entwicklung eines intelligenten Energiemanagements für Flugzeug-Energiesysteme demonstriert. Das Energiemanagement ermöglicht eine deutliche Reduktion der Systemmasse, verbessert die Energieeffizienz und kann damit den Treibstoffverbrauch eines MEA beträchtlich reduzieren. Insbesondere durch die integrierte und frühzeitige Entwicklung des Energiemanagements mit dem elektrischen System in der Modellbeschreibungssprache Modelica lassen sich die Systemkomponenten mit realistischen Lastprofilen dimensionieren und dadurch die Systemmasse reduzieren. Anhand eines elektrischen Referenzsystems wird das Optimierungspotenzial des Energiemanagements bezüglich Massenreduktion und Energieeffizienzsteigerung quantifiziert und am Systemmodell validiert. Es ergibt sich für das Systemmodell eine Reduktion der Systemmasse um 32 % sowie eine leichte Verbesserung der Energieeffizienz. Durch die multiphysikalische Implementierung des Energiemanagements lässt sich dieses auch für das thermische Management im Flugzeug verwenden. Hierbei kann eine deutliche Verbesserung der Energieeffizienz für die Bereitstellung von Kühlleistung erzielt werden. Aufgrund der erreichten Vorteile sollte ein Energiemanagement bei der Entwicklung zukünftiger Flugzeugenergiesysteme in Betracht gezogen werden. Insbesondere beim MEA existiert ein großes Optimierungspotenzial durch das Energiemanagement. Die Ausführungen in der vorliegenden Arbeit sollen als Motivation für die Flugzeugindustrie dienen, mit realistischen Lastprofilen zu dimensionieren und die modellbasierte und integrierte Entwicklung eines Energiemanagements mit den Energiesystemen bereits in frühen Entwicklungsphasen durchzuführen. / Low fuel consumption is a major concern in civil aerospace due to environmental and economic objectives. Hence, conventional hydraulically or pneumatically driven aircraft systems have been replaced by electrically driven systems for decades. This trend is also known as More Electric Aircraft (MEA). In former studies, MEA architectures were more efficient, but much heavier than their conventional counterparts. The present work demonstrates the model-based development of intelligent energy management algorithms for aircraft energy systems based on economic models. This energy management facilitates a significant reduction of system mass, improves energy efficiency and can hence reduce fuel consumption of MEA considerably. In particular, the integrated development of an energy management along with the electrical system in the Modelica modelling language enables sizing of system components with realistic load profiles. Hence, this reduces the system mass. The optimization potential of the energy management is quantified and validated by means of an electrical reference system model. Applying the energy management, the mass of this system model can be reduced by 32 % and the energy efficiency can be improved slightly. Due to the multi-physical modelling of the energy management, it can also be applied to thermal management of aircraft systems. Thus, the energy efficiency of the cooling system can be improved significantly. As a result of the demonstrated benefits, an energy management should be considered for future development of aircraft energy systems. Especially for MEA, there is tremendous optimization potential for the energy management. Hence, the present work shall motivate aircraft industry to size aircraft systems with realistic load profiles and perform a model-based and integrated development of the energy management along with the electrical system in early phases of the system design process.

Political campaigning has become a multi-million dollar business. A substantial proportion of a campaign's budget is spent on voter mobilization, i.e., on identifying and influencing as many people as possible to vote. Based on data, campaigns use statistical tools to provide a basis for deciding who to target. While the data available is usually rich, campaigns have traditionally relied on a rather limited selection of information, often including only previous voting behavior and one or two demographical variables. Statistical procedures that are currently in use include logistic regression or standard classification tree methods like CHAID, but there is a growing interest in employing modern data mining approaches. Along the lines of this development, we propose a modern framework for voter targeting called LORET (for logistic regression trees) that employs trees (with possibly just a single root node) containing logistic regressions (with possibly just an intercept) in every leaf. Thus, they contain logistic regression and classification trees as special cases and allow for a synthesis of both techniques under one umbrella. We explore various flavors of LORET models that (a) compare the effect of using the full set of available variables against using only limited information and (b) investigate their varying effects either as regressors in the logistic model components or as partitioning variables in the tree components. To assess model performance and illustrate targeting, we apply LORET to a data set of 19,634 eligible voters from the 2004 US presidential election. We find that augmenting the standard set of variables (such as age and voting history) together with additional predictor variables (such as the household composition in terms of party affiliation and each individual's rank in the household) clearly improves predictive accuracy. We also find that LORET models based on tree induction outbeat the unpartitioned competitors. Additionally, LORET models using both partitioning variables and regressors in the resulting nodes can improve the efficiency of allocating campaign resources while still providing intelligible models. / Series: Research Report Series / Department of Statistics and Mathematics

The WikiLeaks Afghanistan war logs contain nearly 77,000 reports of incidents in the US-led Afghanistan war, covering the period from January 2004 to December 2009. The recent growth of data on complex social systems and the potential to derive stories from them has shifted the focus of journalistic and scientific attention increasingly toward data-driven journalism and computational social science. In this paper we advocate the usage of modern statistical methods for problems of data journalism and beyond, which may help journalistic and scientific work and lead to additional insight. Using the WikiLeaks Afghanistan war logs for illustration, we present an approach that builds intelligible statistical models for interpretable segments in the data, in this case to explore the fatality rates associated with different circumstances in the Afghanistan war. Our approach combines preprocessing by Latent Dirichlet Allocation (LDA) with model trees. LDA is used to process the natural language information contained in each report summary by estimating latent topics and assigning each report to one of them. Together with other variables these topic assignments serve as splitting variables for finding segments in the data to which local statistical models for the reported number of fatalities are fitted. Segmentation and fitting is carried out with recursive partitioning of negative binomial distributions. We identify segments with different fatality rates that correspond to a small number of topics and other variables as well as their interactions. Furthermore, we carve out the similarities between segments and connect them to stories that have been covered in the media. This gives an unprecedented description of the war in Afghanistan and serves as an example of how data journalism, computational social science and other areas with interest in database data can benefit from modern statistical techniques. (authors' abstract)

Les systèmes avioniques s'étoffent et se complexifient de plus en plus. Avec l'augmentation des capacités de calcul, de nouvelles architectures basées sur le partage de ressources émergent. Effectuer le diagnostic d'un système n'est désormais plus une opération anodine. L'enjeu actuel est donc de mettre en place des techniques de diagnostic performantes tout en optimisant les capacités de monitoring nécessaires.Ce mémoire donne une caractérisation basée modèle d'un système sous diagnostic, puis propose des techniques pour en évaluer les performances de diagnostic, ainsi que celles de son monitoring (relativement à ces performances). Le contexte industriel dans lequel s'inscrit cette thèse amène d'autres contraintes, notamment la prise en compte de la taille des systèmes avioniques à analyser. Cette thèse étudie alors l'applicabilité des techniques introduites dans ce contexte et en propose une adaptation. / Avionics systems become more and more complex. With the improvment of computing possibilities, new architectures based on resources sharing are growing up. Perform diagnosis of a system is no longer a trivial operation. The challenge is to develop efficient techniques of diagnosis while optimizing capabilities of monitoring required.This thesis give a model-based characterization of a system under diagnosis, and proposes techniques to assess diagnostic performances, as well as its monitoring ones (with respect to these diagnostic performances). The industrial context of this thesis brings other constraints, and in particular the need to handle the size of avionics systems to analyze. That thesis then examines the applicability of the introduced techniques to this particular context, and proposes an adaptation.

In the presented thesis we study three methods of pricing European currency barrier options. With help of these methods we value selected barrier options with underlying asset EUR/CZK. In the first chapter we introduce the basic definitions from the world of financial derivatives and we describe our data. In the second chapter we deal with the classical model based on geometric Brownian motion of underlying asset and we prove a theorem of valuating Up-In-barrier option in this model. In the third chapter we introduce a model with stochastic volatility, the Heston model. We calibrate this model to market data and we use it to value our barrier options. In the last chapter we describe a jump diffusion model. Again we calibrate this jump diffusion model to market data and price our barrier options. The aim of this thesis is to decribe and to compare different methods of valuating barrier options. 1

The purpose of this thesis is to evaluate Rapid Control Prototyping which is apart of the Model-Based Design concept that makes it possible to convenientlytest prototype control algorithms directly on the real system. The evaluation ishere done by designing two different controllers, a gain-scheduled P controllerand a linear Model Predictive Controller (mpc), for the lowering of the forks of aforklift.The two controllers are first tested in a simulation environment. The thesis con-tains two different simulation models: one physical where only minor parameteradjustments are done and one estimated black-box model. After evaluating thecontrollers in a simulation environment they are tested on a real forklift with areal-time target machine.The designed controllers have different strengths and weaknesses as one is non-linear and single variable, the P controller, and the other linear and multivariable,thempc. The P controller has a smooth movement in all situations without be-ing slow, unlike thempc. The disadvantage of the P controller compared to thempcis that there is no guarantee that the P controller will keep the speed limit,whereas thempcapproach gives such a guarantee.The better performance of the P controller outweighs the speed limit guaranteeand thus a conclusion is drawn that the nonlinearities of the system has a largereffect than the multivariable aspect. Also, another conclusion drawn is that work-ing with Model-Based Design and Rapid Control Prototyping makes it possibleto test many different ideas on a real forklift without spending a lot of time onimplementation. / Syftet med detta examensarbete är att utvärdera Rapid Control Prototyping vil-ket är en del av modellbaserad utveckling som gör det möjligt att enkelt testamodeller av styralgoritmer direkt på det riktiga systemet. Utvärderingen är gjordgenom att testa två olika regulatorer, en P-regulator med parameterstyrning ochen linjär modelbaserad prediktionsregulator (mpc), för sänkningen av gafflarnapå en truck.De två regulatorerna testas först i en simuleringsmiljö. I arbetet används två olikasimuleringsmodeller: en fysikalisk där endast mindre parameterjusteringar görsoch en estimerad black-box modell. Efter att regulatorerna utvärderas i simule-ringsmiljön testas de även på en riktig truck med hjälp av automatisk kodgenere-ring och exekvering på en dedikerad hårdvaruplattform.De konstruerade regulatorerna har olika för- och nackdelar eftersom en är olinjäroch envariabel, P-regulatorn, och en är linjär men flervariabel,mpc:n. P-regulatornhar en mjuk rörelse i alla lägen utan att bli för långsam, till skillnad frånmpc:n.Nackdelen med P-regulatorn, jämfört medmpc:n är att det inte finns någon ga-ranti för att P-regulatorn håller hastighetsbegränsningen sommpc:n gör.P-regulatorns bättre prestanda överväger garantin om att hålla hastighetsbegräns-ningen och därför dras slutsatsen att olinjäriteterna i systemet överväger effekter-na av det faktum att det också är flervariabelt. En annan slutsats är att modell-baserad utveckling och Rapid Control Prototyping gör det möjligt att testa fleraolika idéer på en riktig gaffeltruck utan att spendera för mycket tid på implemen-tationen.

Les algorithmes d'inférence ou d'optimisation possèdent généralement des hyperparamètres qu'il est nécessaire d'ajuster. Nous nous intéressons ici à l'automatisation de cette étape d'ajustement et considérons différentes méthodes qui y parviennent en apprenant en ligne la structure du problème considéré.La première moitié de cette thèse explore l'ajustement des hyperparamètres en apprentissage artificiel. Après avoir présenté et amélioré le cadre générique de l'optimisation séquentielle à base de modèles (SMBO), nous montrons que SMBO s'applique avec succès à l'ajustement des hyperparamètres de réseaux de neurones profonds. Nous proposons ensuite un algorithme collaboratif d'ajustement qui mime la mémoire qu'ont les humains d'expériences passées avec le même algorithme sur d'autres données.La seconde moitié de cette thèse porte sur les algorithmes MCMC adaptatifs, des algorithmes d'échantillonnage qui explorent des distributions de probabilité souvent complexes en ajustant leurs paramètres internes en ligne. Pour motiver leur étude, nous décrivons d'abord l'observatoire Pierre Auger, une expérience de physique des particules dédiée à l'étude des rayons cosmiques. Nous proposons une première partie du modèle génératif d'Auger et introduisons une procédure d'inférence des paramètres individuels de chaque événement d'Auger qui ne requiert que ce premier modèle. Ensuite, nous remarquons que ce modèle est sujet à un problème connu sous le nom de label switching. Après avoir présenté les solutions existantes, nous proposons AMOR, le premier algorithme MCMC adaptatif doté d'un réétiquetage en ligne qui résout le label switching. Nous présentons une étude empirique et des résultats théoriques de consistance d'AMOR, qui mettent en lumière des liens entre le réétiquetage et la quantification vectorielle / Inference and optimization algorithms usually have hyperparameters that require to be tuned in order to achieve efficiency. We consider here different approaches to efficiently automatize the hyperparameter tuning step by learning online the structure of the addressed problem. The first half of this thesis is devoted to hyperparameter tuning in machine learning. After presenting and improving the generic sequential model-based optimization (SMBO) framework, we show that SMBO successfully applies to the task of tuning the numerous hyperparameters of deep belief networks. We then propose an algorithm that performs tuning across datasets, mimicking the memory that humans have of past experiments with the same algorithm on different datasets. The second half of this thesis deals with adaptive Markov chain Monte Carlo (MCMC) algorithms, sampling-based algorithms that explore complex probability distributions while self-tuning their internal parameters on the fly. We start by describing the Pierre Auger observatory, a large-scale particle physics experiment dedicated to the observation of atmospheric showers triggered by cosmic rays. The models involved in the analysis of Auger data motivated our study of adaptive MCMC. We derive the first part of the Auger generative model and introduce a procedure to perform inference on shower parameters that requires only this bottom part. Our model inherently suffers from label switching, a common difficulty in MCMC inference, which makes marginal inference useless because of redundant modes of the target distribution. After reviewing existing solutions to label switching, we propose AMOR, the first adaptive MCMC algorithm with online relabeling. We give both an empirical and theoretical study of AMOR, unveiling interesting links between relabeling algorithms and vector quantization.