Tephra deposits are one of the possible outcomes of explosive volcanic eruptions and are the result of vertical settling of volcanic particles that have been expelled from the volcanic vent into the atmosphere, following magma fragmentation within the volcanic conduit. Tephra fallout represents the main volcanic hazard to populated areas and critical facilities. Therefore, it is crucial to better understand processes that lead to tephra transport, sedimentation and hazards. In this study, and based on detailed mapping and sampling of the tephra deposit of the 2450 BP Plinian eruption of Pululagua volcano (Ecuador), I investigate tephra deposits through a variety of approaches, including empirical and analytical modeling of tephra thickness and grain size data to infer important eruption source parameters (e.g. column height, total mass ejected, total grain size distribution of the deposit). I also use a statistical approach (smoothed bootstrap with replacement method) to assess the uncertainty in the eruptive parameters. The 2450 BP Pululagua volcanic plume dynamics were also explored through detailed grain size analysis and 1D modeling of tephra accumulation. Finally, I investigate the influence of particle shape on tephra accumulation on the ground through a quantitative and comprehensive study of the shape of volcanic ash. As the global need for energy is expected to grow in the future, many future natural hazard studies will likely involve the assessment of volcanic hazards at critical facilities, including nuclear power plants. I address the potential hazards from tephra fallout, pyroclastic flows and lahars for the Bataan Nuclear Power Plant (Philippines) posed by three nearby volcanoes capable of impacting the site during an explosive eruption. I stress the need for good constraints (stratigraphic analysis and events dating) on past eruptive events to better quantify the probability of future events at potentially active volcanoes, the need for probabilistic approaches in such volcanic hazard assessments to address a broad range of potential eruption scenarios, and the importance of considering coupled volcanic processes (e.g. tephra fallout leading to lahars) in volcanic hazard assessments.

This thesis studies grammar-based approaches in the application of Estimation of Distribution Algorithms (EDA) to the tree representation widely used in Genetic Programming (GP). Although EDA is becoming one of the most active fields in Evolutionary computation (EC), the solution representation in most EDA is a Genetic Algorithms (GA) style linear representation. The more complex tree representations, resembling GP, have received only limited exploration. This is unfortunate, because tree representations provide a natural and expressive way of representing solutions for many problems. This thesis aims to help fill this gap, exploring grammar-based approaches to extending EDA to GP-style tree representations. This thesis firstly provides a comprehensive survey of current research on EDA with emphasis on EDA with GP-style tree representation. The thesis attempts to clarify the relationship between EDA with conventional linear representations and those with a GP-style tree representation, and to reveal the unique difficulties which face this research. Secondly, the thesis identifies desirable properties of probabilistic models for EDA with GP-style tree representation, and derives the PRODIGY framework as a consequence. Thirdly, following the PRODIGY framework, three methods are proposed. The first method is Program Evolution with Explicit Learning (PEEL). Its incremental general-to-specific grammar learning method balances the effectiveness and efficiency of the grammar learning. The second method is Grammar Model-based Program Evolution (GMPE). GMPE realises the PRODIGY framework by introducing elegant inference methods from the formal grammar field. GMPE provides good performance on some problems, but also provides a means to better understand some aspects of conventional GP, especially the building block hypothesis. The third method is Swift GMPE (sGMPE), which is an extension of GMPE, aiming at reducing the computational cost. Fourthly, a more accurate Minimum Message Length metric for grammar learning in PRODIGY is derived in this thesis. This metric leads to improved performance in the GMPE system, but may also be useful in grammar learning in general. It is also relevant to the learning of other probabilistic graphical models.

Graphical techniques for modeling the dependencies of randomvariables have been explored in a variety of different areas includingstatistics, statistical physics, artificial intelligence, speech recognition, image processing, and genetics.Formalisms for manipulating these models have been developedrelatively independently in these research communities. In this paper weexplore hidden Markov models (HMMs) and related structures within the general framework of probabilistic independencenetworks (PINs). The paper contains a self-contained review of the basic principles of PINs.It is shown that the well-known forward-backward (F-B) and Viterbialgorithms for HMMs are special cases of more general inference algorithms forarbitrary PINs. Furthermore, the existence of inference and estimationalgorithms for more general graphical models provides a set of analysistools for HMM practitioners who wish to explore a richer class of HMMstructures.Examples of relatively complex models to handle sensorfusion and coarticulationin speech recognitionare introduced and treated within the graphical model framework toillustrate the advantages of the general approach.

Nous proposons une approche probabiliste afin de déterminer l’impact des changements dans les programmes à objets. Cette approche sert à prédire, pour un changement donné dans une classe du système, l’ensemble des autres classes potentiellement affectées par ce changement. Cette prédiction est donnée sous la forme d’une probabilité qui dépend d’une part, des interactions entre les classes exprimées en termes de nombre d’invocations et d’autre part, des relations extraites à partir du code source. Ces relations sont extraites automatiquement par rétro-ingénierie. Pour la mise en oeuvre de notre approche, nous proposons une approche basée sur les réseaux bayésiens. Après une phase d’apprentissage, ces réseaux prédisent l’ensemble des classes affectées par un changement. L’approche probabiliste proposée est évaluée avec deux scénarios distincts mettant en oeuvre plusieurs types de changements effectués sur différents systèmes. Pour les systèmes qui possèdent des données historiques, l’apprentissage a été réalisé à partir des anciennes versions. Pour les systèmes dont on ne possède pas assez de données relatives aux changements de ses versions antécédentes, l’apprentissage a été réalisé à l’aide des données extraites d’autres systèmes. / We study the possibility of predicting the impact of changes in object-oriented code using bayesian networks. For each change type, we produce a bayesian network that determines the probability that a class is impacted given that another class is changed. Each network takes as input a set of possible relationships between classes. We train our networks using historical data. The proposed impact-prediction approach is evaluated with two different scenarios, various types of changes, and five systems. In the first scenario, we use as training data, the changes performed in the previous versions of the same system. In the second scenario training data is borrowed from systems that are different from the changed one. Our evaluation showed that, in both cases, we obtain very good predictions, even though they are better in the first scenario.

O desafio de encontrar corretamente genes eucarioticos codificadores de proteinas nas sequencias genomicas e um problema em aberto. Neste trabalho, implementamos uma plata- forma, com o objetivo de melhorar a forma com que preditores de genes sao implementados e avaliados. Tres novas ferramentas foram implementadas: ToPS (Toolkit of Probabilistic Models of Sequences) foi o primeiro arcabouco orientado a objetos que fornece ferramentas para implementacao, manipulacao, e combinacao de modelos probabilisticos para representar sequencias de simbolos; MYOP (Make Your Own Predictor) e um sistema que tem como objetivo facilitar a construcao de preditores de genes; e SGEval utiliza grafos de splicing para comparar diferente anotacoes com eventos de splicing alternativos. Utilizamos nossas ferramentas para o desenvolvimentos de preditores de genes em onze genomas distintos: A. thaliana, C. elegans, Z. mays, P. falciparum, D. melanogaster, D. rerio, M. musculus, R. norvegicus, O. sativa, G. max e H. sapiens. Com esse desenvolvimento, estabelecemos um protocolo para implementacao de novos preditores. Alem disso, utilizando a nossa plata- forma, desenvolvemos um fluxo de trabalho para predicao de genes no projeto do genoma da cana de acucar, que ja foi utilizado em 109 sequencias de BAC geradas pelo BIOEN (FAPESP Bioenergy Program). / The challenge of correctly identify eukaryotic protein-coding genes in the genomic se- quences is an open problem. In this work, we implemented a plataform with the aim of improving the way that gene predictors are implemented and evaluated. ToPS (Toolkit of Probabilistic Models of Sequence) was the first object-oriented framework that provides tools for implementation, manipulation, and combination of probabilistic models that represent sequences of symbols. MYOP (Make Your Own Predictor) facilitates the construction of gene predictors. SGEval (Splicing Graph Evaluation) uses splicing graphs to compare dif- ferent annotations with alternative splicing events. We used our plataform to develop gene finders in eleven distinct genomes: A. thaliana, C. elegans, Z. mays, P. falciparum, D. me- lanogaster, D. rerio, M. musculus, R. norvegicus, O. sativa, G. max e H. sapiens. With this development, we established a protocol for implementing new gene predictors. In addi- tion, using our platform, we developed a pipeline to find genes in the 109 sugarcane BAC sequences produced by BIOEN (FAPESP Bioenergy Program).

Le travail que nous présentons dans ce mémoire apporte sa contribution au domaine dela surveillance et de la supervision en ligne des systèmes à événements discrets complexes.Il se place dans un contexte perturbé par l’occurrence d’aléas de fonctionnement d’une partieopérative au sein duquel nous visons à mettre à disposition des équipes de maintenance desoutils pour les aider à localiser rapidement les équipements à l’origine probable de défautsproduits : localiser mieux pour maintenir mieux et donc minimiser encore davantage les tempsde dérives équipements. Si les équipements de production étaient en mesure de détecterde telles dérives, le problème pourrait être considéré comme simple, cependant, la présenced’équipements de métrologie montre le contraire. Aussi, partant du constat que les équipementsde production ne peuvent être dotés d’un système de captage couvrant de manière exhaustivel’ensemble des paramètres à observer, que la fiabilité des capteurs est variable dans le temps,que les contextes de production sont particulièrement stressants, nous avons proposé ici dedévelopper une approche probabiliste basée sur un raisonnement Bayésien permettant d’estimeren temps réel la confiance qui peut être accordée aux opérations réalisées par les équipementsde production. / The work that we present in this paper contributes to the field of supervision, monitoringand control of complex discrete event systems services. It is placed in the context of randomfailure occurrence of operative parts where we focus on providing tools to maintenance teamsby locating the possible origin of potential defect products: better locate to better maintain, soeffectively to minimize more equipment’s time drift. If the production equipment were able todetect such drifts, the problem could be considered simple; however, metrology equipment addsto the complexity. In addition, because of an impossibility to equip the production equipment witha sensor system which comprehensively covers all parameters to be observed, a variable sensorreliability in time and a stressed production environments, we propose a probabilistic approachbased on Bayesian network to estimate real time confidence, which can be used for productionequipment?s operation.

Made available in DSpace on 2014-06-12T15:50:15Z (GMT). No. of bitstreams: 2 arquivo1959_1.pdf: 2352925 bytes, checksum: 90760b286db4ed0dcc12ae48554413a9 (MD5) license.txt: 1748 bytes, checksum: 8a4605be74aa9ea9d79846c1fba20a33 (MD5) Previous issue date: 2008 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / Estudos recentes têm demonstrado que as redes biológicas apresentam características nãoaleatórias, dentre as quais destacamos a arquitetura modular. Neste trabalho, estamos interessados na organização modular das redes de regulação transcricional (RRT), que modelizam as interações entre genes e proteínas que controlam a sua expressão no nível transcricional. Compreender os mecanismos de regulação transcricional é crucial para se explicar a diversidade morfológica e funcional das células. Nós nos propomos a abordar o problema da identificação de módulos regulatórios transcricionais, i.e. grupos de genes co-regulados e seus reguladores, com ênfase no aspecto computacional. Uma distinção importante deste trabalho é que estamos também interessados em estudar o aspecto evolutivo dos módulos transcricionais. Do ponto de vista biológico, a abordagem proposta está fundamentada em três premissas principais: (i) genes co-regulados são controlados por proteínas regulatórias comuns (fatores de transcrição FTs) e, portanto, eles devem apresentar padrões de sequência (motifs) comuns nas suas regiões regulatórias, que correspondem aos sítios de ligação desses FTs, (ii) genes co-regulados respondem coordenadamente a certas condições ambientais e de desenvolvimento e, logo, devem ser co-expressos sob essas condições, e (iii) uma vez que módulos transcricionais são presumivelmente responsáveis por funções biológicas importantes, eles estão sujeitos a uma maior pressão seletiva e, consequentemente, devem ser evolutivamente conservados. Nós definimos, portanto, o conceito de metamódulo regulatório transcricional (MMRT) como grupos de genes compartilhando motifs e exibindo um comportamento de expressão coerente em contextos específicos consistentemente em várias espécies e propomos modelos probabilísticos para descrever o comportamento modular em termos do compartilhamento de elementos regulatórios (motifs), da co-expressão e da conservação evolutiva das associações funcionais entre os genes com base em dados diversos tais como dados de sequência, de expressão e dados filogenéticos

La compréhension et la prédiction du comportement vibro-acoustique des systèmes légers bois du bâtiment constitue un enjeu scientifique d'actualité. En 2015 une étude montrait encore que presque la moitié de ces systèmes constructifs n'offrait pas satisfaction. Un modèle prédictif à l'échelle du bâtiment, en cours de normalisation, permet de prendre en compte la performance individuelle des différents systèmes séparatifs pour remonter à un niveau de performance globale. La difficulté scientifique réside alors dans l'évaluation de la performance individuelle associée à chaque conception admissible, dans un vaste ensemble de systèmes techniquement réalisables. Dans cette recherche, une méthodologie est proposée pour la construction de modèles numériques capables de prendre en compte, aux basses fréquences, la complexité et la diversité des systèmes bois constitués de multiples plaques, poutres, cavités acoustiques et matériaux poroélastiques. En accord avec les procédures d'évaluation normalisées, des modèles déterministes pour les excitations mécaniques du système sont construits. Une approche probabiliste est alors développée en réponse à la problématique des incertitudes liées à la construction légère. Ainsi, en résolvant un problème stochastique inverse utilisant des données expérimentales pour identifier les hyperparamètres de modèles probabilistes développés, il est possible de quantifier la propagation des incertitudes du système à la performance prédite en conditions de laboratoire. Par suite, des configurations optimales, robustes aux incertitudes, sont recherchées. Du fait de la nature combinatoire du problème d'optimisation, un algorithme génétique, particulièrement adapté à un espace de recherche discret ainsi qu'à l'optimisation multi-objectif, est mis en oeuvre. Dans les cas traités, les configurations optimales tendent vers une maximisation de la rigidité structurelle / Being able to understand and predict the vibroacoustic behavior of lightweight wood-based building systems contitute a serious scientific concern. In 2015, acoustic comfort investigation claims that unsatisfactions are expressed with respect to around 50% of such constructions. In particular, low frequency discomfort is target of criticism. A methodology was proposed, currently running through standardisation process, which translates the individual performance of the building systems into a global building performance index. The challenge consequently lies in the prediction of the individual performances in regard to the wide spread of wood based designs. In this research, a methodology is introduced for the construction of computational models able to handle the complexity and diversity of the systems, constituted of multiple boards, stiffeners, cavities and poroelastic media. Structural excitations of the system are constructed according to standard evaluation procedures. Then, a probabilistic approach is undertaken in order to take into account the uncertainty problematic, inherent to lightweight wood based constructions. In particular, stochastic inverse problems are constructed to identify, from experimental measurements, hyperparameters associated with ad hoc probabilistic models. Eventually, uncertainty quantification can be performed in regard to predicted performance in laboratory conditions. Following, robust optimal designs are sought in the presence of uncertainties. No continuous mapping from the search space of the configurations to the space of the fitness functions representative of the objective performance exists and derivatives cannot be defined. By way of consequence, the class of the evolutionnary algorithm, suited to discrete search spaces as well as multi-objective optimisation, is chosen. Considered optimisation problems displayed preferential directions of the genetic algorithm towards stiffest admissible designs

La classification et le regroupement des données fonctionnelles longitudinales ont fait beaucoup de progrès dans les dernières années. Plusieurs méthodes ont été proposées et ont démontré des résultats prometteurs. Pour ce mémoire, on a comparé le comportement des algorithmes de partitionnement sur un ensemble de données décrivant les trajectoires de voitures dans une intersection de Montréal. La motivation est qu’il est coûteux et long de faire la classification manuellement et on démontre dans cet ouvrage qu’il est possible d’obtenir des prédictions adéquates avec les différents algorithmes. Parmi les méthodes utilisées, la méthode distclust utilise l’approche des K-moyennes avec une notion de distance entre les courbes fonctionnelles. On utilise aussi une classification par mélange de densité gaussienne, mclust. Ces deux approches n’étant pas conçues uniquement pour le problème de classification fonctionnelle, on a donc également appliqué des méthodes fonctionnelles spécifiques au problème : fitfclust, funmbclust, funclust et funHDDC. On démontre que les résultats du partitionnement et de la prédiction obtenus par ces approches sont comparables à ceux obtenus par ceux basés sur la distance. Les méthodes fonctionnelles sont préférables, car elles permettent d’utiliser des critères de sélection objectifs comme le AIC et le BIC. On peut donc éviter d’utiliser une partition préétablie pour valider la qualité des algorithmes, et ainsi laisser les données parler d’elles-mêmes. Finalement, on obtient des estimations détaillées de la structure fonctionnelle des courbes, comme sur l’impact de la réduction de données avec une analyse en composantes principales fonctionnelles multivariées. / The study of the clustering of functional data has made a lot of progress in the last couple of years. Multiple methods have been proposed and the respective analysis has shown their eÿciency with some benchmark studies. The objective of this Master’s thesis is to compare those clustering algorithms with datasets from traÿc at an intersection of Montreal. The idea behind this is that the manual classification of these data sets is time-consuming. We show that it is possible to obtain adequate clustering and prediction results with several algorithms. One of the methods that we discussed is distclust : a distance-based algorithm that uses a K-means approach. We will also use a Gaussian mixture density clustering method known as mclust. Although those two techniques are quite e˙ective, they are multi-purpose clustering methods, therefore not tailored to the functional case. With that in mind, we apply four functional clustering methods : fitfclust, funmbclust, funclust, and funHDDC. Our results show that there is no loss in the quality of the clustering between the afore-mentioned functional methods and the multi-purpose ones. We prefer to use the functional ones because they provide a detailed estimation of the functional structure of the trajectory curves. One notable detail is the impact of a dimension reduction done with multivari-ate functional principal components analysis. Furthermore, we can use objective selection criteria such as the AIC and the BIC, and avoid using cluster quality indices that use a pre-existing classification of the data.

In modern games and animation there is a constant strive for more realistic motion. Today a lot of games use motion matching and blending with lots of post-processing steps to produce animations, but these methods often require huge amounts of motions clips while still having problems with realistic joint weights. Using machine learning for generating motion is a fairly new technique, and is proving to be a viable option due to the lower cost and potentially more realistic results. Probabilistic models could be suitable candidates for solving a problem such as this as the are able to model a wide variety of motions due to their built-in randomness. This thesis examines a few different models which could be used for generating motion for character when interacting with terrain, such as when walking up an incline. The main models examined in this thesis are the MoGlow model and a CVAE model. Firstly virtual scenes are built in Unity based upon loads of motion capture clips containing movements interacting with the terrain. A character is then inserted into the scene and the animation clips are played. Data is exported consisting of the character’s joint positions and rotations in relation to the surrounding terrain. This data is then used to train the models using supervised learning. Evaluation of this is done by having character go through an obstacles course of varying terrains, generating motion from the different models. After this foot sliding was measured as well as frame-rates. This was also compared to values from that of a selection of motion capture clips. In addition to this a user study is conducted where the users are asked to rate the quality of generated motion in certain video clips. The results show that both the MoGlow and CVAE models produced movement resembling real human movement on uneven terrain, with the MoGlow model’s results being most similar to that of a the motion capture training data. These were also found to be executable at interactive frame-rates, making them suitable for use in video games. / I moderna spel och animationer finns det en konstant strävan efter mer realistisk rörelse. I dagsläget använder många spel teknologier så som rörelsematchning och flera efterprocessering steg för att producera animationer, men ett problem med dessa metoder är att det oftast krävs enorma mängder rörelse klipp för att kunna anpassas till alla möjliga situationer, samtidigt som man ofta tappar lite av vikten i rörelserna. Användet av maskinginlärning för att generera rörelser är en relativt ny utveckling, och ses som en möjlig lösning till dessa problem. Probabilistka modeller är en typ av modeller som kan användas för detta, eftersom att de kan representera en bred variation av rörelser med samma model, på grund av den underligande slumpmässigheten. Det här pappret kommer att undersöka olika probabilistka modeller som kan användas för att generera rörelse när man även tar hansyn till omgivningen, tex när man går i en uppförsbacke. De huvudsakliga modellerna som kommer undersökas är en MoGlow model och en CVAE model. Först så byggs virtuella scener in Unity utifrån en mängd animationsklipp. Därefter stoppas en karaktär in och de här klippen spelas upp. I detta steg är data exporterad som innehåller karaktärens position och benens rotationer i relation till omgivningen. Denna data används sedan för att träna modellerna med väglett lärande. Evaluering är genomförd genom att ha karaktärer gå igenom hinderbanor uppbyggda av varierande terränger, där modeller genererar rörelser för karaktären. Fotglidande och bildhastighet är avmätt och resultatet av metoderna är jämfört med varandra och med utvald data från inspelade träningsdatan. Utöver detta görs även en användarstudie där personer får ge betyg till generarde rörelser utifrån en mängd videoklipp. Resultaten visar att båda MoGlow och CVAE modellen producerar rörelse som liknar realsiska männsklig rörelse vid interaktion mod ojämn terräng. MoGlow modellen visar resultat mest likt den inspelade data. Alla modeller testade går att kör interaktiva bildhastigheter, vilket gör dem lämpliga för använding i dataspel.