Accurate methods for glottal feature extraction include the use of high-speed video imaging (HSVI). There have been previous attempts to extract these features with the acoustic recording. However, none of these methods compare their results with an objective method, such as HSVI. This thesis tests these acoustic methods against a large diverse population of 46 subjects. Two previously studied acoustic methods, as well as one introduced in this thesis, were compared against two video methods, area and displacement for open quotient (OQ) estimation. The area comparison proved to be somewhat ambiguous and challenging due to thresholding effects. The displacement comparison, which is based on glottal edge tracking, proved to be a more robust comparison method than the area. The first acoustic methods OQ estimate had a relatively small average error of 8.90% and the second method had a relatively large average error of -59.05% compared to the displacement OQ. The newly proposed method had a relatively small error of -13.75% when compared to the displacements OQ. There was some success even though there was relatively high error with the acoustic methods, however, they may be utilized to augment the features collected by HSVI for a more accurate glottal feature estimation.

This thesis consists of five scientific papers dealing with equations related to the optimal switching problem, mainly backward stochastic differential equations and variational inequalities. Besides the scientific papers, the thesis contains an introduction to the optimal switching problem and a brief outline of possible topics for future research. Paper I concerns systems of variational inequalities with operators of Kolmogorov type. We prove a comparison principle for sub- and supersolutions and prove the existence of a solution as the limit of solutions to iteratively defined interconnected obstacle problems. Furthermore, we use regularity results for a related obstacle problem to prove Hölder continuity of this solution. Paper II deals with systems of variational inequalities in which the operator is of non-local type. By using a maximum principle adapted to this non-local setting we prove a comparison principle for sub- and supersolutions. Existence of a solution is proved using this comparison principle and Perron's method. In Paper III we study backward stochastic differential equations in which the solutions are reflected to stay inside a time-dependent domain. The driving process is of Wiener-Poisson type, allowing for jumps. By a penalization technique we prove existence of a solution when the bounding domain has convex and non-increasing time slices. Uniqueness is proved by an argument based on Ito's formula. Paper IV and Paper V concern optimal switching problems under incomplete information. In Paper IV, we construct an entirely simulation based numerical scheme to calculate the value function of such problems. We prove the convergence of this scheme when the underlying processes fit into the framework of Kalman-Bucy filtering. Paper V contains a deterministic approach to incomplete information optimal switching problems. We study a simplistic setting and show that the problem can be reduced to a full information optimal switching problem. Furthermore, we prove that the value of information is positive and that the value function under incomplete information converges to that under full information when the noise in the observation vanishes.

In this paper we present an econometric framework aimed at obtaining projections of income growth in Europe at the regional level. We account for model uncertainty in terms of the choice of explanatory variables, as well as the nature of the spatial spillovers of output growth and human capital investment. Building on recent advances in Bayesian model averaging, we construct projected trajectories of income and human capital simultaneously, while integrating out the effects of other covariates. This approach allows us to assess the potential contribution of future educational attainment to economic growth and income convergence among European regions over the next decades. Our findings suggest that income convergence dynamics and human capital act as important drivers of income growth for the decades to come. In addition we find that the relative return of improving educational attainment levels in terms of economic growth appears to be higher in peripheral European regions. (authors' abstract) / Series: Department of Economics Working Paper Series

Ces dernières années, le contenu disponible sur le Web a augmenté de manière considérable dans ce qu’on appelle communément le Web social. Pour l’utilisateur moyen, il devient de plus en plus difficile de recevoir du contenu de qualité sans se voir rapidement submergé par le flot incessant de publications. Pour les fournisseurs de service, le passage à l’échelle reste problématique. L’objectif de cette thèse est d’aboutir à une meilleure expérience utilisateur à travers la mise en place de systèmes de filtrage et de recommandation. Le filtrage consiste à offrir la possibilité à un utilisateur de ne recevoir qu’un sous ensemble des publications des comptes auxquels il est abonné. Tandis que la recommandation permet la découverte d’information à travers la suggestion de comptes à suivre sur des sujets donnés. Nous avons élaboré MicroFilter un système de filtrage passant à l’échelle capable de gérer des flux issus du Web ainsi que RecLand, un système de recommandation qui tire parti de la topologie du réseau ainsi que du contenu afin de générer des recommandations pertinentes. / In the last years, the amount of available data on the social Web has exploded. For the average user, it became hard to find quality content without being overwhelmed with publications. For service providers, the scalability of such services became a challenging task. The aim of this thesis is to achieve a better user experience by offering the filtering and recommendation features. Filtering consists to provide for a given user, the ability of receiving only a subset of the publications from the direct network. Where recommendation allows content discovery by suggesting relevant content producers on given topics. We developed MicroFilter, a scalable filtering system able to handle Web-like data flows and RecLand, a recommender system that takes advantage of the network topology as well as the content in order to provide relevant recommendations.

Microbes fulfill important ecosystem functions by contributing as drivers of global nutrient cycles. Their distribution patterns are mainly controlled by environmental heterogeneities. So far, little is known about the mode of action of particular environmental drivers on the microbiota, particularly in low productivity habitats. The aim of this thesis was to investigate the relationships between local environmental drivers and the microbial responses at the level of communities, individuals and realized function, using three structurally different model habitats sharing the feature of overall low productivity. Using a hypothesis-based approach and extensive 16S rRNA amplicon mapping of bacterioplankton colonizing the polar Southern Ocean, I identified how the seasonal formation of open-water polynyas and coupled phytoplankton production affected the diversity of surface bacterial communities and resulted in a cascading effect influencing the underlying dark polar water masses. Additional laboratory experiments, with cultures exposed to light, resulted in reduction in alpha diversity and promoted opportunistic populations with most bacterial populations thriving in the cultures typically reflected the dominants in situ. Furthermore it was experimentally tested how induced cyclic water table fluctuations shaping environmental heterogeneity in a constructed wetland on temporal scale, by directly affecting redox conditions. Twelve months of water table fluctuations resulted in enhanced microbial biomass, however a shift in community composition did not lead to a significant increase in pollutant removal efficiency when compared to a static control wetland. I detected phyla that have previously been proposed as key players in anaerobic benzene break-down using a protocol that was developed for single cell activity screening using isotope-substrate uptake and microautoradiography combined with taxonomic identification based on fluorescent in situ hybridization targeting the 16S rRNA. Eventually, I provide an example of how anthropogenic pollution with polyaromatic hydrocarbons induced a strong environmental filtering on intrinsic microbial communities in lake sediments. In conclusion, my studies reveal that microorganisms residing in low productivity habitats are greatly influenced by environmental heterogeneity across both spatial and temporal scales. However, such variation in community composition or overall abundance does not always translate to altered community function.

Exome sequencing, the process of sequencing the set of all known exons simultaneously using next-generation sequencing technology, has dramatically changed the landscape of genetic research and genetic testing. The incredible volume of data produced by these experiments creates challenges in: 1) annotating the affects of observed variants, 2) filtering to remove noise, 3) identifying plausible disease-causing variants, and 4) validating experimental results. Here we will present a series of bioinformatic tools and techniques intended to address these challenges with exome sequencing and associated validation experiments. First, we will present the Automated Sequence Analysis Pipeline (ASAP), a tool for the efficient and automated management, detection and annotation of Sanger sequencing-based genetic testing and variant validation. This pipeline is extended to annotate exome-sequencing derived variants. Exome sequencing experiments produce a great number of variants that do not cause a patient's disease. One of the biggest challenges in exome sequencing experiments is sorting through these false positives to discover the true disease-causing variants. We have developed several techniques to aid in the reduction of these errors. The techniques described include: 1) the construction of a catalog of systematic errors by reprocessing thousands of publically available exomes, 2) a tool for the filtering of variants based on family structure and disease assumptions, and 3) a tool for discovering regions of autozygosity from the exomes of several affected patients in consanguineous pedigrees. Classes of variants that are undiscoverable using current analysis techniques gives rise to false negatives in exome sequencing experiments. We will present a tool, the Retrotransposon Insertion Detector for Exomes (RIDE) that uses the characteristic anomalies present in sequence alignments to detect the insertion of repetitive elements. The process of identifying a the cause of a patient's disease using exome sequencing data has been equated to finding a needle in a stack of needles. Only through the proper annotation of variants and the reduction of the error rates associated with exome sequencing experiments can this task be achieved in an efficient manner.

Filtering is a key step in digital image processing and analysis. It is mainly used for amplification or attenuation of some frequencies depending on the nature of the application. Filtering can either be performed in the spatial domain or in a transformed domain. The selection of the filtering method, filtering domain, and the filter parameters are often driven by the properties of the underlying image. This thesis presents three different kinds of biomedical image filtering applications, where the filter parameters are automatically determined from the underlying images. Filtering can be used for image enhancement. We present a robust image dependent filtering method for intensity inhomogeneity correction of biomedical images. In the presented filtering method, the filter parameters are automatically determined from the grey-weighted distance transform of the magnitude spectrum. An evaluation shows that the filter provides an accurate estimate of intensity inhomogeneity. Filtering can also be used for analysis. The thesis presents a filtering method for heart localization and robust signal detection from video recordings of rat embryos. It presents a strategy to decouple motion artifacts produced by the non-rigid embryonic boundary from the heart. The method also filters out noise and the trend term with the help of empirical mode decomposition. Again, all the filter parameters are determined automatically based on the underlying signal. Transforming the geometry of one image to fit that of another one, so called image registration, can be seen as a filtering operation of the image geometry. To assess the progression of eye disorder, registration between temporal images is often required to determine the movement and development of the blood vessels in the eye. We present a robust method for retinal image registration. The method is based on particle swarm optimization, where the swarm searches for optimal registration parameters based on the direction of its cognitive and social components. An evaluation of the proposed method shows that the method is less susceptible to becoming trapped in local minima than previous methods. With these thesis contributions, we have augmented the filter toolbox for image analysis with methods that adjust to the data at hand.

Collaborative filtering based recommender systems use information about a user's preferences to make personalized predictions about content, such as topics, people, or products, that they might find relevant. As the volume of accessible information and active users on the Internet continues to grow, it becomes increasingly difficult to compute recommendations quickly and accurately over a large dataset. In this study, we will introduce an algorithmic framework built on top of Apache Spark for parallel computation of the neighborhood-based collaborative filtering problem, which allows the algorithm to scale linearly with a growing number of users. We also investigate several different variants of this technique including user and item-based recommendation approaches, correlation and vector-based similarity calculations, and selective down-sampling of user interactions. Finally, we provide an experimental comparison of these techniques on the MovieLens dataset consisting of 10 million movie ratings.

Low-rank matrix factorizations have become increasingly popular to project high dimensional data into latent spaces with small dimensions in order to obtain better understandings of the data and thus more accurate predictions. In particular, they have been widely applied to important applications such as collaborative filtering and social network analysis. In this thesis, I investigate the applications and extensions of the ideas of the low-rank matrix factorization to solve several practically important problems arise from collaborative filtering and social network analysis. A key challenge in recommendation system research is how to effectively profile new users, a problem generally known as \emph{cold-start} recommendation. In the first part of this work, we extend the low-rank matrix factorization by allowing the latent factors to have more complex structures --- decision trees to solve the problem of cold-start recommendations. In particular, we present \emph{functional matrix factorization} (fMF), a novel cold-start recommendation method that solves the problem of adaptive interview construction based on low-rank matrix factorizations. The second part of this work considers the efficiency problem of making recommendations in the context of large user and item spaces. Specifically, we address the problem through learning binary codes for collaborative filtering, which can be viewed as restricting the latent factors in low-rank matrix factorizations to be binary vectors that represent the binary codes for both users and items. In the third part of this work, we investigate the applications of low-rank matrix factorizations in the context of social network analysis. Specifically, we propose a convex optimization approach to discover the hidden network of social influence with low-rank and sparse structure by modeling the recurrent events at different individuals as multi-dimensional Hawkes processes, emphasizing the mutual-excitation nature of the dynamics of event occurrences. The proposed framework combines the estimation of mutually exciting process and the low-rank matrix factorization in a principled manner. In the fourth part of this work, we estimate the triggering kernels for the Hawkes process. In particular, we focus on estimating the triggering kernels from an infinite dimensional functional space with the Euler Lagrange equation, which can be viewed as applying the idea of low-rank factorizations in the functional space.

The work described in this thesis contributes to the development of a system to evaluate sailing performance. This work was motivated by the lack of tools available to evaluate sailing performance. The goal of the work presented is to detect and classify the turns of a sailing yacht. Data was collected using a BlackBerry PlayBook affixed to a J/24 sailing yacht. This data was manually annotated with three types of turn: tack, gybe, and mark rounding. This manually annotated data was used to train classification methods. Classification methods tested were multi-layer perceptrons (MLPs) of two sizes in various committees and nearest- neighbour search. Pre-processing algorithms tested were Kalman filtering, categorization using quantiles, and residual normalization. The best solution was found to be an averaged answer committee of small MLPs, with Kalman filtering and residual normalization performed on the input as pre-processing.