It is estimated that less than ten percent of the world's species have been discovered and described. The main reason for the slow pace of new species description is that the science of taxonomy, as traditionally practiced, can be very laborious: taxonomists have to manually gather and analyze data from large numbers of specimens and identify the smallest subset of external body characters that uniquely diagnoses the new species as distinct from all its known relatives. The pace of data gathering and analysis can be greatly increased by the information technology. In this paper, we propose a content-based image retrieval system for taxonomic research. The system can identify representative body shape characters of known species based on digitized landmarks and provide statistical clues for assisting taxonomists to identify new species or subspecies. The experiments on a taxonomic problem involving species of suckers in the genera Carpiodes demonstrate promising results.

<div>Modern Java application frameworks present significant challenges for existing static analysis algorithms. Such challenges include large-scale code bases, heap-carried dependency, and asynchronous control flow caused by message passing.</div><div>Existing analysis algorithms are not suitable to deal with these challenges. One reason is that analyses are typically designed to operate homogeneously on the whole program. This leads to scalability problems when the analysis algorithms are used on applications built as plug-ins of large frameworks, since the framework code is analyzed together with the application code. Moreover, the asynchronous message passing of the actor model adopted by most modern frameworks leads to control flows which are not modeled by existing analyses.</div><div>This thesis presents several techniques for more powerful debugging and program understanding tools based on slicing. In general, slicing-based techniques aim to discover interesting properties of a large program by only reasoning about the relevant part of the program (typically a small amount of code) precisely, abstracting away the behavior of the rest of the program.</div><div>The key contribution of this thesis is a demand-driven framework to enable precise and scalable analyses on programs built on large frameworks. A slicing algorithm, which can handle heap-carried dependence, is used to identify the program elements relevant to an analysis query. We instantiated the framework to infer correlations between registration call sites and callback methods, and resolve asynchronous control flows caused by asynchronous message passing.</div>

Osteoarthritis (OA) of the knee is a disease that deteriorates the bones and surrounding soft tissue of the affected joint. Categorisation of the disease into grades of severity is subject to errors of measurement and poor observer agreement. There is an urgent need for automated methods to measure radiographic features and remove, as far as possible, the element of subjectivity in assessment. This project creates a fully automated system to analyse all aspects of the knee in radiographs. The methods evaluate explicit and implicit features of: overall shape, trabecular structure, osteophytes, tibial spines and intercondylar notch, and joint space shape. The project develops the first fully automated osteophyte detection algorithms, improved trabeculae features using raw pixel intensities, and a better analysis of joint space using shape models. This project is the first to combine explicit and implicit features across the whole of the knee, and applies these features to classify radiographs using four main outcomes: current OA, current pain, later onset OA, and later onset pain. The results find a strong current OA classification rate, with an Area Under the ROC Curve (AUC) of 0.904 and weighted kappa of 0.49 (0.48-0.51). The remaining later onset and pain experiments report weaker results; these results suggest that radiographic features in Posterior-Anterior (PA) view radiographs have a weak association with clinical and later onset OA.

Le cadre de cette thèse est l'analyse statique modulaire par interprétation abstraite de logiciels en vue de leur vérification automatique. Nous nous intéressons en particulier aux programmes comportant des objets alloués dynamiquement sur un tas et repérés par des pointeurs. Le but final étant de trouver des erreurs dans un programme (problèmes de déréférencements et d'alias) ou de prouver qu'un programme est correct (relativement à ces problèmes) de façon automatique. Isthiaq, Pym, O'Hearn et Reynolds ont développé récemment des logiques de fragmentation (separation logics) qui sont des logiques de Hoare avec un langage d'assertions/de prédicats permettant de démontrer qu'un programme manipulant des pointeurs sur un tas est correct. La sémantique des triplets de la logique ({P}C{P

<p>The identification of processes governing genetic structure at different spatial scales remains a major challenge in evolutionary biology and is of considerable applied interest in conservation biology. In <i>Euphrasia stricta</i> five varieties have been identified (<i>brevipila</i>, <i>gotlandica</i>, <i>stricta</i>, <i>suecica</i> and <i>tenuis</i>) based on differences in habitat, phenology and morphology. In this thesis, I examined genetic variation at AFLP and microsatellite marker loci in relation to variation in habitat and morphology within and among varieties of <i>E. stricta</i> on the island Gotland in the Baltic Sea. The results are discussed in relation to evolutionary processes acting within this species complex. </p><p>In a study conducted at the regional scale, the two early-flowering varieties <i>suecica</i> and <i>tenuis</i> each formed a genetically distinct group, while the three late-flowering varieties <i>brevipila</i>, <i>gotlandica</i> and <i>stricta</i> formed a third group. The results suggest that <i>suecica</i> and <i>tenuis</i> have ancient origins since they are genetically different both from the <i>brevipila</i>/<i>gotlandica</i>/<i>stricta</i> group and from each other despite their similar habitat preferences. This pattern was obtained using both marker systems. Discrepancies between AFLP and microsatellites were found in patterns of isolation by distance and in estimates of expected heterozygosity, <i>H</i>_e.</p><p>Focusing on the mixed genetic group <i>brevipila</i>/<i>gotlandica</i>/<i>stricta</i> and the causes behind their clustering together despite differences in morphology and habitat preferences, I performed a study at a smaller geographic scale. Studying a population of <i>E. stricta</i> I found that, although gene flow within the population was strong, it had not prevented the formation of genetic groups associated with micro-habitat properties. </p><p>An important result for conservation of the rare variety <i>suecica</i> is its distinct genetic separation from variety <i>tenuis</i>. If the aim of conservation is to preserve the uniqueness of <i>suecica,</i> the two varieties should be treated as separated entities. </p>

The identification of processes governing genetic structure at different spatial scales remains a major challenge in evolutionary biology and is of considerable applied interest in conservation biology. In Euphrasia stricta five varieties have been identified (brevipila, gotlandica, stricta, suecica and tenuis) based on differences in habitat, phenology and morphology. In this thesis, I examined genetic variation at AFLP and microsatellite marker loci in relation to variation in habitat and morphology within and among varieties of E. stricta on the island Gotland in the Baltic Sea. The results are discussed in relation to evolutionary processes acting within this species complex. In a study conducted at the regional scale, the two early-flowering varieties suecica and tenuis each formed a genetically distinct group, while the three late-flowering varieties brevipila, gotlandica and stricta formed a third group. The results suggest that suecica and tenuis have ancient origins since they are genetically different both from the brevipila/gotlandica/stricta group and from each other despite their similar habitat preferences. This pattern was obtained using both marker systems. Discrepancies between AFLP and microsatellites were found in patterns of isolation by distance and in estimates of expected heterozygosity, He. Focusing on the mixed genetic group brevipila/gotlandica/stricta and the causes behind their clustering together despite differences in morphology and habitat preferences, I performed a study at a smaller geographic scale. Studying a population of E. stricta I found that, although gene flow within the population was strong, it had not prevented the formation of genetic groups associated with micro-habitat properties. An important result for conservation of the rare variety suecica is its distinct genetic separation from variety tenuis. If the aim of conservation is to preserve the uniqueness of suecica, the two varieties should be treated as separated entities.

<p>The application of pulse shape analysis (PSA) and γ-ray tracking techniques has attracted a great deal of interest in the recent years in fields ranging from nuclear structure studies to medical imaging. These new data analysis methods add position sensitivity as well as directional information for the detected γ-rays to the excellent energy resolution of germanium detectors. This thesis focuses on the application of PSA on planar segmented germanium detectors, divided into three separate studies. The pulse shape analysis technique known as the matrix method was chosen due to its ability to treat events with arbitrary number and combinations of interactions within a single detector. It has been applied in two experiments with the 25-fold segmented planar pixel detector -- imaging and polarization measurements -- as well as in a simulation of upcoming detectors for DESPEC at NuSTAR/FAIR.</p><p>In the first experiment, a point source of ¹³⁷Cs was imaged. Events where the 662 keV γ-rays scattered once and were then absorbed in a different segment were treated by the PSA algorithm in order to find the locations of these interactions. The Compton scattering formula was then used to determine the direction to the source. The experiment has provided a robust test of the performance of the PSA algorithm on multiple interaction events, in particular those with interactions in adjacent segments, as well as allowed to estimate the realistically attainable position resolution. In the second experiment, the response of the detector to polarized photons of 288 keV was studied. The polarization of photons can be measured through the observation of the angular distribution of Compton-scattered photons, Hence the ability to resolve the interaction locations had once again proven useful.</p><p>The third study is focused on the performance of the proposed planar germanium detectors for the DESPEC array. As these detectors have not yet been manufactured at the time of this writing, a set of data simulated in GEANT4 was used. The detector response was calculated for two of the possible segmentation patterns -- that with a single pixelated contact and one where both contacts are segmented into mutually orthogonal strips. In both cases, PSA was applied in order to reconstruct the interaction locations from this response. It was found that the double-sided strip detector can achieve an over-all better position resolution with a given number of readout channels. However, this comes at the expense of a small number of complex events where the reconstruction fails. These results have also been compared to the performance of the 25-fold pixelated detector.</p>

While most statistical applications involve real numbers, some demand complex numbers. Statistical shape analysis is one such area. The complex Bingham distribution is utilized in the shape analysis of landmark data in two dimensions. Previous analysis of data arising from this distribution involved classical statistical techniques. In this report, a full Bayesian inference was carried out on the posterior distribution of the parameter matrix when data arise from a complex Bingham distribution. We utilized a Markov chain Monte Carlo algorithm to sample the posterior distribution of the parameters. A Metropolis-Hastings algorithm sampled the posterior conditional distribution of the eigenvalues while a successive conditional Monte Carlo sampler was used to sample the eigenvectors. The method was successfully verifi ed on simulated data, using both at and informative priors. / text

This dissertation focuses on the mathematical design of a unified shape kernel for geometric computing, with possible applications to computer aided design (CAM) and manufacturing (CAM), solid geometric modelling, free-form modelling of curves and surfaces, feature-based modelling, finite element meshing, computer animation, etc. The generality of such a unified shape kernel grounds on a shape theory for objects in some Euclidean space. Shape does not mean herein only geometry as usual in geometric modelling, but has been extended to other contexts, e. g. topology, homotopy, convexity theory, etc. This shape theory has enabled to make a shape analysis of the current geometric kernels. Significant deficiencies have been then identified in how these geometric kernels represent shapes from different applications. This thesis concludes that it is possible to construct a general shape kernel capable of representing and manipulating general specifications of shape for objects even in higher-dimensional Euclidean spaces, regardless whether such objects are implicitly or parametrically defined, they have ‘incomplete boundaries’ or not, they are structured with more or less detail or subcomplexes, which design sequence has been followed in a modelling session, etc. For this end, the basic constituents of such a general geometric kernel, say a combinatorial data structure and respective Euler operators for n-dimensional regular stratified objects, have been introduced and discussed.

This report describes investigations of the NE213-detector and the possibility to utilize pulse shape analysis to separate neutrons and gammas in a mixed emission field. Neutron fluxes are often contaminated with gammas, to which the detectors are sensitive. Sorting out the unwanted gamma pulses from the interesting neutrons is therefore crucial in many situations, for instance in fusion reactor diagnostics, such as for neutron cameras. This can be done based on pulse shapes, which differ for gammas and neutrons interacting in the NE213-detector. By analyzing the pulse shapes from a digital transient recorder, neutrons can be distinguished from gammas. An experiment with a Cf-252 neutron source was set up and provided data. The separation algorithm was based on charge comparison and gave good results. Furthermore the results of the pulse shape analysis were verified by TOF-measurements. The lowest permissible energy for a reasonable separation was found to be around 0.5 MeV. Some conclusions on the limitations of the equipment were also made.