Conditioning and preconditioning of the variational data assimilation problem

Numerical weather prediction (NWP) centres use numerical models of the atmospheric flow to forecast future weather states from an estimate of the current state. Variational data assimilation (VAR) is used commonly to determine an optimal state estimate that miminizes the errors between observations of the dynamical system and model predictions of the flow. The rate of convergence of the VAR scheme and the sensitivity of the solution to errors in the data are dependent on the condition number of the Hessian of the variational least-squares objective function. The traditional formulation of VAR is ill-conditioned and hence leads to slow convergence and an inaccurate solution. In practice, operational NWP centres precondition the system via a control variable transform to reduce the condition number of the Hessian. In this paper we investigate the conditioning of VAR for a single, periodic, spatially-distributed state variable. We present theoretical bounds on the condition number of the original and preconditioned Hessians and hence demonstrate the improvement produced by the preconditioning. We also investigate theoretically the effect of observation position and error variance on the preconditioned system and show that the problem becomes more ill-conditioned with increasingly dense and accurate observations. Finally, we confirm the theoretical results in an operational setting by giving experimental results from the Met Office variational system.

Conditioning of the weak-constraint variational data assimilation problem for numerical weather prediction

4-Dimensional Variational Data Assimilation (4DVAR) assimilates observations through the minimisation of a least-squares objective function, which is constrained by the model flow. We refer to 4DVAR as strong-constraint 4DVAR (sc4DVAR) in this thesis as it assumes the model is perfect. Relaxing this assumption gives rise to weak-constraint 4DVAR (wc4DVAR), leading to a different minimisation problem with more degrees of freedom. We consider two wc4DVAR formulations in this thesis, the model error formulation and state estimation formulation. The 4DVAR objective function is traditionally solved using gradient-based iterative methods. The principle method used in Numerical Weather Prediction today is the Gauss-Newton approach. This method introduces a linearised `inner-loop' objective function, which upon convergence, updates the solution of the non-linear `outer-loop' objective function. This requires many evaluations of the objective function and its gradient, which emphasises the importance of the Hessian. The eigenvalues and eigenvectors of the Hessian provide insight into the degree of convexity of the objective function, while also indicating the difficulty one may encounter while iterative solving 4DVAR. The condition number of the Hessian is an appropriate measure for the sensitivity of the problem to input data. The condition number can also indicate the rate of convergence and solution accuracy of the minimisation algorithm. This thesis investigates the sensitivity of the solution process minimising both wc4DVAR objective functions to the internal assimilation parameters composing the problem. We gain insight into these sensitivities by bounding the condition number of the Hessians of both objective functions. We also precondition the model error objective function and show improved convergence. We show that both formulations' sensitivities are related to error variance balance, assimilation window length and correlation length-scales using the bounds. We further demonstrate this through numerical experiments on the condition number and data assimilation experiments using linear and non-linear chaotic toy models.

Development of computational tools for the integrated analysis of DNA microarray data with applications in cancer research

The MAP-i Doctoral Program of the Universities of Minho, Aveiro and Porto

Integrating data from heterogeneous DNA microarray platforms

DNA microarrays are one of the most used technologies for gene expression measurement. However, there are several distinct microarray platforms, from different manufacturers, each with its own measurement protocol, resulting in data that can hardly be compared or directly integrated. Data integration from multiple sources aims to improve the assertiveness of statistical tests, reducing the data dimensionality problem. The integration of heterogeneous DNA microarray platforms comprehends a set of tasks that range from the re-annotation of the features used on gene expression, to data normalization and batch effect elimination. In this work, a complete methodology for gene expression data integration and application is proposed, which comprehends a transcript-based re-annotation process and several methods for batch effect attenuation. The integrated data will be used to select the best feature set and learning algorithm for a brain tumor classification case study. The integration will consider data from heterogeneous Agilent and Affymetrix platforms, collected from public gene expression databases, such as The Cancer Genome Atlas and Gene Expression Omnibus.

Comparison of Contemporary Protocols for High-speed Data Transport via 10 Gbps WAN Connections

This work is dedicated to comparison of open source as well as proprietary transport protocols for highspeed data transmission via IP networks. The contemporary common TCP needs significant improvement since it was developed as general-purpose transport protocol and firstly introduced four decades ago. In nowadays networks, TCP fits not all communication needs that society has. Caused of it another transport protocols have been developed and successfully used for e.g. Big Data movement. In scope of this research the following protocols have been investigated for its efficiency on 10Gbps links: UDT, RBUDP, MTP and RWTP. The protocols were tested under different impairments such as Round Trip Time up to 400 ms and packet losses up to 2%. Investigated parameters are the data rate under different conditions of the network, the CPU load by sender andreceiver during the experiments, size of feedback data, CPU usage per Gbps and the amount of feedback data per GiByte of effectively transmitted data. The best performance and fair resources consumption was observed by RWTP. From the opensource projects, the best behavior is showed by RBUDP.

Modelling structural zeros in compositional data

This analysis was stimulated by the real data analysis problem of householdexpenditure data. The full dataset contains expenditure data for a sample of 1224 households. The expenditure is broken down at 2 hierarchical levels: 9 major levels (e.g. housing, food, utilities etc.) and 92 minor levels. There are also 5 factors and 5 covariates at the household level. Not surprisingly, there are a small number of zeros at the major level, but many zeros at the minor level. The question is how best to model the zeros. Clearly, models that tryto add a small amount to the zero terms are not appropriate in general as at least some of the zeros are clearly structural, e.g. alcohol/tobacco for households that are teetotal. The key question then is how to build suitable conditional models. For example, is the sub-composition of spendingexcluding alcohol/tobacco similar for teetotal and non-teetotal households?In other words, we are looking for sub-compositional independence. Also, what determines whether a household is teetotal? Can we assume that it is independent of the composition? In general, whether teetotal will clearly depend on the household level variables, so we need to be able to model this dependence. The other tricky question is that with zeros on more than onecomponent, we need to be able to model dependence and independence of zeros on the different components. Lastly, while some zeros are structural, others may not be, for example, for expenditure on durables, it may be chance as to whether a particular household spends money on durableswithin the sample period. This would clearly be distinguishable if we had longitudinal data, but may still be distinguishable by looking at the distribution, on the assumption that random zeros will usually be for situations where any non-zero expenditure is not small.While this analysis is based on around economic data, the ideas carry over tomany other situations, including geological data, where minerals may be missing for structural reasons (similar to alcohol), or missing because they occur only in random regions which may be missed in a sample (similar to the durables)

The requirements and needs of global data usage in product lifecycle management

This study was done for ABB Ltd. Motors and Generators business unit in Helsinki. In this study, global data movement in large businesses is examined from a product data management (PDM) and enterprise resource planning (ERP) point-of-view. The purpose of this study was to understand and map out how a large global business handles its data in a multiple site structure and how it can be applied in practice. This was done by doing an empirical interview study on five different global businesses with design locations in multiple countries. Their master data management (MDM) solutions were inspected and analyzed to understand which solution would best benefit a large global architecture with many design locations. One working solution is a transactional hub which negates the effects of multisite transfers and reduces lead times. Also, the requirements and limitations of the current MDM architecture were analyzed and possible reform ideas given.  

Contributions des voies vestibulospinale et corticospinale au contrôle des mouvements du bras

Le contrôle des mouvements du bras fait intervenir plusieurs voies provenant du cerveau. Cette thèse, composée principalement de deux études, tente d’éclaircir les contributions des voies tirant leur origine du système vestibulaire et du cortex moteur. Dans la première étude (Raptis et al 2007), impliquant des mouvements d’atteinte, nous avons cerné l’importance des voies descendantes partant du système vestibulaire pour l’équivalence motrice, i.e. la capacité du système moteur à atteindre un but moteur donné lorsque le nombre de degrés de liberté articulaires varie. L’hypothèse émise était que le système vestibulaire joue un rôle essentiel dans l’équivalence motrice. Nous avons comparé la capacité d’équivalence motrice de sujets sains et de patients vestibulodéficients chroniques lors de mouvements nécessitant un contrôle des positions du bras et du tronc. Pendant que leur vision était temporairement bloquée, les sujets devaient soit maintenir une position de l’index pendant une flexion du tronc, soit atteindre une cible dans l’espace péri-personnel en combinant le mouvement du bras avec une flexion du tronc. Lors d’essais déterminés aléatoirement et imprévus par les participants, leur tronc était retenu par un mécanisme électromagnétique s’activant en même temps que le signal de départ. Les sujets sains ont pu préserver la position ou la trajectoire de l’index dans les deux conditions du tronc (libre, bloqué) en adaptant avec une courte latence (60-180 ms) les mouvements articulaires au niveau du coude et de l’épaule. En comparaison, six des sept patients vestibulodéficients chroniques ont présenté des déficits au plan des adaptations angulaires compensatoires. Pour ces patients, entre 30 % et 100 % du mouvement du tronc n’a pas été compensé et a été transmis à la position ou trajectoire de l’index. Ces résultats indiqueraient que les influences vestibulaires évoquées par le mouvement de la tête pendant la flexion du tronc jouent un rôle majeur pour garantir l’équivalence motrice dans ces tâches d’atteinte lorsque le nombre de degrés de liberté articulaires varie. Également, ils démontrent que la plasticité de long terme survenant spontanément après une lésion vestibulaire unilatérale complète ne serait pas suffisante pour permettre au SNC de retrouver un niveau d’équivalence motrice normal dans les actions combinant un déplacement du bras et du tronc. Ces tâches de coordination bras-tronc constituent ainsi une approche inédite et sensible pour l’évaluation clinique des déficits vestibulaires. Elles permettent de sonder une dimension fonctionnelle des influences vestibulaires qui n’était pas prise en compte dans les tests cliniques usuels, dont la sensibilité relativement limitée empêche souvent la détection d’insuffisances vestibulaires six mois après une lésion de ces voies. Avec cette première étude, nous avons donc exploré comment le cerveau et les voies descendantes intègrent des degrés de liberté articulaires supplémentaires dans le contrôle du bras. Dans la seconde étude (Raptis et al 2010), notre but était de clarifier la nature des variables spécifiées par les voies descendantes pour le contrôle d’actions motrices réalisées avec ce membre. Nous avons testé l’hypothèse selon laquelle les voies corticospinales contrôlent la position et les mouvements des bras en modulant la position-seuil (position de référence à partir de laquelle les muscles commencent à être activés en réponse à une déviation de cette référence). Selon ce principe, les voies corticospinales ne spécifieraient pas directement les patrons d’activité EMG, ce qui se refléterait par une dissociation entre l’EMG et l’excitabilité corticospinale pour des positions-seuils différentes. Dans un manipulandum, des participants (n=16) ont modifié leur angle du poignet, d’une position de flexion (45°) à une position d’extension (-25°), et vice-versa. Les forces élastiques passives des muscles ont été compensées avec un moteur couple afin que les sujets puissent égaliser leur activité EMG de base dans les deux positions. L’excitabilité motoneuronale dans ces positions a été comparée à travers l’analyse des réponses EMG évoquées à la suite d’étirements brefs. Dans les deux positions, le niveau d’EMG et l’excitabilité motoneuronale étaient semblables. De plus, ces tests ont permis de montrer que le repositionnement du poignet était associé à une translation de la position-seuil. Par contre, malgré la similitude de l’excitabilité motoneuronale dans ces positions, l’excitabilité corticospinale des muscles du poignet était significativement différente : les impulsions de stimulation magnétique transcrânienne (TMS; à 1.2 MT, sur l’aire du poignet de M1) ont provoqué des potentiels moteurs évoqués (MEP) de plus grande amplitude en flexion pour les fléchisseurs comparativement à la position d’extension et vice-versa pour les extenseurs (p<0.005 pour le groupe). Lorsque les mêmes positions étaient établies après une relaxation profonde, les réponses réflexes et les amplitudes des MEPs ont drastiquement diminué. La relation caractéristique observée entre position physique et amplitude des MEPs dans le positionnement actif s’est aussi estompée lorsque les muscles étaient relâchés. Cette étude suggère que la voie corticospinale, en association avec les autres voies descendantes, participerait au contrôle de la position-seuil, un processus qui prédéterminerait le référentiel spatial dans lequel l’activité EMG émerge. Ce contrôle de la « référence » constituerait un principe commun s’appliquant à la fois au contrôle de la force musculaire, de la position, du mouvement et de la relaxation. Nous avons aussi mis en évidence qu’il est nécessaire, dans les prochaines recherches ou applications utilisant la TMS, de prendre en compte la configuration-seuil des articulations, afin de bien interpréter les réponses musculaires (ou leurs changements) évoquées par cette technique; en effet, la configuration-seuil influencerait de manière notable l’excitabilité corticomotrice, qui peut être considérée comme un indicateur non seulement lors d’activités musculaires, mais aussi cognitives, après apprentissages moteurs ou lésions neurologiques causant des déficits moteurs (ex. spasticité, faiblesse). Considérées dans leur ensemble, ces deux études apportent un éclairage inédit sur des principes fondamentaux du contrôle moteur : nous y illustrons de manière plus large le rôle du système vestibulaire dans les tâches d’atteinte exigeant une coordination entre le bras et son « support » (le tronc) et clarifions l’implication des voies corticomotrices dans la spécification de paramètres élémentaires du contrôle moteur du bras. De plus amples recherches sont cependant nécessaires afin de mieux comprendre comment les systèmes sensoriels et descendants (e.g. vestibulo-, réticulo-, rubro-, propriospinal) participent et interagissent avec les signaux corticofugaux afin de spécifier les seuils neuromusculaires dans le contrôle de la posture et du mouvement.

‘Modeling and Analysis of Competing Risks Data’

there has been much research on analyzing various forms of competing risks data. Nevertheless, there are several occasions in survival studies, where the existing models and methodologies are inadequate for the analysis competing risks data. ldentifiabilty problem and various types of and censoring induce more complications in the analysis of competing risks data than in classical survival analysis. Parametric models are not adequate for the analysis of competing risks data since the assumptions about the underlying lifetime distributions may not hold well. Motivated by this, in the present study. we develop some new inference procedures, which are completely distribution free for the analysis of competing risks data.

Modelling structural zeros in compositional data

This analysis was stimulated by the real data analysis problem of household expenditure data. The full dataset contains expenditure data for a sample of 1224 households. The expenditure is broken down at 2 hierarchical levels: 9 major levels (e.g. housing, food, utilities etc.) and 92 minor levels. There are also 5 factors and 5 covariates at the household level. Not surprisingly, there are a small number of zeros at the major level, but many zeros at the minor level. The question is how best to model the zeros. Clearly, models that try to add a small amount to the zero terms are not appropriate in general as at least some of the zeros are clearly structural, e.g. alcohol/tobacco for households that are teetotal. The key question then is how to build suitable conditional models. For example, is the sub-composition of spending excluding alcohol/tobacco similar for teetotal and non-teetotal households? In other words, we are looking for sub-compositional independence. Also, what determines whether a household is teetotal? Can we assume that it is independent of the composition? In general, whether teetotal will clearly depend on the household level variables, so we need to be able to model this dependence. The other tricky question is that with zeros on more than one component, we need to be able to model dependence and independence of zeros on the different components. Lastly, while some zeros are structural, others may not be, for example, for expenditure on durables, it may be chance as to whether a particular household spends money on durables within the sample period. This would clearly be distinguishable if we had longitudinal data, but may still be distinguishable by looking at the distribution, on the assumption that random zeros will usually be for situations where any non-zero expenditure is not small. While this analysis is based on around economic data, the ideas carry over to many other situations, including geological data, where minerals may be missing for structural reasons (similar to alcohol), or missing because they occur only in random regions which may be missed in a sample (similar to the durables)

A wavenumber independent boundary element method for an acoustic scattering problem

In this paper we consider the impedance boundary value problem for the Helmholtz equation in a half-plane with piecewise constant boundary data, a problem which models, for example, outdoor sound propagation over inhomogeneous. at terrain. To achieve good approximation at high frequencies with a relatively low number of degrees of freedom, we propose a novel Galerkin boundary element method, using a graded mesh with smaller elements adjacent to discontinuities in impedance and a special set of basis functions so that, on each element, the approximation space contains polynomials ( of degree.) multiplied by traces of plane waves on the boundary. We prove stability and convergence and show that the error in computing the total acoustic field is O( N-(v+1) log(1/2) N), where the number of degrees of freedom is proportional to N logN. This error estimate is independent of the wavenumber, and thus the number of degrees of freedom required to achieve a prescribed level of accuracy does not increase as the wavenumber tends to infinity.

A high-wavenumber boundary-element method for an acoustic scattering problem

In this paper we show stability and convergence for a novel Galerkin boundary element method approach to the impedance boundary value problem for the Helmholtz equation in a half-plane with piecewise constant boundary data. This problem models, for example, outdoor sound propagation over inhomogeneous flat terrain. To achieve a good approximation with a relatively low number of degrees of freedom we employ a graded mesh with smaller elements adjacent to discontinuities in impedance, and a special set of basis functions for the Galerkin method so that, on each element, the approximation space consists of polynomials (of degree $\nu$) multiplied by traces of plane waves on the boundary. In the case where the impedance is constant outside an interval $[a,b]$, which only requires the discretization of $[a,b]$, we show theoretically and experimentally that the $L_2$ error in computing the acoustic field on $[a,b]$ is ${\cal O}(\log^{\nu+3/2}|k(b-a)| M^{-(\nu+1)})$, where $M$ is the number of degrees of freedom and $k$ is the wavenumber. This indicates that the proposed method is especially commendable for large intervals or a high wavenumber. In a final section we sketch how the same methodology extends to more general scattering problems.

Efficient non-linear data assimilation in geophysical fluid dynamics

New ways of combining observations with numerical models are discussed in which the size of the state space can be very large, and the model can be highly nonlinear. Also the observations of the system can be related to the model variables in highly nonlinear ways, making this data-assimilation (or inverse) problem highly nonlinear. First we discuss the connection between data assimilation and inverse problems, including regularization. We explore the choice of proposal density in a Particle Filter and show how the ’curse of dimensionality’ might be beaten. In the standard Particle Filter ensembles of model runs are propagated forward in time until observations are encountered, rendering it a pure Monte-Carlo method. In large-dimensional systems this is very inefficient and very large numbers of model runs are needed to solve the data-assimilation problem realistically. In our approach we steer all model runs towards the observations resulting in a much more efficient method. By further ’ensuring almost equal weight’ we avoid performing model runs that are useless in the end. Results are shown for the 40 and 1000 dimensional Lorenz 1995 model.

Extracting quantitative structural parameters for disordered polymers from neutron scattering data

The organization of non-crystalline polymeric materials at a local level, namely on a spatial scale between a few and 100 a, is still unclear in many respects. The determination of the local structure in terms of the configuration and conformation of the polymer chain and of the packing characteristics of the chain in the bulk material represents a challenging problem. Data from wide-angle diffraction experiments are very difficult to interpret due to the very large amount of information that they carry, that is the large number of correlations present in the diffraction patterns.We describe new approaches that permit a detailed analysis of the complex neutron diffraction patterns characterizing polymer melts and glasses. The coupling of different computer modelling strategies with neutron scattering data over a wide Q range allows the extraction of detailed quantitative information on the structural arrangements of the materials of interest. Proceeding from modelling routes as diverse as force field calculations, single-chain modelling and reverse Monte Carlo, we show the successes and pitfalls of each approach in describing model systems, which illustrate the need to attack the data analysis problem simultaneously from several fronts.