La fusion d’informations a été largement étudiée dans le domaine de l’intelligence artificielle. Une information est en général considérée comme imparfaite. Par conséquent, la combinaison de plusieurs sources d’informations (éventuellement hétérogènes) peut conduire à une information plus globale et complète. Dans le domaine de la fusion on distingue généralement les approches probabilistes et non probabilistes dont fait partie la théorie de l’évidence, développée dans les années 70. Cette méthode permet de représenter à la fois, l’incertitude et l’imprécision de l’information, par l’attribution de fonctions de masses qui s’appliquent non pas à une seule hypothèse (ce qui est le cas le plus courant pour les méthodes probabilistes) mais à un ensemble d’hypothèses. Les travaux présentés dans cette thèse concernent la fusion d’informations pour la segmentation d’images.Pour développer cette méthode nous sommes partis de l’algorithme de la « Ligne de Partage des Eaux » (LPE) qui est un des plus utilisés en détection de contours. Intuitivement le principe de la LPE est de considérer l’image comme un relief topographique où la hauteur d’un point correspond à son niveau de gris. On suppose alors que ce relief se remplit d’eau par des sources placées au niveau des minima locaux de l’image, formant ainsi des bassins versants. Les LPE sont alors les barrages construits pour empêcher les eaux provenant de différents bassins de se mélanger. Un problème de cette méthode de détection de contours est que la LPE directement appliquée sur l’image engendre une sur-segmentation, car chaque minimum local engendre une région. Meyer et Beucher ont proposé de résoudre cette question en spécifiant un ensemble de marqueurs qui seront les seules sources d’inondation du relief. L'extraction automatique des marqueurs à partir des images ne conduit pas toujours à un résultat satisfaisant, en particulier dans le cas d'images complexes. Plusieurs méthodes ont été proposées pour déterminer automatiquement ces marqueurs.Nous nous sommes en particulier intéressés à l’approche stochastique d’Angulo et Jeulin qui estiment une fonction de densité de probabilité (fdp) d'un contour (LPE) après M simulations de la segmentation LPE classique. N marqueurs sont choisis aléatoirement pour chaque réalisation. Par conséquent, une valeur de fdp élevée est attribuée aux points de contours correspondant aux fortes réalisations. Mais la décision d’appartenance d’un point à la « classe contour » reste dépendante d’une valeur de seuil. Un résultat unique ne peut donc être obtenu.Pour augmenter la robustesse de cette méthode et l’unicité de sa réponse, nous proposons de combiner des informations grâce à la théorie de l’évidence.La LPE se calcule généralement à partir de l’image gradient, dérivée du premier ordre, qui donne une information globale sur les contours dans l’image. Alors que la matrice Hessienne, matrice des dérivées d’ordre secondaire, donne une information plus locale sur les contours. Notre objectif est donc de combiner ces deux informations de nature complémentaire en utilisant la théorie de l’évidence. Les différentes versions de la fusion sont testées sur des images réelles de la base de données Berkeley. Les résultats sont comparés avec cinq segmentations manuelles fournies, en tant que vérités terrain, avec cette base de données. La qualité des segmentations obtenues par nos méthodes sont fondées sur différentes mesures: l’uniformité, la précision, l’exactitude, la spécificité, la sensibilité ainsi que la distance métrique de Hausdorff / Information fusion has been widely studied in the field of artificial intelligence. Information is generally considered imperfect. Therefore, the combination of several sources of information (possibly heterogeneous) can lead to a more comprehensive and complete information. In the field of fusion are generally distinguished probabilistic approaches and non-probabilistic ones which include the theory of evidence, developed in the 70s. This method represents both the uncertainty and imprecision of the information, by assigning masses not only to a hypothesis (which is the most common case for probabilistic methods) but to a set of hypothesis. The work presented in this thesis concerns the fusion of information for image segmentation.To develop this method we start with the algorithm of Watershed which is one of the most used methods for edge detection. Intuitively the principle of the Watershed is to consider the image as a landscape relief where heights of the different points are associated with grey levels. Assuming that the local minima are pierced with holes and the landscape is immersed in a lake, the water filled up from these minima generate the catchment basins, whereas watershed lines are the dams built to prevent mixing waters coming from different basins.The watershed is practically applied to the gradient magnitude, and a region is associated with each minimum. Therefore the fluctuations in the gradient image and the great number of local minima generate a large set of small regions yielding an over segmented result which can hardly be useful. Meyer and Beucher proposed seeded watershed or marked-controlled watershed to surmount this oversegmentation problem. The essential idea of the method is to specify a set of markers (or seeds) to be considered as the only minima to be flooded by water. The number of detected objects is therefore equal to the number of seeds and the result is then markers dependent. The automatic extraction of markers from the images does not lead to a satisfying result especially in the case of complex images. Several methods have been proposed for automatically determining these markers.We are particularly interested in the stochastic approach of Angulo and Jeulin who calculate a probability density function (pdf) of contours after M simulations of segmentation using conventional watershed with N markers randomly selected for each simulation. Therefore, a high pdf value is assigned to strong contour points that are more detected through the process. But the decision that a point belong to the "contour class" remains dependent on a threshold value. A single result cannot be obtained.To increase the robustness of this method and the uniqueness of its response, we propose to combine information with the theory of evidence.The watershed is generally calculated on the gradient image, first order derivative, which gives comprehensive information on the contours in the image.While the Hessian matrix, matrix of second order derivatives, gives more local information on the contours. Our goal is to combine these two complementary information using the theory of evidence. The method is tested on real images from the Berkeley database. The results are compared with five manual segmentation provided as ground truth, with this database. The quality of the segmentation obtained by our methods is tested with different measures: uniformity, precision, recall, specificity, sensitivity and the Hausdorff metric distance

Doctor of Philosophy / Department of Entomology / Ming-Shun Chen / Gerald E. Wilde / Analysis of transcriptomes from salivary glands and midgut of the Hessian fly [Mayetiola destructor (Say)] identified a diverse set of cDNAs that were categorized into five groups, group I – V, based on their phylogenetic relationship. All five of these groups may encode putative protease inhibitors based on structural similarity with known proteins. The sequences of these putative proteins among different groups are highly diversified. However, sequence identity and structural analysis of the proteins revealed that all of them contained high cysteine residues that were completely conserved at their respective positions among these otherwise diversified proteins. Analysis of bacterial artificial chromosome (BAC) DNA for two groups, group I (11A6) and group II (14A4), indicated that group I might be a single copy gene or genes with low copy number whereas group II exists as multiple copies clustered within the Hessian fly genome. To test the inhibitory activity and specificity of these putative proteins, recombinant proteins were generated. Enzymatic analysis of the recombinant proteins against commercial and insect gut proteases demonstrated that recombinant proteins indeed are strong inhibitors of proteases with different specificities. Northern analysis of the representative members of five groups revealed that the group I-IV genes were expressed exclusively in the larval stage with variations among groups at different larval stages. The group V (11C4) genes were expressed in the late larval and pupal stage. Tissue specific gene expression analysis revealed that group I-IV genes were predominantly expressed in malpighian tubules whereas the group V genes were abundantly expressed in the salivary glands. Localization experiments with the antibody for representative members from group II (14A4) demonstrated that the protein was predominantly localized in the malpighian tubules and in low amounts in the midgut, suggesting that malpighian tubules are the primary tissue of 14A4 inhibitor synthesis. The overall results indicated that the Hessian fly contains a complex network of genes that code for protease inhibitors which regulate protease activities through different developmental stages of the insect.

Doctor of Philosophy / Department of Agronomy / Gina Brown-Guedira / Allan K. Fritz / Aegilops tauschii, the D-genome donor of modern wheat, has served as an important source of genetic variation in wheat breeding. The objective of this research was to characterize and utilize multiple Ae. tauschii-derived pest resistance genes contained in the germplasm KS96WGRC40. Two Ae. tauschii-derived genes, H23 and Cmc4, provide resistance to the Hessian fly (HF) and wheat curl mite (WCM), respectively. A linkage analysis of a testcross population estimated 32.67% recombination between H23 and Cmc4 on chromosome 6DS in wheat indicating that the two genes are not tightly linked as previous mapping reports show. Haplotype data of recombinant lines and physical mapping of linked microsatellite markers located Cmc4 distal to H23. Haplotype data indicated that both KS89WGRC04 and KS96WGRC40 have the distal portion of 6DS derived from Ae. tauschii. Microsatellite primer pairs BARC183 and GDM036 were more useful than the previously published linked markers in identifying lines carrying Cmc4 and H23, respectively. Through phenotypic selection and advancement within the testcross population, three TC1F2:4 lines were identified as homozygous resistant for H23 and Cmc4 and have the complete terminal segment of 6DS from Ae. tauschii. Two lines are more desirable than the original germplasm releases and can serve as a source of resistance to both HF and WCM in an elite background. A linkage analysis of a segregating recombinant inbred line population identified an Ae. tauschii-derived gene of major effect conferring resistance to Septoria leaf blotch (STB) and another Ae. tauschii-derived gene of major effect conferring resistance to soil-borne wheat mosaic virus (SBWMV) in the germplasm KS96WGRC40. The STB resistance gene in KS96WGRC40 is located in the distal 40% of the short arm of chromosome 7D flanked by microsatellite markers Xgwm044 and Xbarc126. Two previously reported STB genes, Stb4 and Stb5, have been mapped to 7DS in the same region as the STB resistance gene in KS96WGRC40. The uniqueness of the STB resistance genes on 7DS is questionable. The SBWMV resistance gene in KS96WGRC40 is located on chromosome 5DL linked to microsatellite marker Xcfd010. The SBWMV resistance gene within KS96WGRC40 was derived from TA2397 via KS95WGRC33.

Distributed parameter systems (DPS) comprise an important class of engineering systems ranging from "traditional" such as tubular reactors, to cutting edge processes such as nano-scale coatings. DPS have been studied extensively and significant advances have been noted, enabling their accurate simulation. To this end a variety of tools have been developed. However, extending these advances for systems design is not a trivial task . Rigorous design and operation policies entail systematic procedures for optimisation and control. These tasks are "upper-level" and utilize existing models and simulators. The higher the accuracy of the underlying models, the more the design procedure benefits. However, employing such models in the context of conventional algorithms may lead to inefficient formulations. The optimisation and control of DPS is a challenging task. These systems are typically discretised over a computational mesh, leading to large-scale problems. Handling the resulting large-scale systems may prove to be an intimidating task and requires special methodologies. Furthermore, it is often the case that the underlying physical phenomena span various temporal and spatial scales, thus complicating the analysis. Stiffness may also potentially be exhibited in the (nonlinear) models of such phenomena. The objective of this work is to design reliable and practical procedures for the optimisation and control of DPS. It has been observed in many systems of engineering interest that although they are described by infinite-dimensional Partial Differential Equations (PDEs) resulting in large discretisation problems, their behaviour has a finite number of significant components , as a result of their dissipative nature. This property has been exploited in various systematic model reduction techniques. Of key importance in this work is the identification of a low-dimensional dominant subspace for the system. This subspace is heuristically found to correspond to part of the eigenspectrum of the system and can therefore be identified efficiently using iterative matrix-free techniques. In this light, only low-dimensional Jacobians and Hessian matrices are involved in the formulation of the proposed algorithms, which are projections of the original matrices onto appropriate low-dimensional subspaces, computed efficiently with directional perturbations.The optimisation algorithm presented employs a 2-step projection scheme, firstly onto the dominant subspace of the system (corresponding to the right-most eigenvalues of the linearised system) and secondly onto the subspace of decision variables. This algorithm is inspired by reduced Hessian Sequential Quadratic Programming methods and therefore locates a local optimum of the nonlinear programming problem given by solving a sequence of reduced quadratic programming (QP) subproblems . This optimisation algorithm is appropriate for systems with a relatively small number of decision variables. Inequality constraints can be accommodated following a penalty-based strategy which aggregates all constraints using an appropriate function , or by employing a partial reduction technique in which only equality constraints are considered for the reduction and the inequalities are linearised and passed on to the QP subproblem . The control algorithm presented is based on the online adaptive construction of low-order linear models used in the context of a linear Model Predictive Control (MPC) algorithm , in which the discrete-time state-space model is recomputed at every sampling time in a receding horizon fashion. Successive linearisation around the current state on the closed-loop trajectory is combined with model reduction, resulting in an efficient procedure for the computation of reduced linearised models, projected onto the dominant subspace of the system. In this case, this subspace corresponds to the eigenvalues of largest magnitude of the discretised dynamical system. Control actions are computed from low-order QP problems solved efficiently online.The optimisation and control algorithms presented may employ input/output simulators (such as commercial packages) extending their use to upper-level tasks. They are also suitable for systems governed by microscopic rules, the equations of which do not exist in closed form. Illustrative case studies are presented, based on tubular reactor models, which exhibit rich parametric behaviour.

In recent years, mobile robots have gained tremendous attention from the entire society: the industry is aiming at selling more intelligent products while the academia is improving their performance from all perspectives. Real world examples include autnomous driving vehicles, multirotors, legged robots, etc. One of the challenging tasks commonly faced by all game players, and all robotics platforms, is to plan motion or locomotion of the robot, calculate an optimal trajectory according to certain criterion and control it accordingly. Difficulty of solving such task usually arises from high-dimensionality and complexity of the system dynamics, fast changing conditions imposed as constraints and necessity for real-time deployment. This work proposes a method over the aforementioned mission by solving an optimal control problem in a receding horizon fashion. Unlike the existing Sequential Linear Quadratic [1] algorithm which is a continuous-time variant of Differential Dynamic Programming [2], we tackle the problem in a discretized multiple shooting fashion. Sequential Quadratic Programming is employed as optimization technique to solve the constrained Nonlinear Programming iteratively. Moreover, we apply trust region method in the sub Quadratic Programming to handle potential indefiniteness of Hessian matrix as well as to improve robustness of the solver. Simulation and benchmark with previous method have been conducted on robotics platforms to show the effectiveness of our solution and superiority under certain circumstances. Experiments have demonstrated that our method is capable of generating trajectories under complicated scenarios where the Hessian matrix contains negative eigenvalues (e.g. obstacle avoidance). / De senaste åren har mobila robotar fått enorm uppmärksamhet från hela samhället: branschen siktar på att sälja mer intelligenta produkter samtidigt som akademin förbättrar sina prestationer ur alla perspektiv. Exempel på verkligheten inkluderar autonoma körande fordon, multirotorer, robotar med ben, etc. En av de utmanande uppgifterna som vanligtvis alla spelare och alla robotplattformar står inför är att planera robotens rörelse eller rörelse, beräkna en optimal bana enligt vissa kriterier och kontrollera det därefter. Svårigheter att lösa en sådan uppgift beror vanligtvis på hög dimensionalitet och komplexitet hos systemdynamiken, snabbt föränderliga villkor som åläggs som begränsningar och nödvändighet för realtidsdistribution. Detta arbete föreslår en metod över det tidigare nämnda uppdraget genom att lösa ett optimalt kontrollproblem på ett vikande horisont. Till skillnad från den befintliga Sequential Linear Quadratic [1] algoritmen som är en kontinuerlig tidsvariant av Differential Dynamic Programming [2], tar vi oss an problemet på ett diskretiserat multipelfotograferingssätt. Sekventiell kvadratisk programmering används som optimeringsteknik för att lösa den begränsade olinjära programmeringen iterativt. Dessutom tillämpar vi trust region-metoden i den sub-kvadratiska programmeringen för att hantera potentiell obestämdhet av hessisk matris samt för att förbättra lösarens robusthet. Simulering och benchmark med tidigare metod har utförts på robotplattformar för att visa effektiviteten hos vår lösning och överlägsenhet under vissa omständigheter. Experiment har visat att vår metod är kapabel att generera banor under komplicerade scenarier där den hessiska matrisen innehåller negativa egenvärden (t.ex. undvikande av hinder).

Actuellement, le principal obstacle à la mise en œuvre de la FWI élastique en trois dimensions sur des cas d'étude réalistes réside dans le coût de calcul associé aux taches de modélisation sismique. Pour surmonter cette difficulté, je propose deux contributions. Tout d'abord, je propose de calculer le gradient de la fonctionnelle avec la méthode de l'état adjoint à partir d'une forme symétrisée des équations de l'élastodynamique formulées sous forme d'un système du premier ordre en vitesse-contrainte. Cette formulation auto-adjointe des équations de l'élastodynamique permet de calculer les champs incidents et adjoints intervenant dans l'expression du gradient avec un seul opérateur de modélisation numérique. Le gradient ainsi calculé facilite également l'interfaçage de plusieurs outils de modélisation avec l'algorithme d'inversion. Deuxièmement, j'explore dans cette thèse dans quelle mesure les encodages des sources avec des algorithmes d'optimisation du second-ordre de quasi-Newton et de Newton tronqué permettait de réduire encore le coût de la FWI. Finalement, le problème d'optimisation associé à la FWI est mal posé, nécessitant ainsi d'ajouter des contraintes de régularisation à la fonctionnelle à minimiser. Je montre ici comment une régularisation fondée sur la variation totale du modèle fournissait une représentation adéquate des modèles du sous-sol en préservant le caractère discontinu des interfaces lithologiques. Pour améliorer les images du sous-sol, je propose un algorithme de débruitage fondé sur une variation totale locale au sein duquel j'incorpore l'information structurale fournie par une image migrée pour préserver les structures de faible dimension. / Currently, the main limitation to perform 3D elastic full waveform inversion on a production level is the computational cost it represents. With this in mind, we provide two contributions. First, we develop a self adjoint formulation of the isotropic first order velocity-stress elastic equations that allow to implement only one forward modeling operator in the gradient computation. Second, we combine Newton and quasi-Newton optimization methods with source encoding techniques to see to what extent the computational cost could be further reduced. Finally, the optimization process associated to FWI is ill posed and requires regularization constraints. I show that the total variation of the model as a regularization term provides and adequate description of earth models, preserving the discontinuous character of the lithological layers. To improve the quality of the images, we propose a local total variation denoising algorithm based on the incorporation of the information provided by a migrated image.

<p>Posmatra se problem optimizacije bez ograničenja. Za re&scaron;avanje<br />problema&nbsp; optimizacije bez ograničenja postoji mno&scaron;tvo raznovrsnih<br />metoda. Istraživanje ovde motivisano je potrebom za metodama koje<br />će brzo konvergirati.<br />Cilj je sistematizacija poznatih rezultata, kao i teorijska i numerička<br />analiza mogućnosti uvođenja parametra u gradijentne metode.<br />Najpre se razmatra problem minimizacije konveksne funkcije vi&scaron;e<br />promenljivih.<br />Problem minimizacije konveksne funkcije vi&scaron;e promenljivih ovde se<br />re&scaron;ava bez izračunavanja matrice hesijana, &scaron;to je naročito aktuelno za<br />sisteme velikih dimenzija, kao i za probleme optimizacije kod kojih<br />ne raspolažemo ni tačnom vredno&scaron;ću funkcije cilja, ni tačnom<br />vredno&scaron;ću gradijenta. Deo motivacije za istraživanjem ovde leži i u<br />postojanju problema kod kojih je funkcija cilja rezultat simulacija.<br />Numerički rezultati, predstavljeni u Glavi 6, pokazuju da uvođenje<br />izvesnog parametra može biti korisno, odnosno, dovodi do ubrzanja<br />određenog metoda optimizacije.<br />Takođe se predstavlja jedan novi hibridni metod konjugovanog<br />gradijenta, kod koga je parametar konjugovanog gradijenta<br />konveksna kombinacija dva poznata parametra konjugovanog<br />gradijenta.<br />U prvoj glavi opisuje se motivacija kao i osnovni pojmovi potrebni za<br />praćenje preostalih glava.<br />U drugoj glavi daje se pregled nekih gradijentnih metoda prvog i<br />drugog reda.<br />Četvrta glava sadrži pregled osnovnih pojmova i nekih rezultata<br />vezanih za metode konjugovanih gradijenata.<br />Pomenute glave su tu radi pregleda nekih poznatih rezultata, dok se<br />originalni doprinos predstavlja u trećoj, petoj i &scaron;estoj glavi.<br />U trećoj glavi se opisuje izvesna modifikacija određenog metoda u<br />kome se koristi multiplikativni parametar, izabran na slučajan način.<br />Dokazuje se linearna konvergencija tako formiranog novog metoda.<br />Peta glava sadrži originalne rezultate koji se odnose na metode<br />konjugovanih gradijenata. Naime, u ovoj glavi predstavlja se novi<br />hibridni metod konjugovanih gradijenata, koji je konveksna<br />kombinacija dva poznata metoda konjugovanih gradijenata.<br />U &scaron;estoj glavi se daju rezultati numeričkih eksperimenata, izvr&scaron;enih<br />na&nbsp; izvesnom skupu test funkcija, koji se odnose na metode iz treće i<br />pete glave. Implementacija svih razmatranih algoritama rađena je u<br />paketu MATHEMATICA. Kriterijum upoređivanja je vreme rada<br />centralne procesorske jedinice.6</p> / <p>The problem under consideration is an unconstrained optimization<br />problem. There are many different methods made in aim to solve the<br />optimization problems.&nbsp; The investigation made here is motivated by<br />the fact that the methods which converge fast are necessary.<br />The main goal is the systematization of some known results and also<br />theoretical and numerical analysis of the possibilities to int roduce<br />some parameters within gradient methods.<br />Firstly, the minimization problem is considered, where the objective<br />function is a convex, multivar iable function. This problem is solved<br />here without the calculation of Hessian, and such solution is very<br />important, for example, when the&nbsp; big dimension systems are solved,<br />and also for solving optimization problems with unknown values of<br />the objective function and its gradient. Partially, this investigation is<br />motivated by the existence of problems where the objective function<br />is the result of simulations.<br />Numerical results, presented in&nbsp; Chapter&nbsp; 6, show that the introduction<br />of a parameter is useful, i.e., such introduction results by the<br />acceleration of the known optimization method.<br />Further, one new hybrid conjugate gradient method is presented, in<br />which the conjugate gradient parameter is a convex combination of<br />two known conjugate gradient parameters.<br />In the first chapter, there is motivation and also the basic co ncepts<br />which are necessary for the other chapters.<br />The second chapter contains the survey of some first order and<br />second order gradient methods.<br />The fourth chapter contains the survey of some basic concepts and<br />results corresponding to conjugate gradient methods.<br />The first, the second and the fourth&nbsp; chapters are here to help in<br />considering of some known results, and the original results are<br />presented in the chapters 3,5 and 6.<br />In the third chapter, a modification of one unco nstrained optimization<br />method is presented, in which the randomly chosen multiplicative<br />parameter is used. Also, the linear convergence of such modification<br />is proved.<br />The fifth chapter contains the original results, corresponding to<br />conjugate gradient methods. Namely, one new hybrid conjugate<br />gradient method is presented, and this&nbsp; method is the convex<br />combination of two known conjugate gradient methods.<br />The sixth chapter consists of the numerical results, performed on a set<br />of test functions, corresponding to methods in the chapters 3 and 5.<br />Implementation of all considered algorithms is made in Mathematica.<br />The comparison criterion is CPU time.</p> / <p>The problem under consideration is an unconstrained optimization<br />problem. There are many different methods made in aim to solve the<br />optimization problems.&nbsp; The investigation made here is motivated by<br />the fact that the methods which converge fast are necessary.<br />The main goal is the systematization of some known results and also<br />theoretical and numerical analysis of the possibilities to int roduce<br />some parameters within gradient methods.<br />Firstly, the minimization problem is considered, where the objective<br />function is a convex, multivar iable function. This problem is solved<br />here without the calculation of Hessian, and such solution is very<br />important, for example, when the&nbsp; big dimension systems are solved,<br />and also for solving optimization problems with unknown values of<br />the objective function and its gradient. Partially, this investigation is<br />motivated by the existence of problems where the objective function<br />is the result of simulations.<br />Numerical results, presented in&nbsp; Chapter&nbsp; 6, show that the introduction<br />of a parameter is useful, i.e., such introduction results by the<br />acceleration of the known optimization method.<br />Further, one new hybrid conjugate gradient method is presented, in<br />which the conjugate gradient parameter is a convex combination of<br />two known conjugate gradient parameters.<br />In the first chapter, there is motivation and also the basic co ncepts<br />which are necessary for the other chapters.<br />Key&nbsp; Words Documentation&nbsp; 97<br />The second chapter contains the survey of some first order and<br />second order gradient methods.<br />The fourth chapter contains the survey of some basic concepts and<br />results corresponding to conjugate gradient methods.<br />The first, the second and the fourth&nbsp; chapters are here to help in<br />considering of some known results, and the original results are<br />presented in the chapters 3,5 and 6.<br />In the third chapter, a modification of one unco nstrained optimization<br />method is presented, in which the randomly chosen multiplicative<br />parameter is used. Also, the linear convergence of such modification<br />is proved.<br />The fifth chapter contains the original results, corresponding to<br />conjugate gradient methods. Namely, one new hybrid conjugate<br />gradient method is presented, and this&nbsp; method is the convex<br />combination of two known conjugate gradient methods.<br />The sixth chapter consists of the numerical results, performed on a set<br />of test functions, corresponding to methods in the chapters 3 and 5.<br />Implementation of all considered algorithms is made in Mathematica.<br />The comparison criterion is CPU time</p>

Cette thèse est consacrée dans un premier temps à l’élaboration d’un modèle variationnel dedébruitage d’ordre deux, faisant intervenir l’espace BV 2 des fonctions à hessien borné. Nous nous inspirons ici directement du célèbre modèle de Rudin, Osher et Fatemi (ROF), remplaçant la minimisation de la variation totale de la fonction par la minimisation de la variation totale seconde, c’est à dire la variation totale de ses dérivées. Le but est ici d’obtenir un modèle aussi performant que le modèle ROF, permettant de plus de résoudre le problème de l’effet staircasing que celui-ci engendre. Le modèle que nous étudions ici semble efficace, entraînant toutefois l’apparition d’un léger effet de flou. C’est afin de réduire cet effet que nous introduisons finalement un modèle mixte, permettant d’obtenir des solutions à la fois non constantes par morceaux et sans effet de flou au niveau des détails. Dans une seconde partie, nous nous intéressons au problème d’extraction de texture. Un modèle reconnu comme étant l’un des plus performants est le modèle T V -L1, qui consiste simplement à remplacer dans le modèle ROF la norme L2 du terme d’attache aux données par la norme L1. Nous proposons ici une méthode originale permettant de résoudre ce problème utilisant des méthodes de Lagrangien augmenté. Pour les mêmes raisons que dans le cas du débruitage, nous introduisons également le modèle T V 2-L1, consistant encore une fois à remplacer la variation totale par la variation totale seconde. Un modèle d’extraction de texture mixte est enfin très brièvement introduit. Ce manuscrit est ponctué d’un vaste chapitre dédié aux tests numériques. / This thesis is devoted in a first part to the elaboration of a second order variational modelfor image denoising, using the BV 2 space of bounded hessian functions. We here take a leaf out of the well known Rudin, Osher and Fatemi (ROF) model, where we replace the minimization of the total variation of the function with the minimization of the second order total variation of the function, that is to say the total variation of its partial derivatives. The goal is to get a competitive model with no staircasing effect that generates the ROF model anymore. The model we study seems to be efficient, but generates a blurry effect. In order to deal with it, we introduce a mixed model that permits to get solutions with no staircasing and without blurry effect on details. In a second part, we take an interset to the texture extraction problem. A model known as one of the most efficient is the T V -L1 model. It just consits in replacing the L2 norm of the fitting data term with the L1 norm.We propose here an original way to solve this problem by the use of augmented Lagrangian methods. For the same reason than for the denoising case, we also take an interest to the T V 2-L1 model, replacing again the total variation of the function by the second order total variation. A mixed model for texture extraction is finally briefly introduced. This manuscript ends with a huge chapter of numerical tests.

<p>The focus of this thesis is on numerical algorithms for efficient solution of QTL analysis problem in genetics.</p><p>Firstly, we consider QTL mapping problems where a standard least-squares model is used for computing the model fit. We develop optimization methods for the local problems in a hybrid global-local optimization scheme for determining the optimal set of QTL locations. Here, the local problems have constant bound constraints and may be non-convex and/or flat in one or more directions. We propose an enhanced quasi-Newton method and also implement several schemes for constrained optimization. The algorithms are adopted to the QTL optimization problems. We show that it is possible to use the new schemes to solve problems with up to 6 QTLs efficiently and accurately, and that the work is reduced with up to two orders magnitude compared to using only global optimization.</p><p>Secondly, we study numerical methods for QTL mapping where variance component estimation and a REML model is used. This results in a non-linear optimization problem for computing the model fit in each set of QTL locations. Here, we compare different optimization schemes and adopt them for the specifics of the problem. The results show that our version of the active set method is efficient and robust, which is not the case for methods used earlier. We also study the matrix operations performed inside the optimization loop, and develop more efficient algorithms for the REML computations. We develop a scheme for reducing the number of objective function evaluations, and we accelerate the computations of the derivatives of the log-likelihood by introducing an efficient scheme for computing the inverse of the variance-covariance matrix and other components of the derivatives of the log-likelihood.</p>

The focus of this thesis is on numerical algorithms for efficient solution of QTL analysis problem in genetics. Firstly, we consider QTL mapping problems where a standard least-squares model is used for computing the model fit. We develop optimization methods for the local problems in a hybrid global-local optimization scheme for determining the optimal set of QTL locations. Here, the local problems have constant bound constraints and may be non-convex and/or flat in one or more directions. We propose an enhanced quasi-Newton method and also implement several schemes for constrained optimization. The algorithms are adopted to the QTL optimization problems. We show that it is possible to use the new schemes to solve problems with up to 6 QTLs efficiently and accurately, and that the work is reduced with up to two orders magnitude compared to using only global optimization. Secondly, we study numerical methods for QTL mapping where variance component estimation and a REML model is used. This results in a non-linear optimization problem for computing the model fit in each set of QTL locations. Here, we compare different optimization schemes and adopt them for the specifics of the problem. The results show that our version of the active set method is efficient and robust, which is not the case for methods used earlier. We also study the matrix operations performed inside the optimization loop, and develop more efficient algorithms for the REML computations. We develop a scheme for reducing the number of objective function evaluations, and we accelerate the computations of the derivatives of the log-likelihood by introducing an efficient scheme for computing the inverse of the variance-covariance matrix and other components of the derivatives of the log-likelihood.