Submitted by Cristiane Chim (cristiane.chim@ucpel.edu.br) on 2017-02-14T11:02:21Z No. of bitstreams: 1 JulianoRB.pdf: 1568199 bytes, checksum: d4d56c5973ce16deee6885df8b5cc27f (MD5) / Made available in DSpace on 2017-02-14T11:02:21Z (GMT). No. of bitstreams: 1 JulianoRB.pdf: 1568199 bytes, checksum: d4d56c5973ce16deee6885df8b5cc27f (MD5) Previous issue date: 2016-12-14 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - CAPES# / #2075167498588264571# / #600 / Acoustic echo is a phenomenon caused by the reflection of sound when it is perceived by its emission source with some delay. It appears in full duplex communication systems in a detrimental manner to the interaction between the speakers, reducing the intelligibility of exchanged information. Because it is a very unpleasant effect, acoustic echo attenuation or canceling procedures, if possible, become necessary procedures to mitigate such phenomenon. Several techniques have been reported in the literature in order to cancel acoustic echo, which includes the application of structures based on adaptive filtering. Typically, strategies where the adaptive filtering is used, it is assumed that the path traveled by the acoustic echo can be modeled as a linear time-invariant system. However, because the elements that form such a path, such as amplifiers and speakers, present non-linear characteristics, linear models fail to adequately represent the whole path traveled by the acoustic echo. In recent approaches it has been assumed that the echo response should obey non-linear and non-stationary models, whereas in acoustic environments the signal to be modeled can be modified by the movement of people and objects. Thus, the appropriate response representation model becomes essential for satisfactory adaptive system’s performance on echo identification and cancellation. This work aims to propose a new technique in the nonlinear and time variant estimation from the application of the Kalman filtering theory, which presents a good answer in the identification of non-stationary systems, in a Hilbert Space defined by Reproductive Kernel, where linearity can be treated linearly, using as an example of application the cancellation of acoustic echo / O eco acústico é um fenômeno causado pela reflexão do som quando este é percebido por sua fonte emissora com algum atraso, agindo, nos sistemas de comunicação full duplex de forma prejudicial na interação entre os interlocutores, ocasionando a redução da inteligibilidade das informações trocadas. Dessa forma, por se tratar de um efeito bastante desagradável, faz-se necessária a aplicação de procedimentos cuja finalidade seja atenuar, ou se possível, eliminar tal fenômeno. Diversas técnicas têm sido apresentadas na literatura com o objetivo de cancelar o eco acústico, onde se destaca a aplicação de estruturas baseadas na filtragem adaptativa. Normalmente, em estratégias onde a filtragem adaptativa é utilizada, assume-se que o caminho percorrido pelo eco acústico pode ser modelado como um sistema linear e invariante no tempo. Entretanto, em virtude do conjunto de elementos que formam tal caminho, tais como os amplificadores e os alto-falantes, apresentarem características não lineares, os modelos lineares não conseguem representar de forma adequada todo o caminho percorrido pelo eco acústico. Considerando que em ambientes acústicos o sinal a ser modelado pode ser modificado pela circulação de pessoas e movimentação de objetos, em recentes abordagens, têm-se assumido que a resposta do eco deve obedecer a modelos não lineares e não estacionários. Dessa forma, o modelo adequado de representação da resposta se torna fundamental para que o sistema adaptativo utilizado na identificação e cancelamento do eco tenha um desempenho satisfatório. Este trabalho propõe uma nova técnica na estimação não linear e variante no tempo a partir da aplicação da teoria de filtragem de Kalman, que apresenta uma boa resposta na identificação de sistemas não estacionários, em um Espaço de Hilbert definido por Kernel Reprodutivo, onde a não linearidade poderá ser tratada de forma linear, utilizando como exemplo de aplicação o cancelamento do eco acústico. Palavras

Today’s computer systems develop towards less energy consumption while keeping high performance. These are contradictory requirement and pose a great challenge. A good example of an application were this is used is the smartphone. The constraints are on long battery time while getting high performance required by future 2D/3D applications. A solution to this is heterogeneous systems that have components that are specialized in different tasks and can execute them fast with low energy consumption. These could be specialized i.e. encoding/decoding, encryption/decryption, image processing or communication. At the apartment of Computer Architecture and Parallel Processing Laboratory (CAPPL) at New Jersey Institute of Technology (NJIT) a vector co-processor has been developed. The Vector co-processor has the unusual feature of being able to receive instructions from multiple hosts (scalar cores). In addition to this a test system with a couple of scalar processors using the vector processor has been developed. This thesis describes this processor and its test system. It also shows the development of math applications involving matrix operations. This results in the conclusions of the vector co-processing saving substantial amount of energy while speeding up the execution of the applications. In addition to this the thesis will describe an extension of the vector co-processor design that makes it possible to monitor the throughput of instructions and data in the processor.

Structured representation of input physical patterns as a set of local features has been useful for a veriety of robotics and human computer interaction (HCI) applications. It enables a stable understanding of the variable inputs. However, this representation does not fit the conventional machine learning algorithms and distance metrics because they assume vector inputs. To learn from input patterns with variable structure is thus challenging. To address this problem, I propose a general and systematic method to design distance metrics between structured inputs that can be used in conventional learning algorithms. Based on the observation of the stability in the geometric distributions of local features over the physical patterns across similar inputs, this is done combining the local similarities and the conformity of the geometric relationship between local features. The produced distance metrics, called “parametric kernels”, are positive semi-definite and require almost linear time to compute. To demonstrate the general applicability and the efficacy of this approach, I designed and applied parametric kernels to handwritten character recognition, on-line face recognition, and object detection from laser range finder sensor data. Parametric kernels achieve recognition rates competitive to state-of-the-art approaches in these tasks. / text

A Linguagem RS é destinada a programação de núcleos reativos centralizados. Tais núcleos são responsáveis por toda a lógica de um sistema reativo, manipulando os sinais de entrada, realizando as reações e gerando os sinais de saída. Sendo sua idéia inicial tratar apenas processos centralizados, não houve a preocupação com a distribuição. Este trabalho tem como principal objetivo apresentar os aspectos introduzidos de uma nova versão para a Linguagem e para o Compilador RS, que possibilitam a execução de programas distribuídos. Além da possibilidade de execução de sistemas reativos distribuídos, foi acrescentado à Linguagem RS extensões já previstas na sua criação, como sinais inibidores, regras de exclusão mútua e concomitância, a possibilidade de disparo de mais de uma regra em um mesmo instante e a limpeza léxica do código fonte RS. As modificações incorporadas nesta nova versão da linguagem, foram efetivadas através de um novo compilador, chamado de Compilador RS 5.0. O protótipo implementado oferece a geração de três formatos de código: o formato padrão da linguagem RS (os autômatos e as regras correspondentes), códigos na linguagem C para a simulação dos autômatos (tanto para programas distribuídos quanto não-distribuídos) e arquivos no formato portável OC, que é um formato de código objeto padrão para as linguagens reativas. Para a distribuição e implementação da Linguagem RS foi necessária a criação de um novo núcleo de comunicação do MDX, que é responsável pela comunicação dos autômatos RSD. Este núcleo é dividido em três partes. A primeira trata da definição de um modelo formal com as mudanças necessárias para que a linguagem RS consiga trabalhar de forma distribuída, a segunda mostra o projeto do novo núcleo MDX e a terceira apresenta a implementação em C e MDX dos autômatos gerados pelo Compilador RS 5.0. Por fim, exemplos de aplicação desta nova linguagem são apresentados, onde podem ser vistos a importância e o acréscimo proporcionado por este trabalho tanto à linguagem RS quanto à programação de sistemas reativos síncronos. / The RS language is intended to the programming of centralized reactive kernels. Such kernels are responsible for the logic of a reactive system, manipulating the input signals, carrying through the reactions and generating the output signals. Being its initial idea to treat only centered processes, it did not have the concern with the distribution. The main objective of this work is to describe the process of creation of a new version for the Language and Compiler RS, that make possible the execution of distributed programs. Beyond the possibility of execution distributed reactive systems, it was added to RS language foreseen extensions already in its creation, as inhibiting signals, rules of manual exclusion and concurrence, the possibility of detonation of more than a rule in one exactly instant and the lexical cleanness of the RS code source. The modifications incorporated in this new version of the language, had been accomplished through a new compiler, called Compiler RS 5.0. The implemented archetype offers the generation of three formats of code: the standard format of RS language (the corresponding automatons and rules), codes in the language C for the simulation of the automatons and archives in OC portable format, that is a object format code standard for the reactive languages. For the distribution and implementation of Language RS was necessary the creation of a new kernel of communication of the MDX, that is responsible for the communication of RSD automatons. It is divided in three parts. The first one deals with the definition of a formal model that defines the necessary changes so that RS language obtains to work of distributed form, the second shows the design of new MDX kernel and third presents the implementation in C and MDX of the automatons generated for Compiler RS 5.0. Finally, examples of application of this new language are presented, where the importance and the proportionate upgrade for this work to RS language how to the programming of synchronous reactive systems can in such a way be seen.

A Linguagem RS é destinada a programação de núcleos reativos centralizados. Tais núcleos são responsáveis por toda a lógica de um sistema reativo, manipulando os sinais de entrada, realizando as reações e gerando os sinais de saída. Sendo sua idéia inicial tratar apenas processos centralizados, não houve a preocupação com a distribuição. Este trabalho tem como principal objetivo apresentar os aspectos introduzidos de uma nova versão para a Linguagem e para o Compilador RS, que possibilitam a execução de programas distribuídos. Além da possibilidade de execução de sistemas reativos distribuídos, foi acrescentado à Linguagem RS extensões já previstas na sua criação, como sinais inibidores, regras de exclusão mútua e concomitância, a possibilidade de disparo de mais de uma regra em um mesmo instante e a limpeza léxica do código fonte RS. As modificações incorporadas nesta nova versão da linguagem, foram efetivadas através de um novo compilador, chamado de Compilador RS 5.0. O protótipo implementado oferece a geração de três formatos de código: o formato padrão da linguagem RS (os autômatos e as regras correspondentes), códigos na linguagem C para a simulação dos autômatos (tanto para programas distribuídos quanto não-distribuídos) e arquivos no formato portável OC, que é um formato de código objeto padrão para as linguagens reativas. Para a distribuição e implementação da Linguagem RS foi necessária a criação de um novo núcleo de comunicação do MDX, que é responsável pela comunicação dos autômatos RSD. Este núcleo é dividido em três partes. A primeira trata da definição de um modelo formal com as mudanças necessárias para que a linguagem RS consiga trabalhar de forma distribuída, a segunda mostra o projeto do novo núcleo MDX e a terceira apresenta a implementação em C e MDX dos autômatos gerados pelo Compilador RS 5.0. Por fim, exemplos de aplicação desta nova linguagem são apresentados, onde podem ser vistos a importância e o acréscimo proporcionado por este trabalho tanto à linguagem RS quanto à programação de sistemas reativos síncronos. / The RS language is intended to the programming of centralized reactive kernels. Such kernels are responsible for the logic of a reactive system, manipulating the input signals, carrying through the reactions and generating the output signals. Being its initial idea to treat only centered processes, it did not have the concern with the distribution. The main objective of this work is to describe the process of creation of a new version for the Language and Compiler RS, that make possible the execution of distributed programs. Beyond the possibility of execution distributed reactive systems, it was added to RS language foreseen extensions already in its creation, as inhibiting signals, rules of manual exclusion and concurrence, the possibility of detonation of more than a rule in one exactly instant and the lexical cleanness of the RS code source. The modifications incorporated in this new version of the language, had been accomplished through a new compiler, called Compiler RS 5.0. The implemented archetype offers the generation of three formats of code: the standard format of RS language (the corresponding automatons and rules), codes in the language C for the simulation of the automatons and archives in OC portable format, that is a object format code standard for the reactive languages. For the distribution and implementation of Language RS was necessary the creation of a new kernel of communication of the MDX, that is responsible for the communication of RSD automatons. It is divided in three parts. The first one deals with the definition of a formal model that defines the necessary changes so that RS language obtains to work of distributed form, the second shows the design of new MDX kernel and third presents the implementation in C and MDX of the automatons generated for Compiler RS 5.0. Finally, examples of application of this new language are presented, where the importance and the proportionate upgrade for this work to RS language how to the programming of synchronous reactive systems can in such a way be seen.

Made available in DSpace on 2016-12-08T16:44:51Z (GMT). No. of bitstreams: 1 PGPV14MA180.pdf: 350816 bytes, checksum: 21cfa9750f5bc20f9040d6522a471d04 (MD5) Previous issue date: 2014-12-11 / In southern Brazil, Fusarium Head Blight (FHB) caused by the fungus Fusarium graminearum is severe disease because the fungus is abundant in winter and summer crop residues, which helps inoculation in susceptible maize hybrids when have excessive periods of rain after pollination phase. This research aimed to determine the best time to spray fungicide during maize silking to control FHB. The experiment was conducted in field at the experimental area of Agroveterinarias Sciences Center at the Santa Catarina State University (CAV/UDESC) in Lages, Santa Catarina. Were seeded simple hybrids P32R48 H and Maximus TL TG Viptera during two periods in the growing seasons of 2012/13 and 2013/14. After five days of stigma emergence, the ears were artificially inoculated with 1 x 105 macroconídios.mL-1 concentration of F. graminearum injecting 2 ml of inoculum solution in stigma-style channel. Inoculations happen in 03/08/2013, 03/22/2013, 12/26/2013 and 01/19/2014. The fungicide was a mixture of azoxystrobin + cyproconazole (300 mL c.p/ha) + carbendazim (1 L c.p/ha) + nimbus mineral oil (0.5% v.v/ha). The treats were six fungicide spray times: 144, 96 and 48 hours before inoculation; 48, 96 and 144 hours after inoculation and a control, only inoculated without fungicide application. The experiment design was randomized block with four replications. The fungicide was applied with a manual CO2 gas sprayer. The variables evaluated were severity of FHB in the ears (SGE), percentage of damaged kernels (PGA), incidence of F. graminearum in the grains (IFgG) and grain yield (RG). Data were subjected to variance analysis, Tukey´s test and simple correlation analysis. The hybrid P32R48H was more susceptible than Maximus TL TG Viptera to FHB. Sprays 48 hours before and after inoculation were effective to reduce SGE in approximately 50% and 41%; to reduce PGA by 33% and 50%; to reduce IFgG by 27%; and to increase RG by 9% and 7%, respectively when compared to control. There is a negative linear correlation between SGE, PGA, IFgG and RG variables. The yield and grain quality were improved by fungicide spray between the 3rd and 7th day period after stigma issue / Na região Sul do Brasil, a podridão de espiga de giberela, causada pelo fungo Fusarium graminearum, é uma doença preocupante, pois o fungo é abundante nos resíduos culturais das gramíneas de inverno e de verão, o que facilita a sua inoculação nas espigas de híbridos suscetíveis em safras com excesso de chuva após a fase da polinização. Este trabalho de pesquisa teve como objetivo definir o melhor momento para pulverização de fungicida durante o estádio do espigamento do milho para controlar a podridão de espiga de giberela. O experimento foi conduzido à campo, na área experimental do Centro de Ciências Agroveterinárias da Universidade do Estado de Santa Catarina (CAV/UDESC) no município de Lages, Santa Catarina. Foram semeados os híbridos simples P32R48 H e Maximus TL TG Viptera, em duas épocas nas safras agrícolas de 2012/13 e 2013/14. Após cinco dias da emissão dos estigmas, as espigas foram inoculadas artificialmente com uma concentração de 1 x 105 macroconídios.mL-1 de F. graminearum injetando 2 mL da solução de inóculo no canal do estilo-estigma. As inoculações foram realizadas nos dias 08/03/2013, 22/03/2013, 26/12/2013 e 19/01/2014. O fungicida utilizado foi a mistura de azoxistrobina + ciproconazole (300 mL p.c/ha) + carbendazim (1 L p.c/ha) + oléo mineral Nimbus (0,5% v.v./ha). Os tratamentos constituiram-se em seis momentos de pulverização do fungicida: 144, 96 e 48 horas antes da inoculação; 48, 96 e 144 horas depois da inoculação e uma testemunha, somente inoculado sem aplicação de fungicida. A distribuição dos tratamentos seguiu o modelo do delineamento em blocos casualizados, com quatro repetições. O fungicida foi aplicado com o auxílio de pulverizador costal com pressão produzida por gás CO2. As variáveis avaliadas foram severidade da giberela nas espigas (SGE), porcentagem de grãos ardidos (PGA), incidência de F. graminearum nos grãos (IFgG) e rendimento de grãos (RG). Os dados obtidos foram submetidos à análise de variância, ao teste de Tukey e a análise de correlação simples. O híbrido P32R48H apresentou-se mais suscetível a podridão de giberela. As pulverizações 48 horas antes e depois da inoculação foram eficientes em reduzir a SGE em aproximadamente 50% e 41%; em reduzir a PGA em 33% e 50%; em reduzir a IFgG em 27%; e em aumentar o RG em 9% e 7%, respectivamente quando comparados ao tratamento sem fungicida. Existe associação entre as variáveis SGE, PGA, IFgG e RG. Pode-se aumentar o rendimento e melhorar a qualidade de grãos com a aplicação de fungicida no período do 3º ao 7º dia após a emissão dos estigmas

Cette thèse s'inscrit dans le contexte de diagnostic assisté par ordinateur pour la localisation subcellulaire des protéines dans les images microscopiques. L'objectif est la conception et le développement d'un système de classification automatique permettant d'identifier le compartiment cellulaire dans lequel une protéine d'intérêt exerce son activité biologique. Afin de surmonter les difficultés rencontrées pour discerner les compartiments cellulaires présents dans les images microscopiques, les systèmes décrits dans la littérature proposent d'extraire plusieurs descripteurs associés à une combinaison de classifieurs. Dans cette thèse, nous proposons un schéma de classification différent répondant mieux aux besoins de généricité et de flexibilité pour traiter différentes bases d'images.Dans le but de fournir une caractérisation riche des images microscopiques, nous proposons un nouveau système de représentation permettant d'englober de multiples descripteurs visuels identifiés dans les différentes approches d'extraction de caractéristiques : locale, fréquentielle, globale et par région. Nous formulons ensuite le problème de fusion et de sélection des caractéristiques sous forme d'un problème de sélection de noyaux. Basé sur l'apprentissage de noyaux multiples (MKL), les tâches de sélection et de fusion de caractéristiques sont considérées simultanément. Les expériences effectuées montrent que la plateforme de classification proposée est à la fois plus simple, plus générique et souvent plus performante que les autres approches de la littérature. Dans le but d'approfondir notre étude sur l'apprentissage de noyaux multiples, nous définissons un nouveau formalisme d'apprentissage MKL réalisé en deux étapes. Cette contribution consiste à proposer trois termes régularisant liés à la résolution du problème d'apprentissage des poids associés à une combinaison linéaire de noyaux, problème reformulé en un problème de classification à vaste marge dans l'espace des couples. Le premier terme régularisant proposé assure une sélection parcimonieuse des noyaux. Les deux autres termes ont été conçus afin de tenir compte de la similarité entre les noyaux via une métrique basée sur la corrélation. Les différentes expérimentations réalisées montrent que le formalisme proposé permet d'obtenir des résultats de même ordre que les méthodes de référence, mais offrant l'avantage d'utiliser moins de fonctions noyaux. / This thesis arises in the context of computer aided analysis for subcellular protein localization in microscopic images. The aim is the establishment of an automatic classification system allowing to identify the cellular compartment in which a protein of interest exerts its biological activity. In order to overcome the difficulties in attempting to discern the cellular compartments in microscopic images, the existing state-of-art systems use several descriptors to train an ensemble of classifiers. In this thesis, we propose a different classification scheme wich better cope with the requirement of genericity and flexibility to treat various image datasets. Aiming to provide an efficient image characterization of microscopic images, a new feature system combining local, frequency-domain, global, and region-based features is proposed. Then, we formulate the problem of heterogeneous feature fusion as a kernel selection problem. Using multiple kernel learning, the problems of optimal feature sets selection and classifier training are simultaneously resolved. The proposed combination scheme leads to a simple and a generic framework capable of providing a high performance for microscopy image classification. Extensive experiments were carried out using widely-used and best known datasets. When compared with the state-of-the-art systems, our framework is more generic and outperforms other classification systems. To further expand our study on multiple kernel learning, we introduce a new formalism for learning with multiple kernels performed in two steps. This contribution consists in proposing three regularized terms with in the minimization of kernels weights problem, formulated as a classification problem using Separators with Vast Margin on the space of pairs of data. The first term ensures that kernels selection leads to a sparse representation. While the second and the third terms introduce the concept of kernels similarity by using a correlation measure. Experiments on various biomedical image datasets show a promising performance of our method compared to states of art methods.

Quantum computing is an emerging field with potential applications in machine learning. This research project aimed to compare the performance of a quantum kernel to that of a classical kernel in machine learning binary classification tasks. Two Support Vector Machines, a popular classification model, was implemented for the respective Variational Quantum kernel and the classical Radial Basis Function kernel and tested on the same sets of artificial quantum-based testing data. The results show that the quantum kernel significantly outperformed the classical kernel for the specific type of data and parameters used in the study. The findings suggest that quantum kernels have the potential to improve machine learning performance for certain types of problems, such as search engines and self-driving vehicles. Further research is, however, needed to confirm their utility in general situations. / Kvantberäkning är ett växande forskningsområde med möjliga tillämpningar inom maskininlärning. I detta forskningsprojekt jämfördes prestandan hos en klassisk kärna med den hos en kvantkärna i binär klassificering för maskininlärninguppgifter, och implikationerna av resultaten diskuterades. Genom att implementera två stödvektormaskiner, en populär klassifikationsmodell, för respektive variabel kvantkärna och klassisk radiell basfunktionskärna kunde vi direkt testa båda kärnorna på samma uppsättning av artificiella kvant-baserad testdata. Resultaten visar på betydande prestandafördelar för kvantkärnan jämfört med den klassiska kärnan när det gäller denna specifika typ av data och de parametrar som användes i vår studie. Vi drar slutsatsen att kvantkärnor inom maskininlärning har potential att överträffa klassiska kärnor, men att mer forskning krävs för att fastställa om detta har någon nytta i allmänna situationer. Om det finns betydande prestandafördelar kan det finnas många tillämpningar, till exempel för sökmotorer och självkörande fordon.

We state a sufficient condition for a fusion system to be saturated. This is then used to investigate localities with kernels: that is, localities that are (in a particular way) extensions of groups by localities. As an application of these results, we define and study certain products in fusion systems and localities, thus giving a new method to construct saturated subsystems of fusion systems.

Le but de ce travail est de présenter des résultats à propos des noyaux sauvages étales. Soit $p$ un nombre premier. Les noyaux sauvages étales d'un corps de nombres $F$ (qui sont dénotés par $WK^{ét}_{2i}(F)$ avec $i\in \mathbb{Z}$) sont des généralisations cohomologiques de la $p$-partie du noyau sauvage classique $WK_{2}(F)$, qui est le sous-groupe de $K_2(F)$ constitué par les symboles qui sont triviaux pour tout symbole de Hilbert local. Ces noyaux sauvages étales sont des $\mathbb{Z}_p$-modules et l'on sait qu'ils sont finis lorsque $i\geq 1$ (et même, suivant les conventions, si $i=0$) : on conjecture en plus qu'ils soient toujours finis (conjecture de Schneider). Dans la suite, on va supposer que cette conjecture est satisfaite. On va s'intéresser en particulier à deux problèmes. Le premier, qui est étudié dans les Chapitres 2 et 3, est la déterminations des structures de groupe qui sont réalisables comme noyaux sauvages étales. En d'autres termes, si l'on se donne un corps de nombres $F$, un $p$-groupe abélien fini $X$ et un nombre entier $i\in\mathbb{Z}$, on peut se demander s'il existe une extension finie $E/F$ telle que $WK^{ét}_{2i}(E)\cong X$. Une question semblable a été étudiée pour les $p$-groupes des classes et il y a un relation précise entre les $p$-groupes des classes et les noyaux sauvages étales. Par conséquent, on peut espérer traduire les résultats classiques dans le contexte des noyaux sauvages étales. Peut-être est-il intéressant de donner ici une courte récapitulation sur le problème de réalisation classique pour les $p$-groupes des classes. Essentiellement, deux techniques sont utilisées. D'un coté, pour un corps de nombres $F$ fixé, l'on étudie la $p$-tour des corps des classes de Hilbert de $F$ : Yahagi a montré que cette tour est infinie si et seulement s'il n'y a pas d'extensions finies $E/F$ dont le $p$-groupe des classes soit trivial. De plus, si la tour est finie, alors toute structure de $p$-groupe abélien apparaît comme $p$-groupe des classes pour quelque extension finie $E/F$. De l'autre coté, une fois que l'on sait que pour un corps de nombres $F$ fixé, il existe une extension finie dont le $p$-groupe de classes est trivial, alors on peut se servir de la théorie du corps des classes et de la théorie des genres pour trouver, pour n'importe quel $p$-groupe abélien fini $X$, une extension finie $E/F$ telle que le $p$-groupe des classes de $E$ est isomorphe à $X$. En effet, la traduction du résultat de Yahagi dans le contexte des noyaux sauvages étales n'est pas tout à fait immédiate : la relation entre le groupe des classes et le noyau sauvage étale d'un corps de nombres $F$ s'écrit dans le langage de $\Gamma$-modules, où $\Gamma$ est le groupe de Galois sur $F$ de la $\mathbb{Z}_p$-extension cyclotomique de $F(\mu_p)$. La façon la plus naturelle pour s'approcher du problème est donc de considérer le problème de réalisabilité pour les modules d'Iwasawa. Ce problème a été étudié (parmi d'autres auteurs) par Ozaki : il a montré que pour tout $\Lambda$-module fini $X$, il existe un corps de nombres $k$ tel que le module d'Iwasawa de $k$ (c'est à dire la limite projective des $p$-groupes des classes le long de la tour cyclotomique) est isomorphe à $X$. Les techniques utilisées sont inspirées à celles de Yahagi et en fait elles s'appuient d'une façon fondamentale du fait que $p$ ne divise pas le nombre des classes de $\mathbb{Q}$. Pour obtenir la traduction de ce résultat en termes de noyaux sauvages étales il faut considérer plutôt $\mathbb{Q}(\mu_p)$ -plus précisément un sous-corps convenable de $\mathbb{Q}(\mu_p)$. Bien entendu, le nombre des classes de ce sous-corps n'est plus premier avec $p$ (du moment que $p$ peut être irrégulier). D'autre part, si $p$ est régulier, la preuve d'Ozaki peut être adaptée (comme l'on montre dans le Chapitre 2). / The aim of the present work is to prove some results about étale wild kernels. Let $p$ be an odd prime. Etale wild kernels of a number field $F$ (which are denoted $WK^{ét}_{2i}(F)$ for $i\in \mathbb{Z}$) are cohomological generalizations of the $p$-part of the classical wild kernel $WK_{2}(F)$, which is the subgroup of $K_2(F)$ made up by symbols which are trivial for any local Hilbert symbol. Etale wild kernels are $\mathbb{Z}_p$-modules which are known to be finite if $i\geq1$ (and even if $i=0$, depending on the chosen convention): actually they are conjectured to be always finite (the Schneider conjecture). In the following we will suppose that this is always the case. Two problems are studied in detail. The first, which is analyzed in Chapter 2 and Chapter 3, is to determine which group structures are realizable for étale wild kernels. In other words, given a number field $F$, a finite abelian $p$-group $X$ and $i\in \mathbb{Z}$, one can ask if there exists a finite extension $E/F$ such that $WK^{ét}_{2i}(E)\cong X$. A similar problem has been studied for $p$-class groups and there are precise relations between the $p$-class group and étale wild kernels. Therefore one may expect to translate results from $p$-class groups to étale wild kernels. It is maybe useful to give here a short account on the classical realizability problem for $p$-class groups. Essentially two kind of techniques are used. On the one hand, for a fixed number field $F$, one studies the Hilbert $p$-class field tower of $F$: it has been shown by Yahagi that the Hilbert $p$-class tower of $F$ is infinite if and only if there is no finite extension $E/F$ whose $p$-class group is trivial. Furthermore, if the Hilbert $p$-class tower of $F$ is finite, then every finite abelian $p$-group structure appears as $p$-class group of some finite extension $E/F$. On the other hand, once we know that for a fixed number field $F$ there exists a finite extension whose $p$-class group is trivial, then class field theory and genus theory are used to exhibit, for any finite abelian $p$-group $X$, a finite extension $E/F$ such that the $p$-class group of $E$ is isomorphic to $X$. Actually, the translation of Yahagi's result in terms of étale wild kernels is not immediate: the relation between the class groups and étale wild kernels of a number field $F$ is expressed in terms of $\Gamma$-modules structures, where $\Gamma$ is the Galois group over $F$ of the cyclotomic $\mathbb{Z}_p$-extension of $F(\mu_p)$. The most natural way to approach the problem is then to consider the realizability problem for Iwasawa modules. This problem is studied (among many others) by Ozaki: he proved that for any finite $\Lambda$-module $X$, there exists a number field $k$ such that the Iwasawa module of $k$ (i.e. the projective limit of $p$-class groups along the cyclotomic $\mathbb{Z}_p$-extension) is isomorphic to $X$. The techniques used are inspired to those by Yahagi and actually Ozaki makes fundamental use of the fact that $p$ does not divide the class number of $\mathbb{Q}$. To get the translation of this result in terms of étale wild kernels one has to consider $\mathbb{Q}(\mu_p)$ -more precisely a suitable subfield of $\mathbb{Q}(\mu_p)$ depending on $i$- instead of $\mathbb{Q}$. Here the problem is that the class number of this suitable subfield is no more coprime with $p$ (as $p$ may be irregular). If this is not the case anyway, the proof of Ozaki can be adapted as it is shown in Chapter 2.