The dissertation deals with clustering algorithms and transforming regression prob-lems into classification problems. The main contributions of the dissertation are twofold; first, to improve (speed up) the clustering algorithms and second, to develop a strict learn-ing environment for solving regression problems as classification tasks by using support vector machines (SVMs). An extension to the most popular unsupervised clustering meth-od, k-means algorithm, is proposed, dubbed k-means2 (k-means squared) algorithm, appli-cable to ultra large datasets. The main idea is based on using a small portion of the dataset in the first stage of the clustering. Thus, the centers of such a smaller dataset are computed much faster than if computing the centers based on the whole dataset. These final centers of the first stage are naturally much closer to the locations of the final centers rendering a great reduction in the total computational cost. For large datasets the speed up in computa-tion exhibited a trend which is shown to be high and rising with the increase in the size of the dataset. The total transient time for the fast stage was found to depend largely on the portion of the dataset selected in the stage. For medium size datasets it has been shown that an 8-10% portion of data used in the fast stage is a reasonable choice. The centers of the 8-10% samples computed during the fast stage may oscillate towards the final centers' positions of the fast stage along the centers' movement path. The slow stage will start with the final centers of the fast phase and the paths of the centers in the second stage will be much shorter than the ones of a classic k-means algorithm. Additionally, the oscillations of the slow stage centers' trajectories along the path to the final centers' positions are also greatly minimized. In the second part of the dissertation, a novel approach of posing a solution of re-gression problems as the multiclass classification tasks within the common framework of kernel machines is proposed. Based on such an approach both the nonlinear (NL) regression problems and NL multiclass classification tasks will be solved as multiclass classification problems by using SVMs. The accuracy of an approximating classification (hyper)Surface (averaged over several benchmarking data sets used in this study) to the data points over a given high-dimensional input space created by a nonlinear multiclass classifier is slightly superior to the solution obtained by regression (hyper)Surface. In terms of the CPU time needed for training (i.e. for tuning the hyperparameters of the models), the nonlinear SVM classifier also shows significant advantages. Here, the comparisons between the solutions obtained by an SVM solving given regression problem as a classic SVM regressor and as the SVM classifier have been performed. In order to transform a regression problem into a classification task, four possible discretizations of a continuous output (target) vector y are introduced and compared. A very strict double (nested) cross-validation technique has been used for measuring the performances of regression and multiclass classification SVMs. In order to carry out fair comparisons, SVMs are used for solving both tasks - regression and multiclass classification. The readily available and most popular benchmarking SVM tool, LibSVM, was used in all experiments. The results in solving twelve benchmarking regression tasks shown here will present SVM regression and classification algorithms as strongly competing models where each approach shows merits for a specific class of high-dimensional function approximation problems.

Today's cloud landscape has evolved computing infrastructure into a dynamic, high utilization, service-oriented paradigm. This shift has enabled the commoditization of large-scale storage and distributed computation, allowing engineers to tackle previously untenable problems without large upfront investment. A key enabler of flexibility in the cloud is the ability to transfer running virtual machines across subnets or even datacenters using live migration. However, live migration can be a costly process, one that has the potential to interfere with other applications not involved with the migration. This work investigates storage interference through experimentation with real-world systems and well-established benchmarks. In order to address migration interference in general, a buffering technique is presented that offloads the migration's read, eliminating interference in the majority of scenarios.

Cette thèse a pour objectif de démontrer la richesse et la variété des dispositifs scéniques utilisés par Molière à Paris, à partir de 1658. Sur ce thème, il n’y eut guère jusqu’ici que des travaux peu convaincants basés sur l’étude iconographique des frontispices. Cette étude s’ouvre sur un rappel circonstancié du contexte technique, matériel et scénographique au XVIIe siècle avec notamment les descriptions précises des grandes innovations dues aux Italiens ; puis, à l’aide de notes, de manuscrits ou de contrats, les spécificités architecturales du théâtre du Palais-Royal, qui fut pour Molière le principal espace de création, sont détaillés. Enfin, l’ensemble des comédies du dramaturge est regroupé en deux grandes catégories : les comédies données à la Ville, comme L’École des femmes, Le Tartuffe ou L’Avare ; celles-ci n’ont pas bénéficié de commentaires qui puissent éclairer sur le dispositif scénique. Mais le décryptage minutieux du texte a pallié cette lacune ; il met en évidence un rapport quasi constant entre action et décoration, et confirme l’intérêt manifeste de l’auteur-comédien pour la mise en scène. La seconde catégorie concerne les comédies données pour la Cour, souvent appelées « comédies-ballets » ou comédies « mêlées de danse et de musique » ; celles-ci ont fait l’objet d’amples relations contemporaines. Bénéficiant de moyens financiers considérables, elles furent caractéristiques par leur débordement de luxe et d’invention. / This thesis aims to demonstrate the richness and variety of scenic devices used by Molière in Paris beginning in 1658. On this subject, there have been only a few unconvincing works to date, and these have been based on iconographic study and frontispieces. The thesis begins with an account of the detailed technical background, materials and stage design in the seventeenth century. It is particularly committed to locating and describing in detail the major innovations resulting from Italian influence. Through an examination of notes, manuscripts and unpublished theater contracts, it has sometimes even been possible to specify the architectural characteristics of the theater of the Palais-Royal, which was Molière's main creative space. The thesis then groups all of the comedies of the playwright into two broad categories. The first category, comedies played “à la Ville” (e.g., L’École des femmes, Tartuffe and L’Avare) has not benefited from commentaries that could clarify the settings that were used. Detailed textual analysis helps to overcome this problem; it highlights an almost constant ratio between action and decoration, and confirms the author-actor Molière’s obvious interest in staging. The second category concerns comedies played for the Court, often called "comedies-ballets" and comedies of "mingled music and dance," which have received more thorough contemporary commentary. These comedies benefited from considerable financial resources and are characterized by their overflowing luxury, scenic writing and invention.

At its core, a nanopore detector has a nanometer-scale biological membrane across which a voltage is applied. The voltage draws a DNA molecule into an á-hemolysin channel in the membrane. Consequently, a distinctive channel current blockade signal is created as the molecule flexes and interacts with the channel. This flexing of the molecule is characterized by different blockade levels in the channel current signal. Previous experiments have shown that a nanopore detector is sufficiently sensitive such that nearly identical DNA molecules were classified successfully using machine learning techniques such as Hidden Markov Models and Support Vector Machines in a channel current based signal analysis platform [4-9]. In this paper, methods for improving feature extraction are presented to improve both classification and to provide biologists and chemists with a better understanding of the physical properties of a given molecule.

An SVM-based clustering algorithm is introduced that clusters data with no a priori knowledge of input classes. The algorithm initializes by first running a binary SVM classifier against a data set with each vector in the set randomly labeled. Once this initialization step is complete, the SVM confidence parameters for classification on each of the training instances can be accessed. The lowest confidence data (e.g., the worst of the mislabeled data) then has its labels switched to the other class label. The SVM is then re-run on the data set (with partly re-labeled data). The repetition of the above process improves the separability until there is no misclassification. Variations on this type of clustering approach are shown.

This work is an investigation into reconstructing fragmented ASCII files based on content analysis motivated by a desire to demonstrate machine learning's applicability to Digital Forensics. Using a categorized corpus of Usenet, Bulletin Board Systems, and other assorted documents a series of experiments are conducted using machine learning techniques to train classifiers which are able to identify fragments belonging to the same original file. The primary machine learning method used is the Support Vector Machine with a variety of feature extractions to train from. Additional work is done in training committees of SVMs to boost the classification power over the individual SVMs, as well as the development of a method to tune SVM kernel parameters using a genetic algorithm. Attention is given to the applicability of Information Retrieval techniques to file fragments, as well as an analysis of textual artifacts which are not present in standard dictionaries.

Malware detection and analysis is a major part of computer security. There is an arm race between security experts and malware developers to develop various techniques to secure computer systems and to find ways to circumvent these security methods. In recent years process heap-based attacks have increased significantly. These attacks exploit the system under attack via the heap, typically by using a heap spraying attack. The main drawback with existing techniques is that they either consume too many resources or are complicated to implement. Our work in this thesis focuses on new methods which offloads process heap analysis for guest Virtual Machines (VM) to the privileged domain using Virtual Machine Introspection (VMI) in a Cloud environment. VMI provides us with a seamless, non-intrusive and invisible (to malwares) way of observing the memory and state of VMs without raising red flags for the malwares.

Cloud computing has brought many benefits to the IT industry, and could reduce the cost and facilitate the growth of businesses specially the startup companies which don’t have enough financial resources to build their own IT infrastructure. One of the main reason that companies hesitate to use cloud services is the security issues that the cloud computing technology has. This thesis at the beginning has an overview on the cloud computing concept and then reviews the cloud security vulnerabilities according to the cloud security alliance, then it describes the cloud denial of service and will focus on analyzing the Slow HTTP DOS attack and then will analyze the direct and indirect impact of these attacks on virtual machines. We decided to analyze the HTTP slow rate attacks because of the craftiness and covered characteristic also the catastrophic impact of the Slow HTTP attack whether it’s lunched on the cloud component or lunched from the cloud. There are some researches on the different way that a web server or web service can be protected against slow HTTP attacks, but there is a research gap about the impact of the attack on virtual environment or whether this attack has cross VM impact or not. This thesis investigates the impact of Slow HTTP attack on virtualization environment and will analyze the direct and indirect impact of these attack. For analyzing the Slow HTTP attacks, Slow headers, Slow body and Slow read are implemented using Slowhttptest and OWASP Switchblade software, and Wireshark is used to capture the traffic. For analyzing the impact of the attack, attacks are lunched on VirtualBox and the impact of the attack on the victim VM and neighbor VM is measured.

Le moteur électrique est l’un des organes principaux d’un véhicule électrique. Sa température, notamment celle des bobines, doit être réduite pour éviter toute dégradation. Le refroidissement par l’extérieur, comme avec une chemise d’eau dans le carter, apparait limité car les pertes générées dans les bobines doivent traverser des zones où la conduction thermique est très mauvaise. L’extraction des calories au cœur de la machine est préférable, mais les échanges thermiques avec l’air sont modérés. En application automobile, le moteur électrique est situé à proximité d’un circuit d’huile de lubrification. Le refroidissement par l'huile en contact direct avec les bobines est étudié.La thèse s’est d’abord attachée à la recherche bibliographique étendue sur les différentes solutions de refroidissement de moteur. Ensuite, les transferts thermiques à l’intérieur du moteur ont été modélisés par méthode nodale. A travers une étude de sensibilité, les principales améliorations thermiques passives ont été dégagées, puis les systèmes de refroidissement eux-mêmes ont été modélisés. Enfin, des essais ont été réalisés sur un banc spécialement conçu. Pour cette partie expérimentale, le refroidissement direct des bobines par circulation d’huile a été étudié en détail. Différents types d’injecteurs d’huile sur les têtes de bobine ont été testés dans diverses conditions de vitesse du rotor, température et débit d’huile.L’objectif de cette thèse est d’analyser l’ensemble des problématiques thermiques liées aux solutions de refroidissement à huile. Il s’agit d’une étude comparative de la performance des solutions à huile entre elles et avec celle d’un refroidissement à eau plus conventionnel. / Electric motor is one of the most important elements of an electric vehicle. Some elements, particularly the windings, can be affected by rising heat. External cooling, as water jacket in the case, appears to be limited because the losses generated in windings must pass through zones where conduction is very poor. Cooling in the core of the machine is preferable, but heat transfer with air is poor. Due to the presence of lubricating oil in the vicinity of the motor and the heat transfer enhancement that such a liquid provides, oil circulation on the windings has been considered.The research was first dedicated to an extensive bibliography on the different solutions of motor cooling. Then heat transfer within the motor was modelled by using the lumped system analysis. Thanks to a sensitivity analysis, the main parameters affecting temperature have been identified before cooling systems were modelled. Finally, tests were performed on a specially designed bench. Oil was introduced at each side of the machine to directly cool the stator coil end-windings. Several oil injection patterns were tested. The influence of the oil flow rate, rotation speed and oil temperature has been investigated.The objective of this PHD study is to analyse all the thermal issues related to the oil cooling systems. This is a comparative study of the performance of the oil cooling solutions. Comparison is also done with conventional water cooling.

As Máquinas Virtuais ganharam grande importância com o advento de processadores multi-core (na plataforma x86) e com o barateamento de componentes de hardware, como a memória. Por conta desse substancial aumento do poder computacional, surgiu o desafio de tirar proveito dos recursos ociosos encontrados nos ambientes corporativos, cada vez mais populados por equipamentos multi-core e com vários Gigabytes de memória. A virtualização, mesmo sendo um conceito já antigo, tornou-se novamente popular neste cenário, pois com ela foi possível utilizar melhor os recursos computacionais, agora abundantes. Este trabalho tem como principal foco estudar algumas das principais técnicas de gerência de recursos computacionais em ambientes virtualizados. Apesar de muitos dos conceitos aplicados nos projetos de Monitores de Máquinas Virtuais terem sido portados de Sistemas Operacionais convencionais com pouca, ou nenhuma, alteração; alguns dos recursos ainda são difíceis de virtualizar com eficiência devido a paradigmas herdados desses mesmos Sistemas Operacionais. Por fim, é apresentado o Memory Dispatcher (MD), um mecanismo de gerenciamento de memória, com o objetivo principal de distribuir a memória entre as Máquinas Virtuais de modo mais eficaz. Este mecanismo, implementado em C, foi testado no Monitor de Máquinas Virtuais Xen e apresentou ganhos de memória de até 70%. / Virtual Machines have gained great importance with advent of multi-core processors (on platform x86) and with low cost of hardware parts, like physical memory. Due to this computational power improvement a new challenge to take advantage of idle resources has been created. The virtualization technology, even being an old concept became popular in research centers and corporations. With this technology idle resources now can be exploited. This work has the objective to show the main techniques to manage computational resources in virtual environments. Although many of current concepts used in Virtual Machine Monitors project has been ported, with minimal changes, from conventional Operating Systems there are some resources that are difficult to virtualize with efficiency due to old paradigms still present in Operating Systems projects. Finally, the Memory Dispatcher (MD) is presented, a mechanism used to memory management. The main objective of MD is to improve the memory share among Virtual Machines. This mechanism was developed in C and it was tested in Xen Virtual Machine Monitor. The MD showed memory gains up to 70%.