Distibuted denial of service (DDoS) attacks are getting more and more common in society as the number of devices connected to the Internet is increasing. To reduce the impact of such attacks it is important to detect them as soon as possible. Many papers have investigated how well different machine learning algorithms can detect DDoS attacks. However, most papers are focusing on supervised learning algorithms which require a lot of labeled data, which is hard to find. This thesis compares two unsupervised learning algorithms, an autoencoder and principal component analysis (PCA), in how well they detect DDoS attacks. The models are implemented in the Python libraries Keras, Tensorflow and scikit-learn. They are then trained and tested with data that has its origin in the CICDDOS2019 dataset. There are normal data and nine different types of DDoS attacks in the used dataset. The models are compared by computing the Receiver Operating Characteristic (ROC) curve and its Area Under the Curve (AUC) score, and the F1 score of the models. For both measures the mean value of the results of all attack types are used. The computations show that the autoencoder perform better than PCA with respect to both the mean AUC score (0.981 compared to 0.967) and the mean F1 score (0.987 compared to 0.978). The thesis goes on to discussing why the autoencoder performs better than PCA and, finally draws conclusions based on the insights of the analysis.

The primary objective of this dissertation is to extend the classical Principal Components Analysis (PCA), aiming to reduce the dimensionality of a large number of Normal interrelated variables, in two directions. The first is to go beyond the static (contemporaneous or synchronous) covariance matrix among these interrelated variables to include certain forms of temporal (over time) dependence. The second direction takes the form of extending the PCA model beyond the Normal multivariate distribution to the Elliptically Symmetric family of distributions, which includes the Normal, the Student's t, the Laplace and the Pearson type II distributions as special cases. The result of these extensions is called the Generalized principal component analysis (GPCA). The GPCA is illustrated using both Monte Carlo simulations as well as an empirical study, in an attempt to demonstrate the enhanced reliability of these more general factor models in the context of out-of-sample forecasting. The empirical study examines the predictive capacity of the GPCA method in the context of Exchange Rate Forecasting, showing how the GPCA method dominates forecasts based on existing standard methods, including the random walk models, with or without including macroeconomic fundamentals. / Ph. D. / Factor models are employed to capture the hidden factors behind the movement among a set of variables. It uses the variation and co-variation between these variables to construct a fewer latent variables that can explain the variation in the data in hand. The principal component analysis (PCA) is the most popular among these factor models. I have developed new Factor models that are employed to reduce the dimensionality of a large set of data by extracting a small number of independent/latent factors which represent a large proportion of the variability in the particular data set. These factor models, called the generalized principal component analysis (GPCA), are extensions of the classical principal component analysis (PCA), which can account for both contemporaneous and temporal dependence based on non-Gaussian multivariate distributions. Using Monte Carlo simulations along with an empirical study, I demonstrate the enhanced reliability of my methodology in the context of out-of-sample forecasting. In the empirical study, I examine the predictability power of the GPCA method in the context of “Exchange Rate Forecasting”. I find that the GPCA method dominates forecasts based on existing standard methods as well as random walk models, with or without including macroeconomic fundamentals.

Thèse réalisée en cotutelle avec l'université Montpellier2 dans le laboratoire de pharmacologie moléculaire de Jean-Philippe Pin à l'institut de génomique fonctionnelle (IGF), Montpellier, France. / Les récepteurs couplés aux protéines G (RCPG) sont une famille très diversifiée de protéines membranaires capables de répondre à un grand nombre de signaux chimiques tels que des photons, des molécules odorantes, ou des hormones. En plus de cette diversité, l’étude des RCPG montre que des associations protéiques spécifiques multiplient les possibilités de signalisation de chacun de ces récepteurs. En permettant d’atténuer, de potentialiser, ou de générer une nouvelle voie de signalisation, l’association des RCPG en oligomères s’avère une importante source de diversité. L’utilisation du transfert d’énergie de résonance de bioluminescence (BRET) qui permet de détecter les interactions protéiques a révélé de nombreuses associations de RCPG. Durant cette thèse, des outils ont été développés pour combiner efficacement le BRET à des essais de complémentation de protéines (PCA) dans le but de savoir si l’oligomérisation des RCPG pouvait impliquer plus de deux récepteurs. Les résultats présentés montrent que les récepteurs de chimiokines CXCR4 et CCR2 forment des homo et hétéro tétramères, et que l’activation d’un dimère CCR2 peut moduler la conformation d’un dimère CXCR4 par un changement conformationnel trans-récepteur. La coopérativité négative de liaison de ligand qui a été démontrée auparavant entre CXCR4 et CCR2 dans des lymphocytes T CD4+ exprimant les récepteurs de manière endogène confirme la validité biologique de cette interaction. Les données présentées suggèrent également que ces complexes peuvent engager les effecteurs Gαi et β-arrestine2, indiquant qu’ils représentent la forme fonctionnelle de ces récepteurs. Enfin, nous avons pu confirmer que chaque récepteur de l’hétérodimère CXCR4-CCR2 est impliqué dans l’engagement des effecteurs lors de l’activation de CCR2. Un autre niveau de complexité dans la signalisation des RCPG est atteint par leur capacité à coupler de multiples protéines G. La liaison du facteur dérivé des cellules stromales (SDF-1) au récepteur CXCR4 permet la migration des lymphocytes T par une voie de signalisation dépendante de la protéine Gαi. Nous avons pu démontrer en revanche que la migration des cellules de cancer du sein était initiée par un couplage de CXCR4 à la voie Gα13-Rho pour former des métastases dans des organes distants. Enfin, un dernier niveau de régulation des RCPG a été abordé par l’étude de la phosphorylation de CXCR4 suite à son activation, qui permet la désensibilisation du récepteur et l’engagement de voies de signalisation dépendantes de la β-arrestine. Il apparaît que la désensibilisation de la voie du calcium serait médiée par la phosphorylation de CXCR4 par les kinases des RCPG (GRK) GRK2 et GRK6 et le recrutement de β- arrestine2, alors GRK3, GRK6 et la β-arrestine1 potentialiseraient l’activation des kinases régulées par les signaux extracellulaires (ERK1/2). Nous suggérons également que c’est la phosphorylation de l’extrémité C-terminale de CXCR4 qui permettrait son association avec la β-arrestine. / G protein-coupled receptors (GPCRs) are a diverse family of membrane proteins capable of responding to a large number of extracellular stimuli including photons, odorant molecules and hormones. In addition to this diversity, it has been shown that GPCRs form specific protein:protein interactions, multiplying the signalling possibilities of each of these receptors. With the ability to diminish, to potentiate or even generate new signalling pathways, the oligomeric association of GPCRs plays an important role in generating this diversity. The use of bioluminescence resonance energy transfer (BRET), which allows the detection of interactions among proteins, has revealed numerous associations between GPCRs. During this thesis, tools have been developed that effectively combine BRET with protein complementation assays (PCA) with the goal of determining if interactions between GPCRs could involve more than two receptors. The results show that the chemokine receptors CXCR4 and CCR2 form both homo and hetero tetramers, and that the activation of a dimer of CCR2 can modulate the conformation of a CXCR4 dimer through a transreceptor conformational change. Negative cooperativity of ligand binding has previously been demonstrated between CXCR4 and CCR2 in CD4+ T lymphocytes endogenously expressing the receptors, confirming the biological validity of this interaction. The data presented also suggests that these complexes can engage the effector proteins Gαi and β- arrestin 2, indicating that they represent a functional form of the receptors. Furthermore, we have confirmed that each receptor of the CXCR4-CCR2 heterodimer is implicated in the engagement of effectors during the activation of CCR2. An additional level of complexity in GPCR-promoted signaling exists in their capacity to couple of multiple G proteins. Binding of stromal cell-derived factor-1 (SDF-1) to CXCR4 is known to promote T lymphocyte migration through a Gαi-dependent signalling pathway. In addition to this mechanism, we have demonstrated that breast cancer cell migration can initiated by a coupling of CXCR4 to the Gα13-Rho pathway, leading to the formation of metastases in distant organs. Finally, a novel level of GPCR regulation was revealed through the study of CXCR4 phosphorylation following its activation, which leads to the desensitization of the receptor and the engagement of β-arrestin-dependent signalling pathways. It appears that the desensitization of calcium signalling is mediated through the phosphorylation of CXCR4 by the GPCR kinases (GRKs) GRK2 and GRK6 and the recruitment of β-arrestin 2, whereas GRK3, GRK6 and β-arrestin 1 potentiate the activation of extracellular regulated kinase (ERK1/2). We also propose that the phosphorylation of the far C-terminal tail of CXCR4 is required for the interaction between the receptor and β-arrestin.

Dans un marché en permanente progression et pour répondre aux besoins des brasseurs en malt de qualité, la maîtrise du procédé de maltage est indispensable. La qualité du malt est fortement dépendante des conditions opératoires, en particulier des conditions de trempe, mais également de la qualité de la matière première : l'orge. Dans cette étude, nous avons établi des modèles polynomiaux qui mettent en relation les conditions opératoires et la qualité du malt. Ces modèles ont été couplés à nos algorithmes génétiques et nous ont permis de déterminer les conditions optimales de maltage, soit pour atteindre une qualité ciblée de malt (friabilité), soit pour permettre un maltage à faible teneur en eau (pour réduire la consommation en eau et maîtriser les coûts environnementaux de production) tout en conservant une qualité acceptable de malt. Cependant, la variabilité de la matière première est un facteur limitant de notre approche. Les modèles établis sont en effet très sensibles à l'espèce d'orge (printemps, hiver) ou encore à la variété d'orge utilisée. Les modèles sont surtout très dépendants de l'année de récolte. Les variations observées sur les propriétés d'une année de récolte à une autre sont mal caractérisées et ne sont donc pas intégrées dans nos modèles. Elles empêchent ainsi de capitaliser l'information expérimentale au cours du temps. Certaines propriétés structurelles de l'orge (porosité, dureté) ont été envisagées comme nouveaux facteurs pour mieux caractériser la matière première mais ils n'ont pas permis d'expliquer les variations observés en malterie.Afin de caractériser la matière première, 394 échantillons d'orge issus de 3 années de récolte différentes 2009-2010-2011 ont été analysés par spectroscopie MIR. Les analyses ACP ont confirmé l'effet notable des années de récolte, des espèces, des variétés voire des lieux de culture sur les propriétés de l'orge. Une régression PLS a permis, pour certaines années et pour certaines espèces, de prédire les teneurs en protéines et en béta-glucanes de l'orge à partir des spectres MIR. Cependant, ces résultats, pourtant prometteurs, se heurtent toujours à la variabilité. Ces nouveaux modèles PLS peuvent toutefois être exploités pour mettre en place des stratégies de pilotage du procédé de maltage à partir de mesures spectroscopiques MIR / In a continuously growing market and in order to meet the needs of Brewers in high quality malt, control of the malting process is a great challenge. Malt quality is highly dependent on the malting process operating conditions, especially on the steeping conditions, but also the quality of the raw material: barley. In this study, we established polynomial models that relate the operating conditions and the malt quality. These models have been coupled with our genetic algorithms to determine the optimal steeping conditions, either to obtain a targeted quality of malt (friability), or to allow a malting at low water content while maintaining acceptable quality of malt (to reduce water consumption and control the environmental costs of malt production). However, the variability of the raw material is a limiting factor for our approach. Established models are very sensitive to the species (spring and winter barley) or to the barley variety. The models are especially highly dependent on the crop year. Variations on the properties of a crop from one to another year are poorly characterized and are not incorporated in our models. They thus prevent us to capitalize experimental information over time. Some structural properties of barley (porosity, hardness) were considered as new factors to better characterize barley but they did not explain the observed variations.To characterize barley, 394 samples from 3 years of different crops 2009-2010-2011 were analysed by MIR spectroscopy. ACP analyses have confirmed the significant effect of the crop-years, species, varieties and sometimes of places of harvest on the properties of barley. A PLS regression allowed, for some years and for some species, to predict content of protein and beta-glucans of barley using MIR spectra. These results thus still face product variability, however, these new PLS models are very promising and could be exploited to implement control strategies in malting process using MIR spectroscopic measurements

Les approches de surveillance de l’intégrité des structures ont été proposées pour permettre un contrôle continu de l’état des structures en intégrant à celle-ci des capteurs intelligents. En effet, ce contrôle continu doit être effectué pour s’assurer du bon fonctionnement de celles-ci car la présence d’un défaut dans la structure peut aboutir à un accident catastrophique. Cependant, la variation des conditions environnementales et opérationnelles (CEO) dans lesquelles la structure évolue, impacte sévèrement les signaux collectés ce qui induit parfois une mauvaise interprétation de la présence du défaut dans la structure. Dans ce travail de thèse, l’application des méthodes d’apprentissage statistiques classiques a été envisagée dans le cas des structures tubulaires. Ici, les effets des paramètres de mesures sur la robustesse de ces méthodes ont été investiguées. Ensuite, deux approches ont été proposées pour remédier aux effets des CEO. La première approche suppose que la base de données des signaux de référence est suffisamment riche en variation des CEO. Dans ce cas, une estimation parcimonieuse du signal mesuré est calculée. Puis, l’erreur d’estimation est utilisée comme indicateur de défaut. Tandis que la deuxième approche est utilisée dans le cas où la base de données des signaux des références contient une variation limitée des CEO mais on suppose que celles-ci varient lentement. Dans ce cas, une mise à jour du modèle de l’état sain est effectuée en appliquant l’analyse en composante principale (PCA) par fenêtre mobile. Dans les deux approches, la localisation du défaut a été assurée en utilisant une fenêtre glissante sur le signal provenant de l’état endommagé. / To ensure better working conditions of civil and engineering structures, inspections must be made on a regular basis. However, these inspections could be labor-intensive and cost-consuming. In this context, structural health monitoring (SHM) systems using permanently attached transducers were proposed to ensure continuous damage diagnostic of these structures. In SHM, damage detection is generally based on comparison between the healthy state signals and the current signals. Nevertheless, the environmental and operational conditions will have an effect on the healthy state signals. If these effects are not taken into account they would result in false indication of damage (false alarm). In this thesis, classical machine learning methods used for damage detection have been applied in the case of pipelines. The effects of some measurements parameters on the robustness of these methods have been investigated. Afterthat, two approaches were proposed for damage diagnostic depending on the database of reference signals. If this database contains large variation of these EOCs, a sparse estimation of the current signal is calculated. Then, the estimation error is used as an indication of the presence of damage. Otherwise, if this database is acquired at limited range of EOCs, moving window PCA can be applied to update the model of the healthy state provided that the EOCs show slow and continuous variation. In both approaches, damage localization was ensured using a sliding window over the damaged pipe signal.

A qualidade microbiológica de medicamentos é fundamental para garantir sua eficácia e segurança. Os métodos convencionais para identificação microbiana em produtos não estéreis são amplamente utilizados, entretanto são demorados e trabalhosos. O objetivo deste trabalho é desenvolver método microbiológico rápido (MMR) para a identificação de contaminantes em produtos farmacêuticos utilizando a espectrofotometria de infravermelho com transformada de Fourier com reflectância total atenuada (FTIR-ATR). Análise de componentes principais (PCA) e análise de discriminantes (LDA) foram utilizadas para obter um modelo de predição com a capacidade de diferenciar o crescimento de oriundo de contaminação por Bacillus subtilis (ATCC 6633), Candida albicans (ATCC 10231), Enterococcus faecium (ATCC 8459), Escherichia coli (ATCC 8739), Micrococcus luteus (ATCC 10240), Pseudomonas aeruginosa (ATCC 9027), Salmonella Typhimurium (ATCC 14028), Staphylococcus aureus (ATCC 6538) e Staphylococcus epidermidis (ATCC 12228). Os espectros de FTIR-ATR forneceram informações quanto à composição de proteínas, DNA/RNA, lipídeos e carboidratos provenientes do crescimento microbiano. As identificações microbianas fornecidas pelo modelo PCA/LDA baseado no método FTIR-ATR foram compatíveis com aquelas obtidas pelos métodos microbiológicos convencionais. O método de identificação microbiana rápida por FTIR-ATR foi validado quanto à sensibilidade (93,5%), especificidade (83,3%) e limite de detecção (17-23 UFC/mL de amostra). Portanto, o MMR proposto neste trabalho pode ser usado para fornecer uma identificação rápida de contaminantes microbianos em produtos farmacêuticos. / Microbiological quality of pharmaceuticals is fundamental in ensuring efficacy and safety of medicines. Conventional methods for microbial identification in non-sterile drugs are widely used, however are time-consuming and laborious. The aim of this paper was to develop a rapid microbiological method (RMM) for identification of contaminants in pharmaceutical products using Fourier transform infrared with attenuated total reflectance spectrometry (FTIR-ATR). Principal components analysis (PCA) and linear discriminant analysis (LDA) were used to obtain a predictive model with capable to distinguish Bacillus subtilis (ATCC 6633), Candida albicans (ATCC 10231), Enterococcus faecium (ATCC 8459), Escherichia coli (ATCC 8739), Micrococcus luteus (ATCC 10240), Pseudomonas aeruginosa (ATCC 9027), Salmonella Typhimurium (ATCC 14028), Staphylococcus aureus (ATCC 6538), and Staphylococcus epidermidis (ATCC 12228) microbial growth. FTIR-ATR spectra provide information of protein, DNA/RNA, lipids, and carbohydrates constitution of microbial growth. Microbial identification provided by PCA/LDA based on FTIR-ATR method were compatible to those obtained using conventional microbiological methods. FTIR-ATR method for rapid identification of microbial contaminants in pharmaceutical products was validated by assessing the sensitivity (93.5%), specificity (83.3%), and limit of detection (17-23 CFU/mL of sample). Therefore, the RMM proposed in this work may be used to provide a rapid identification of microbial contaminants in pharmaceutical products.

O paradigma intruso-residente vem sendo intensamente empregado em estudos para avaliar a memória de reconhecimento social em roedores. Tipicamente, ratos adultos, denominados residentes, são expostos a dois encontros sucessivos, de 5 minutos cada, com um mesmo rato juvenil ou com ratos juvenis diferentes, denominados intrusos; o intervalo de tempo entre encontros é 30 minutos. A quantidade de comportamentos sociais do residente (no segundo encontro) em relação a um intruso familiar é substancialmente menor do que o observado no primeiro encontro, o que não ocorre quando o segundo encontro envolve um juvenil novo; esse resultado caracteriza a memoria de reconhecimento social. Se o intervalo de tempo entre os encontros é aumentado para 60 minutos, a redução da investigação social do intruso familiar por parte do residente desaparece, levando à conclusão de que a memória de reconhecimento social seria um mecanismo para retenção temporária de informações. O objetivo central do presente trabalho foi contribuir para o entendimento da memória de reconhecimento social em ratos. Foram realizados três experimentos. No primeiro experimento avaliou-se se a expressão de comportamentos sociais e também da memória de reconhecimento social estão sujeitos à modulação temporal. No segundo experimento avaliou-se em que extensão o aumento do tempo de exposição ao intruso durante o primeiro encontro resulta num aumento da duração da memoria de reconhecimento social. No terceiro experimento avaliou-se se um procedimento de rotina na maioria dos laboratorios, o transporte dos animais da sala de experimentos para o biotério, interfere na memória de reconhecimento social, quando realizado 0,5 ou 6 horas após o primeiro encontro. Os resultados mostraram que (1) a expressão de comportamentos sociais e a memória de reconhecimento social estão sujeitos à modulação temporal, sendo mais intensos quando os testes são realizados na fase inativa (Capítulo 2), de modo que este fator deve ser levado em consideração quando do planejamento de experimentos envolvendo sociabilidade, (2) o aumento da duração do primeiro encontro para 2 horas ou mais revelou uma memória de reconhecimento social que dura pelo menos 24 horas (Capítulo 3), permitindo questionar que se trate de um dispositivo de curta duração, e (3) o transporte dos animais para o biotério 0,5 horas, mas não 6 horas, depois do primeiro encontro, prejudica a memória de reconhecimento social (Capítulo 4), indicando que se deve estar atento às rotinas laboratoriais pois as mesmas podem interferir no desempenho dos animais em testes de memória. Em associação com essas relevantes observações experimentais, foram propostas estratégias de análise dos dados gerados com esse tipo de experimentação e também discussões conceituais sobre a caracterização da memória de reconhecimento social, que contribuem marcadamente para essa área de estudos. / The intruder-resident paradigm has been extensively employed in studies of social recognition memory in rodents. Typically, adult rats, named residents, are exposed to two 5-min successive encounters with the same juvenile intruder or with two different juveniles; the time interval between the encounters is 30 min. The amount of social behaviors exhibited by the resident rats towards the same intruder juvenile in the second encounter is substantially smaller when compared to both that seen in the first encounter and that seen towards a different juvenile; these results characterize social recognition memory. When the time interval between encounters is increased to 60 min, that reduction of the investigation towards the familiar juvenile intruder vanishes, which is seen as evidence that social recognition memory corresponds to a short-term memory mechanism. The aim of this study was to contribute for our understanding of social recognition memory in rats. Three experiments were run. The first experiment evaluated to which extent both social behaviors and social recognition memory are influenced by temporal phase effects. The second experiment evaluated to which extent the increase in the duration of the first encounter renders social recognition memory longer. The third experiment evaluated to which extent the transportation of the resident rats from the experimental room to the animal facilities either 0.5 or 6 hours after the first encounter, interferes with social recognition memory. The results showed that (1) the expression of social behaviors and of the social recognition memory are modulated temporal phase effects, being stronger when animals are tested in their inactive phase (Chapter 2); thus, this aspect has to be considered in studies on sociability, (2) the increase of the first encounter duration for 2 hours or longer renders social recognition memory to last at least 24 hours (Chapter 3); this allows to question that social recognition memory corresponds to a short-term memory mechanism, and (3) transportation of the resident rats to the animal facilities 0.5, but not 6 hours, after the end of the first encounter disrupts social recognition memory (Chapter 4), indicating that one has to be cautious about usual laboratory routines, because they may interfere with performance of the memory tasks when executed a short time after training the animals.Associated with these relevant experimental observations, these studies allowed proposing novel strategies for data analysis and discussing conceptual issues about the characterization of social recognition memory that give a substantial contribution for this area.

A esquistossomose é uma das principais causas de morbidade em países Tropicais e Subtropicais, gerando graves consequências socioeconômicas. Atualmente, os fármacos disponíveis para o tratamento da desta doença são praziquantel e oxamniquina, porém relatos de baixa susceptibilidade do parasita a esses medicamentos sugerem a necessidade de novas estratégias terapêuticas para o tratamento da doença. Todavia, existe pouco interesse da indústria farmacêutica no desenvolvimento de fármacos contra doenças tropicais e negligenciadas, entre as quais se encontra a esquistossomose. Devido a estes fatores, o presente trabalho teve por objetivo geral utilizar ferramentas computacionais para identificar inibidores da SmDHFR candidatos a novos fármacos. Avaliou-se as características exclusivas para a proteína de S. mansoni por meio de uma análise das sequências FASTA em comparação com a DHFR de outros organismos. A fim de garantir a ação seletiva dessas moléculas frente a enzima do parasita, os campos moleculares de interação seletivos para SmDHFR foram calculados e empregados na construção do modelo farmacofórico, o qual foi utilizado na triagem virtual de inibidores de SmDHFR. Os estudos computacionais realizados nos permitiram a seleção de 20 moléculas com uma boa complementariedade com o modelo farmacofórico gerado e com potencial para serem inibidores de SmDHFR. / Schistosomiasis is one of morbidity\'s main causes in tropical and subtropical countries, which leads to serious socioeconomic consequences. Praziquantel and oxamniquina are the drugs currently available for treating this disease, but reports points that the parasite has been resistant to both drugs, which suggests the need for new therapeutic strategies for the treatment of this disease. However, there is little interest in the pharmaceutical industry in developing drugs against neglected tropical diseases, including schistosomiasis. Due to these factors, the present work has the general objective to use computational tools to identify SmDHFR inhibitors which could be good candidates for developing new drugs. Evaluation of the exclusive characteristics of the S. mansoni protein were performed by FASTA sequence analyses in comparison to DHFR from other organisms. In order to guarantee the selective action of these molecules against the parasite enzyme, the molecular interaction fields selective for SmDHFR were calculated and used in the construction of the pharmacophoric model, which was further used in the virtual screening of SmDHFR inhibitors. Computational studies were performed and those led us to 20 molecules with a good complementarity with the pharmacophoric model that was previously generated and with potential to be SmDHFR inhibitors.

O Reservatório do Rio Grande faz parte do Complexo Billings, apesar de ser separado do Corpo Central pela barragem Anchieta. Esse reservatório está localizado na Região Metropolitana de São Paulo e é aproveitado para usos múltiplos, incluindo abastecimento público, recreação, lazer e corpo receptor de dejetos industriais e domésticos. No presente trabalho, foi estudada a coluna d’água de 6 pontos de coleta ao longo desse reservatório, em duas excursões a campo, uma em março e outra em outubro de 2004. Foram feitas medições de pH, temperatura, condutividade elétrica, oxigênio dissolvido (OD), transparência da coluna d’água, material em suspensão, sólidos totais, clorofila a, feoftina e compostos nitrogenados. Os dados foram tratados estatisticamente por meio de análise multivariada não paramétrica tipo PCA. Durante o mês de março, houve estratificação térmica, de pH e de condutividade elétrica. No mês de outubro houve maiores concentrações de clorofila a e menor transparência do disco de Secchi. O Índice de Estado Trófico, calculado a partir da clorofila a e a partir da profundidade do disco de Secchi, revelaram um ambiente meso a eutrófico. Foram registrados teores de NH4+, OD e clorofila a referentes a classes de uso menos nobres, de acordo com a Resolução CONAMA n° 357 de 17/03/2005. A análise estatística mostrou que os dois primeiros eixos explicam 55% da variância dos dados. O nitrogênio total e o pH são as variáveis que mais contribuem para compor os eixos 1 e 2, respectivamente, refletindo a importância do estado trófico para a caracterização do reservatório e a carga de esgoto que o corpo d’água recebe. Os pontos de amostragem se segregaram de acordo com a data de coleta, sem demonstrar padrão no sentido montante-jusante. O Reservatório do Rio Grande, Complexo Billings, está localizado na Região Metropolitana de São Paulo, e é um importante local de recreação e de onde é extraída água para o abastecimento de cidades da Grande São Paulo. Os sedimentos lacustres são um reservatório concentrado de contaminantes, dentre os quais se destacam os metais-traço. Esses metais podem ser incorporados à biota, acumulados e transferidos ao longo da rede trófica (bioacumulação e biomagnificação) ou permanecer biologicamente indisponíveis na forma de complexos insolúveis ou muito estáveis e refratários. Em ambiente anóxido, o sulfeto pode controlar a disponibilidade de íons metálicos livre, pois reage com os metais para formar um sal insolúvel potencialmente não biodisponível. Neste trabalho, foi utilizado método de extração ácida fraca a frio para a determinação dos Sulfetos Volatilizáveis por Acidificação (SVA) e dos íons metálicos extraídos simultaneamente (MES). SVA foi quantificado por espectrofotometria visível, enquanto Cd, Cu, Ni, Pb, Cr e Zn por ICP-AES. Matéria orgânica (MO) e a porcentagem de silte e argila, fases importantes na complexação de metais, também foram quantificadas. Os resultados foram comparados com valores-guias de qualidade de sedimento sugeridos pela USEPA (relação molar ∑ MES-SVA) e pelo Conselho Canadense do Ministério do Meio Ambiente (CCME) e também com Valores de Referência Regionais (VRR) estabelecidos para a Bacia do Alto-Tietê. Os resultados mostraram haver excesso de sulfeto em relação aos teores de metais, o que indica um baixo potencial tóxico. Houve um aumento do teor de SVA da ordem de 100 vezes do ponto 4 (montante) ao ponto 26 (jusante). No entanto, as concentrações de metais foram superiores ao VRR, especialmente Cu (102,7 vezes maior), Cd (46,2 vezes) e Pb (12,5 vezes). Também foram registradas concentrações maiores que as sugeridas pelo CCME (9,3; 8,3 e 2,9 vezes maior para Cu, Pb e Cd, respectivamente). À exceção do Zn, todos os metais analisados apresentaram incremento no sentido montante-jusante. A análise estatística tipo PCA demonstrou que a variação dos dados de metais possui correlação com a variação de MO, ao invés de SVA, o que sugere a importância da MO no controle da biodisponibilidade de metais. / The Rio Grande branch is a part of Billings Complex and is isolated from the central body by Anchieta Dam. This reservoir is located in São Paulo Metropolitan Region and has multiple uses, including water supply, recreation, leisure, receptor of domestic and industrial wastewater. At the present work, the water column was studied in 6 stations along the reservoir, in two different days (one in March and the other in October, 2004). We measured pH, temperature, electric conductivity, dissolved oxygen (DO), transparence of water column, suspended material, total solid, chlorophyll a, feoftin, nitrogenated substances. The data was treated by means of a PCA, a multivariate non-parametric analysis. During March, the water column was stratified because of temperature, pH and electric conductivity. During October, there were higher chlorophyll a concentrations and lower transparence. Trofic State Index was calculated from chlorophyll a and from Secchi disc data, and showed a meso to eutrofic environment. Some NH4+, DO and chlorophyll a concentrations are those of classes designated for less noble uses according to Brazilian law CONAMA n° 357 of 17/03/2005. The statistical analysis showed that the first two axis where responsible for 55% of the total data variance. Total nitrogen and pH were the variables that contributed most for the composition of axis 1 and 2. This reflects the importance of trofic state for the reservoir’s characteristics. The stations were segregated according to the sampling date, with no high-low direction pattern. Rio Grande reservoir, Billings Complex, is located in São Paulo Metropolitan Region. It is an important recreation place from where water is extracted in order to supply cities nearby. Lacustrine sediments are concentrated pool for chemicals, among which are trace metals. These metals may be taken up by organisms, accumulated by them and transferred along the food web (bioaccumulation and biomagnification) or may remain biologically unavailable as insoluble, stable and refractory complexes. In anoxic environments, sulfide may be responsible for metal free ion availability control, since it reacts with metal ions and forms an insoluble salt, potentially not bioavailable. At the present work, it was used a weak-acid extraction method for the acid volatile sulfide (AVS) and the simultaneously extracted metals (SEM) determination. Cd, Cu, Ni, Pb, Cr and Zn were quantified by means of an ICP-AES and AVS by means of visible spectrophotometer. Organic matter (OM) and clay-silt percentiles, important phases for metal complexation, were also measured. Results were compared to sediment quality guidelines suggested by USEPA (∑ MES-SVA molar relation) and by Canadian Council of Ministers of Environment (CCME) as well by Regional Reference Values (RRV) established for the Alto-Tietê basin. There was an increase of AVS concentration of 100 folds from station 4 (near the riverine region) to station 26 (near the dam). Results showed excess of sulfide compared with metals, what means a low toxic potential. On the other hand, metal concentrations were above RRV, specially Cu (102,7 times greater), Cd (46,2 times greater) and Pb (12,5 times greater). There were also concentrations greater than values suggested by CCME (9,3; 8,3 and 2,9 times greater for Cu, Pb and Cd, respectively). Excepted by Zn, all metals showed increasing patterns from upper to lower regions of the reservoir. PCA statistical analysis demonstrated that the metal data variation has a correlation with OM variation, instead with AVS, and suggests the importance of OM in controlling metal bioavailability in the studied system.

On suppose que des vecteurs de données de grande dimension arrivant en ligne sont des observations indépendantes d'un vecteur aléatoire. Dans le second chapitre, ce dernier, noté Z, est partitionné en deux vecteurs R et S et les observations sont supposées identiquement distribuées. On définit alors une méthode récursive d'estimation séquentielle des r premiers facteurs de l'ACP projetée de R par rapport à S. On étudie ensuite le cas particulier de l'analyse canonique, puis de l'analyse factorielle discriminante et enfin de l'analyse factorielle des correspondances. Dans chacun de ces cas, on définit plusieurs processus spécifiques à l'analyse envisagée. Dans le troisième chapitre, on suppose que l'espérance En du vecteur aléatoire Zn dont sont issues les observations varie dans le temps. On note Rn = Zn - En et on suppose que les vecteurs Rn forment un échantillon indépendant et identiquement distribué d'un vecteur aléatoire R. On définit plusieurs processus d'approximation stochastique pour estimer des vecteurs directeurs des axes principaux d'une analyse en composantes principales (ACP) partielle de R. On applique ensuite ce résultat au cas particulier de l'analyse canonique généralisée (ACG) partielle après avoir défini un processus d'approximation stochastique de type Robbins-Monro de l'inverse d'une matrice de covariance. Dans le quatrième chapitre, on considère le cas où à la fois l'espérance et la matrice de covariance de Zn varient dans le temps. On donne finalement des résultats de simulation dans le chapitre 5 / High dimensional data are supposed to be independent on-line observations of a random vector. In the second chapter, the latter is denoted by Z and sliced into two random vectors R et S and data are supposed to be identically distributed. A recursive method of sequential estimation of the factors of the projected PCA of R with respect to S is defined. Next, some particular cases are investigated : canonical correlation analysis, canonical discriminant analysis and canonical correspondence analysis ; in each case, several specific methods for the estimation of the factors are proposed. In the third chapter, data are observations of the random vector Zn whose expectation En varies with time. Let Rn = Zn - En be and suppose that the vectors Rn form an independent and identically distributed sample of a random vector R. Stochastic approximation processes are used to estimate on-line direction vectors of the principal axes of a partial principal components analysis (PCA) of ~Z. This is applied next to the particular case of a partial generalized canonical correlation analysis (gCCA) after defining a stochastic approximation process of the Robbins-Monro type to estimate recursively the inverse of a covariance matrix. In the fourth chapter, the case when both expectation and covariance matrix of Zn vary with time n is considered. Finally, simulation results are given in chapter 5