Les espèces de poissons migrateurs (saumon atlantique, Salmo salar, en particulier) requièrent des conditions thermiques bien spécifiques. Ils sont très sensibles aux températures de l’eau et aux fortes variations estivales. Sur les trente dernières années, l’étude menée sur la Garonne (France) révèle une augmentation des températures estivales associée à un allongement de la durée des périodes chaudes. L’impact de cette modification du régime thermique sur la survie et la reproduction des espèces migratoires est également mis en évidence. Cette étude est menée sur un tronçon de Garonne, situé entre l’amont de Toulouse et l’amont de la retenue deMalause. Ce secteur est fortement touché par cette problématique avec en moyenne 2°C d’écart entre l’amont et l’aval et des températures supérieures à 25°C régulièrement atteintes. Le régime hydrologique de ce tronçon est fortement déficitaire (selon le SDAGE Adour-Garonne), la sensibilité au flux de surface est forte à cause de son lit large et peu profond, les pressions anthropiques sont importantes, ce sont autant de pistes dont l’impact sur le régime thermique est étudié. Une large collection de données hydrologiques et climatiques est exploitée afin de déterminer les processus en jeu dans l’évolution du régime thermique de ce tronçon de fleuve. Des études en tendances et corrélations et des modèles statistiques permettent de mettre en évidence d’une part la relation forte qui existe entre les températures de l’air et les températures de l’eau et d’autre part l’importance des faibles débits durant les périodes estivales. L’estimation des moyennes journalières de température de l’eau à Malause au moyen de modèles statistiques et déterministes donne de bons résultats pour les températures élevées ainsi que pour les franchissements de seuils liés aux conditions de migrations des amphilalins.Enfin un modèle numérique monodimensionnel de résolution de l’équation de transport thermique et des équations de St-Venant est développé. La physique du modèle tant au niveau hydraulique (prise en compte de fortes variabilités de pente, d’ouvrages, etc.) que thermique (apports latéraux, flux de surfaces, flux de conduction avec le lit) permet d’analyser l’évolution des différents flux qui participent au réchauffement du cours d’eau. Une évolution future à l’aide des sorties des modèles de l’IPCC est explorée et des méthodes éventuelles de restauration des conditions de températures favorables pour les espèces piscicoles sont analysées. / Fish species with strong thermal requirements (i.e. Atlantic salmon) are very sensitive to temperature evolution and particularly to large increases. An investigation conducted on the Garonne River (France) during the last three decades revealed global water warming along with an increase of the high temperature period duration. Large impact of this evolution on the survival and breeding of migrating fish species was also reported. Study was thus conducted on a specific reach of the Garonne River located between the immediate upstream of Toulouse and the upstream of the Malause dam. The issue of water temperature warming is particularly relevant on this reach, with an average increase of 2°C between upstream and downstream and temperatures above 25°C frequently reported. Potential causes are numerous: drastic low-flow regime (quoting SDAGE Adour-Garonne), impacts of surface fluxes that are important due to bed shape (wide and shallow), anthropogenic impacts, etc. Large amount of climatic and hydraulic data are used to make a clear determination of the processes involved in the thermal regime evolution of this reach. Trend and correlation analyses and use of statistical models indicate the strong relation between stream temperature and climate. Low flows also seem to be related to water temperatures during summer periods. Statistic and deterministic models give good results in estimating high daily mean water temperatures (RMSE ranging from 0.99°C to 1.22°C) and predicting water temperatures threshold crossings related to the migrating conditions of Atlantic salmon.Finally, a one-dimensional numerical model that solves both shallow water and thermal equations is developed. Both the formulation of the St-Venant equations (high variability in slope, gates …) and the phenomena taken into account in the water temperature model (lateral influx, surface fluxes, bed conduction …) allows studying the evolution of fluxes driving water temperature evolution. Future evolution of the water temperature at the 2050 horizon is also evaluated using IPCC models output and potential solutions to restore favorable stream temperatures conditions for fishes are analyzed.

A voz tem sido objeto de estudos em diferentes áreas da ciência. Nas últimas duas décadas os pesquisadores demonstraram a presença do caos na produção de voz. O objetivo deste trabalho é estabelecer padrões visuais de sinais de voz humana através da técnica não linear de reconstrução de espaço de fase e associá-los com suas correspondentes análises auditivo perceptiva e acústica. Foram analisados sinais de voz humana de sujeitos de ambos os gêneros, na faixa etária de 19 a 39 anos. Foram gravadas as vogais sustentadas /a/, /e/ e /i/ do português do Brasil, com uma taxa de amostragem de 22.050 Hz. Os sinais foram analisados a fim de obter medidas acústicas (Jitter, Shimmer e Coeficiente de Excesso). Foi utilizada a técnica de reconstrução de espaço de fase para descrever as características de dinâmica não linear dos sinais de voz, e para análise perceptivo auditiva foi utilizada a escala RASAT. Os resultados demonstram que métodos de dinâmica não linear como a reconstrução do espaço de fase, parece ser uma técnica apropriada para análise de sinais de voz, incorporando os componentes caótico e determinístico da voz humana. É sugerido que análise da dinâmica não linear não substitui as técnicas existentes, embora possa aperfeiçoar e complementar os métodos de análise disponíveis para os profissionais da saúde, como fonoaudiólogos e otorrinolaringólogos / Human voice has been the focus of study for different areas of science. Researches in the last two decades have demonstrated the existence of chaos in human voice production. The purpose of this work is to use nonlinear dynamics techniques in the analysis of normal voices from healthy subjects and correlate them to traditional acoustic parameters as well as perceptual analysis. Human voice signals from healthy subjects, both male and female, ranging in age from 19 to 39 years old were analyzed. Sustained vowel sounds /a/, /e/ and /i/, from brazilian Portuguese were recorded at a sampling rate of 22,050 Hz and analyzed in order to obtain acoustic measures (Jitter, Shimmer and coefficient of excess – EX). The phase space reconstruction technique was used to describe the nonlinear dynamic characteristics of voice signal samples. The results show, that non-linear dynamical method as phase space reconstruction seems to be a suitable technique for voice signals analysis, due to the chaotic component of the human voice. It is suggested, that non-linear dynamic analysis does not replace existing techniques instead, it may improve and complement the recent voice analysis methods available for health professionals, speech therapist and clinician

Two controller performances are assessed for generalization in the path following task of autonomously backing up a tractor-trailer. Starting from random locations and orientations, paths are generated to loading docks with arbitrary pose using Dubins Curves. The combination vehicles can be varied in wheelbase, hitch length, weight distributions, and tire cornering stiffness. The closed form calculation of the gains for the Linear Quadratic Regulator (LQR) rely heavily on having an accurate model of the plant. However, real-world applications cannot expect to have an updated model for each new trailer. Finding alternative robust controllers when the trailer model is changed was the motivation of this research. Reinforcement learning, with neural networks as their function approximators, can allow for generalized control from its learned experience that is characterized by a scalar reward value. The Linear Quadratic Regulator and the Deep Deterministic Policy Gradient (DDPG) are compared for robust control when the trailer is changed. This investigation quantifies the capabilities and limitations of both controllers in simulation using a kinematic model. The controllers are evaluated for generalization by altering the kinematic model trailer wheelbase, hitch length, and velocity from the nominal case. In order to close the gap from simulation and reality, the control methods are also assessed with sensor noise and various controller frequencies. The root mean squared and maximum errors from the path are used as metrics, including the number of times the controllers cause the vehicle to jackknife or reach the goal. Considering the runs where the LQR did not cause the trailer to jackknife, the LQR tended to have slightly better precision. DDPG, however, controlled the trailer successfully on the paths where the LQR jackknifed. Reinforcement learning was found to sacrifice a short term reward, such as precision, to maximize the future expected reward like reaching the loading dock. The reinforcement learning agent learned a policy that imposed nonlinear constraints such that it never jackknifed, even when it wasn't the trailer it trained on.

The diploma thesis deals with nonlinear dynamical systems with emphasis on typical phenomena like bifurcation or chaotic behavior. The basic theoretical knowledge is applied to analysis of selected (chaotic) models, namely, Lorenz, Rössler and Chen system. The practical part of the work is then focused on a numerical simulation to confirm the correctness of the theoretical results. In particular, an algorithm for calculating the largest Lyapunov exponent is created (under the MATLAB environment). It represents the main tool for indicating chaos in a system.

In Finance, the use of Automated Trading Systems (ATS) on markets is growing every year and the trades generated by an algorithm now account for most of orders that arrive at stock exchanges (Kissell, 2020). Historically, these systems were based on advanced statistical methods and signal processing designed to extract trading signals from financial data. The recent success of Machine Learning has attracted the interest of the financial community. Reinforcement Learning is a subcategory of machine learning and has been broadly applied by investors and researchers in building trading systems (Kissell, 2020). In this thesis, we address the issue that deep reinforcement learning may be susceptible to sampling errors and over-fitting and propose a robust deep reinforcement learning method that integrates techniques from reinforcement learning and robust optimization. We back-test and compare the performance of the developed algorithm, Robust DDPG, with UBAH (Uniform Buy and Hold) benchmark and other RL algorithms and show that the robust algorithm of this research can reduce the downside risk of an investment strategy significantly and can ensure a safer path for the investor’s portfolio value.

The diploma thesis is focused on the use of heuristic and metaheuristic methods to stabilization and controlling the selected systems distinguished by the deterministic chaos behavior. There are discussed parameterization of chosen optimization methods, which are the genetic algorithm, simulated annealing and pattern search. The thesis also introduced the suitable controlling methods and the definition of the objective function. In the theoretical part of the thesis there is a brief introduction to the deterministic chaos theory. The next chapters describes the most common and deployed methods in~the~control theory, especially OGY and Pyragas methods. The practical part of the thesis is divided into two chapters. The first one describes the~stabilization of the artifical chaotic systems with the time delayed Pyragas method - TDAS and its modification ETDAS. The second chapter shows the real chaotic system control. The Duffing oscillator system was chosen to serve this purpose.

Disertační práce se zabývá rychlým vyhledáváním regulárních výrazů v síťovém provozu s použitím technologie FPGA. Vyhledávání regulárních výrazů v síťovém provozu je výpočetně náročnou operací využívanou převážně v oblasti síťové bezpečnosti a v oblasti monitorování provozu vysokorychlostních počítačových sítí. Současná řešení neumožňují dosáhnout požadovaných multigigabitových propustností při dodržení všech požadavků, které jsou na vyhledávací jednotky kladeny. Nejvyšších propustností dosahují implementace založené na využití inovativních hardwarových architektur implementovaných v FPGA případně v ASIC. Tato disertační práce popisuje nové architektury vyhledávací jednotky, které jsou vhodné pro implementaci jak v FPGA tak v ASIC. Základní myšlenkou navržených architektur je využití perfektní hashovací funkce pro implementaci přechodové tabulky konečného automatu. Dále byla navržena architektura, která umožňuje uživateli zanést malou pravděpodobnost chyby při vyhledávání a tím snížit paměťové nároky vyhledávací jednotky. Disertační práce analyzuje vliv pravděpodobnosti této chyby na celkovou spolehlivost systému a srovnává ji s řešením používaným v současnosti. V rámci disertační práce byla provedena měření vlastností regulárních výrazů používaných při analýze provozu moderních počítačových sítí. Z provedené analýzy vyplývá, že velká část regulárních výrazů je vhodná pro implementaci pomocí navržených architektur. Pro dosažení vysoké propustnosti vyhledávací jednotky práce navrhuje nový algoritmus transformace abecedy, který umožňuje, aby vyhledávací jednotka zpracovala více znaků v jednom kroku. Na rozdíl od současných metod, navržený algoritmus umožňuje konstrukci automatu zpracovávajícího libovolný počet symbolů v jednom taktu. Implementované architektury dosahují v porovnání se současnými metodami úspory paměti zlepšení až 200MB.

Les processus markoviens déterministes par morceaux (PDMP) forment une vaste classe de processus stochastiques caractérisés par une évolution déterministe entre des sauts à mécanisme aléatoire. Ce sont des processus de type hybride, avec une composante discrète de mode et une composante d’état qui évolue dans un espace continu. Entre les sauts du processus, la composante continue évolue de façon déterministe, puis au moment du saut un noyau markovien sélectionne la nouvelle valeur des composantes discrète et continue. Dans cette thèse, nous construisons des PDMP évoluant dans des espaces de mesures (de dimension infinie), pour modéliser des population de cellules en tenant compte des caractéristiques individuelles de chaque cellule. Nous exposons notre construction des PDMP sur des espaces de mesure, et nous établissons leur caractère markovien. Sur ces processus à valeur mesure, nous étudions un problème d'arrêt optimal. Un problème d'arrêt optimal revient à choisir le meilleur temps d'arrêt pour optimiser l'espérance d'une certaine fonctionnelle de notre processus, ce qu'on appelle fonction valeur. On montre que cette fonction valeur est solution des équations de programmation dynamique et on construit une famille de temps d'arrêt $epsilon$-optimaux. Dans un second temps, nous nous intéressons à un PDMP en dimension finie, le TCP, pour lequel on construit un schéma d'Euler afin de l'approcher. Ce choix de modèle simple permet d'estimer différents types d'erreurs. Nous présentons des simulations numériques illustrant les résultats obtenus. / Piecewise deterministic Markov processes (PDMP) form a large class of stochastic processes characterized by a deterministic evolution between random jumps. They fall into the class of hybrid processes with a discrete mode and an Euclidean component (called the state variable). Between the jumps, the continuous component evolves deterministically, then a jump occurs and a Markov kernel selects the new value of the discrete and continuous components. In this thesis, we extend the construction of PDMPs to state variables taking values in some measure spaces with infinite dimension. The aim is to model cells populations keeping track of the information about each cell. We study our measured-valued PDMP and we show their Markov property. With thoses processes, we study a optimal stopping problem. The goal of an optimal stopping problem is to find the best admissible stopping time in order to optimize some function of our process. We show that the value fonction can be recursively constructed using dynamic programming equations. We construct some $epsilon$-optimal stopping times for our optimal stopping problem. Then, we study a simple finite-dimension real-valued PDMP, the TCP process. We use Euler scheme to approximate it, and we estimate some types of errors. We illustrate the results with numerical simulations.

Image charge detection is presented as a possible candidate to realise deterministic ion implantation. The deterministic placement of single impurities in solid substrates will enable a variety of novel applications, using their quantum mechanical properties for sensors or qubit registers. In this work, experimental techniques are used together with theoretical calculations to develop, characterise and optimise the detection of charged objects in a single pass through an image charge detector. In the main experimental part, ion bunches are employed as a model system for highly charged ions in proof-of-principle measurements with detector prototypes built in our labs. Image charge signals are characterised in the time and frequency domain. Using a statistical measurement and data analysis protocol, the noise and signal probability density functions are determined to calculate error and detection rates. It was found that even at an extremely low signal-to-noise ratio of 2, error rates can be suppressed effectively for high fidelity implantation. Aiming to improve the sensitivity, the maximum possible signal-to-noise ratio is calculated and discussed in dependence on the design parameters of an optimised image charge detector and the kinetic ion parameters. Lastly, a new ion implantation set-up combining focused ion beam technology with a source able to produce highly charged ions is introduced.

In this thesis, two novel methods for Inverse Uncertainty Quantification are benchmarked against the more established methods of Monte Carlo sampling of output parameters(MC) and Maximum Likelihood Estimation (MLE). Inverse Uncertainty Quantification (IUQ) is the process of how to best estimate the values of the input parameters in a simulation, and the uncertainty of said estimation, given a measurement of the output parameters. The two new methods are Deterministic Sampling (DS) and Weight Fixing (WF). Deterministic sampling uses a set of sampled points such that the set of points has the same statistic as the output. For each point, the corresponding point of the input is found to be able to calculate the statistics of the input. Weight fixing uses random samples from the rough region around the input to create a linear problem that involves finding the right weights so that the output has the right statistic. The benchmarking between the four methods shows that both DS and WF are comparably accurate to both MC and MLE in most cases tested in this thesis. It was also found that both DS and WF uses approximately the same amount of function calls as MLE and all three methods use a lot fewer function calls to the simulation than MC. It was discovered that WF is not always able to find a solution. This is probably because the methods used for WF are not the optimal method for what they are supposed to do. Finding more optimal methods for WF is something that could be investigated further.