<p>Gaussian Process Regression (GPR) is a widely used surrogate model in efficient global optimization (EGO) due to its capability to provide uncertainty estimates in the prediction. The cost of creating a GPR model for large data sets is high. On the other hand, neural network (NN) models scale better compared to GPR as the number of samples increase. Unfortunately, the uncertainty estimates for NN prediction are not readily available. In this work, a scalable algorithm is developed for EGO using NN-based prediction and uncertainty (EGONN). Initially, two different NNs are created using two different data sets. The first NN models the output based on the input values in the first data set while the second NN models the prediction error of the first NN using the second data set. The next infill point is added to the first data set based on criteria like expected improvement or prediction uncertainty. EGONN is demonstrated on the optimization of the Forrester function and a constrained Branin function and is compared with EGO. The convergence criteria is based on the maximum number of infill points in both cases. The algorithm is able to reach the optimum point within the given budget. The EGONN is extended to handle constraints explicitly and is utilized for aerodynamic shape optimization of the RAE 2822 airfoil in transonic viscous flow at a free-stream Mach number of 0.734 and a Reynolds number of 6.5 million. The results obtained from EGONN are compared with the results from gradient-based optimization (GBO) using adjoints. The optimum shape obtained from EGONN is comparable to the shape obtained from GBO and is able to eliminate the shock. The drag coefficient is reduced from 200 drag counts to 114 and is close to 110 drag counts obtained from GBO. The EGONN is also extended to handle uncertainty quantification (uqEGONN) using prediction uncertainty as an infill method. The convergence criteria is based on the relative change of summary statistics such as mean and standard deviation of an uncertain quantity. The uqEGONN is tested on Ishigami function with an initial sample size of 100 samples and the algorithm terminates after 70 infill points. The statistics obtained from uqEGONN (using only 170 function evaluations) are close to the values obtained from directly evaluating the function one million times. uqEGONN is demonstrated on to quantifying the uncertainty in the airfoil performance due to geometric variations. The algorithm terminates within 100 computational fluid dynamics (CFD) analyses and the statistics obtained from the algorithm are close to the one obtained from 1000 direct CFD based evaluations.</p>

<p>Uncertainty analysis is of great use both for calculating outputs that are more akin to real<br> flight, and for optimization to more robust shapes. However, implementation of uncertainty<br> has been a longstanding challenge in the field of aerodynamics due to the computational cost<br> of simulations. Geometric uncertainty in particular is often left unexplored in favor of uncer-<br> tainties in freestream parameters, turbulence models, or computational error. Therefore, this<br> work proposes a method of geometric uncertainty analysis for aerodynamic shapes that miti-<br> gates the barriers to its feasible computation. The process takes a two- or three-dimensional<br> shape and utilizes a combination of multifidelity meshes and Gaussian process regression<br> (GPR) surrogates in a multifidelity Monte Carlo (MFMC) algorithm. Multifidelity meshes<br> allow for finer sampling with a given budget, making the surrogates more accurate. GPR<br> surrogates are made practical to use by parameterizing major factors in geometric uncer-<br> tainty with only four variables in 2-D and five in 3-D. In both cases, two parameters control<br> the heights of steps that occur on the top and bottom of airfoils where leading and trailing<br> edge devices are attached. Two more parameters control the height and length of waves<br> that can occur in an ideally smooth shape during manufacturing. A fifth parameter controls<br> the depth of span-wise skin buckling waves along a 3-D wing. Parameters are defined to<br> be uniformly distributed with a maximum size of 0.4 mm and 0.15 mm for steps and waves<br> to remain within common manufacturing tolerances. The analysis chain is demonstrated<br> with two test cases. The first, the RAE2822 airfoil, uses transonic freestream parameters<br> set by the ADODG Benchmark Case 2. The results show a mean drag of nearly 10 counts<br> above the deterministic case with fixed lift, and a 2 count increase for a fixed angle of attack<br> version of the case. Each case also has small variations in lift and angle of attack of about<br> 0.5 counts and 0.08◦, respectively. Variances for each of the three tracked outputs show that<br> more variability is possible, and even likely. The ONERA M6 transonic wing, popular due<br> to the extensive experimental data available for computational validation, is the second test<br> case. Variation is found to be less substantial here, with a mean drag increase of 0.5 counts,<br> and a mean lift increase of 0.1 counts. Furthermore, the MFMC algorithm enables accurate<br> results with only a few hours of wall time in addition to GPR training. </p>

В работе рассматривается n-мерная дискретная модель, которая описывает динамику взаимодействия n потребителей. В рамках детерминированного анализа были построены карты режимов и бифуркационные диаграммы, описаны бифуркационные сценарии. Были обнаружены и описаны зоны мультистабильности системы, построены бассейны притяжения аттракторов. Далее в работе рассматривается стохастический вариант модели. Было изучено воздействие на систему аддитивного и параметрического шумов. С помощью функции стохастической чувствительности был проведен сравнительный анализ чувствительности равновесий и циклов. Опираясь на метод доверительных областей получены значения интенсивности шума, при которых наблюдаются индуцированные шумом явления. / The paper considers n-dimensional discrete model that describes the interaction dynamics of n consumers. As a part of the deterministic analysis, 2- and 1- parameter bifurcation diagrams were plotted, bifurcation scenarios were described. Multistability zones of the system were found and investigated, basins of attraction were plotted. Then, a stochastic version of the model is studied. The effect of additive and parametric noise on the system was described. Using the stochastic sensitivity function, a comparative analysis of the sensitivity of equilibria and cycles was carried out. Based on the method of confidence domains, the values of noise intensity, at which noise-induced phenomena can be observed, are obtained.

Organizations specifically the airlines industry are increasingly facing the challenges of operational efficiency measurement. During the last years enormous attention has been given to the assessment and improvement of the performance of productive systems. However, literatures show that there are limitations of the existing models to measure efficiency uniformly and exhaustively across the airlines. The problems are due to lack of the technical efficiency measuring model which unifies and integrates different measuring models into a single model.Therefore, this thesis investigates assessment of the operational performance of world major airlines by employing integrated comparative models to address the above problems. In this study, technical efficiency is addressed among many performance issues by using three types of modes of performance measurement: a non parametric one, represented by Data Envelopment Analysis (DEA) and; a parametric one, represented by Stochastic Frontier Analysis (SFA) and the Balance Scorecard (BSC) which is a strategic management tools. Unlike most of the previous studies, this study integrates the BSC concepts into DEA and SFA model. To evaluate technical efficiency of major international airlines, the study use panel of unbalanced data for the year 2007-2014 to make integrated comparative analysis. The research project incorporates seven leading variables and four lagging variables taken from BSC concept to implement into the DEA and SFA. All the three models of performance measurements have their own strength and limitation if they are used alone. But if the three models are integrated and combined together, they would yield better comparative and quality of efficiency assessment. Therefore, the study primarily developed a model beginning from the theoretical framework assumption into building of a unified comparative model of integrated comparative operational efficiency assessment of airlines. The research design and methodology uses secondary data collection i.e. annual reports and business reports of airlines which are collected from the airlines own website. The huge amount of financial and operational data cannot be collected by using primary data collection method as it would make it practically impossible and expensive. So by employing secondary data collection method saves time, money and a panel data can be accessed and generated easily. Hence, from 100 world major airlines population which are ranked by revenue, simple random sampling is used to select 80 samples airlines for this study. First, the BSC identifies the input and output variables. Next, the DEA model ranks the efficiency measurement, identifies the slack variables and benchmarks the airlines. Third, the SFA model identifies technical efficiency, the random error and technical inefficiency. Finally, the technical efficiency estimates obtained from the two techniques are analyzed comparatively. The research makes further analysis of particular case of the Ethiopian Airlines in relation to the most efficient and inefficient airlines and in comparison of the regional analysis. After extensive tests have been conducted, ‘Balanced Frontier Envelopment’ model is developed. According to this model, it is a paramount to measure efficiency with combining the strength of three models together and gives better results than the previous one or two combined models. The developed and integrated strategic model enhances measuring of the operating technical efficiency of airlines. This model benefit the airlines industry in many ways such as minimizing the cost and maximizing profit through managing technical efficiency which lead into the success of the airlines. From the model perspective, therefore, result of DEA model is much higher than the result of SFA model. DEA model is easier to manipulate than the SFA model because the former does not need the functional form while the later requires a functional form. Furthermore, according to the efficiency finding of the study, first, the European regional airlines are relatively more efficient than the rest of regions in the world. Second, the North America regional airlines are the second more efficient regional airlines in the world. Third, the Ethiopian airlines are the most efficient in Africa when we compare among Egyptair, Kenyan Airways and South African Airways. Fourth, high revenue does not necessarily leads to the technical efficiency of the firm. / Business Management / D.B.L. (Business Leadership)

Cette thèse est consacrée à l'étude de quelques problèmes dans le domaine des équations différentielles stochastiques rétrogrades (EDSR), et leurs applications aux équations aux dérivées partielles.Dans le premier chapitre, nous introduisons la notion d'équation différentielle doublement stochastique rétrograde (EDDSR) avec condition terminale singulière. Nous étudions d’abord les EDDSR avec générateur monotone, et obtenons ensuite un résultat d'existence par un schéma d'approximation. Une dernière section établit le lien avec les équations aux dérivées partielles stochastiques, via l'approche solution faible développée par Bally, Matoussi en 2001.Le deuxième chapitre est consacré aux EDSR avec condition terminale singulière et sauts. Comme dans le chapitre précédent la partie délicate sera de prouver la continuité en T. Nous formulons des conditions suffisantes sur les sauts afin d'obtenir cette dernière. Une section établit ensuite le lien entre solution minimale de l'EDSR et équations intégro-différentielles. Enfin le dernier chapitre est dédié aux équations différentielles stochastiques rétrogrades du second ordre (2EDSR) doublement réfléchies. Nous avons établi l'existence et l'unicité de telles équations. Ainsi, il nous a fallu dans un premier temps nous concentrer sur le problème de réflexion par barrière supérieure des 2EDSR. Nous avons ensuite combiné ces résultats à ceux existants afin de donner un cadre correct aux 2EDSRDR. L'unicité est conséquence d'une propriété de représentation et l'existence est obtenue en utilisant les espaces shiftés, et les distributions de probabilité conditionnelles régulières. Enfin une application aux jeux de Dynkin et aux options Israëliennes est traitée dans la dernière section. / This thesis is devoted to the study of some problems in the field of backward stochastic differential equations (BSDE), and their applications to partial differential equations.In the first chapter, we introduce the notion of backward doubly stochastic differential equations (BDSDE) with singular terminal condition. A first work consists to study the case of BDSDE with monotone generator. We then obtain existing result by an approximating scheme built considering a truncation of the terminal condition. The last part of this chapter aim to establish the link with stochastic partial differential equations, using a weak solution approach developed by Bally, Matoussi in 2001.The second chapter is devoted to the BSDEs with singular terminal conditions and jumps. As in the previous chapter the tricky part will be to prove continuity in T. We formulate sufficient conditions on the jumps in order to obtain it. A section is then dedicated to establish a link between a minimal solution of our BSDE and partial integro-differential equations.The last chapter is dedicated to doubly reflected second order backward stochastic differential equations (2DRBSDE). We have been looking to establish existence and uniqueness for such equations. In order to obtain this, we had to focus first on the upper reflection problem for 2BSDEs. We combined then these results to those already existing to give a well-posedness context to 2DRBSDE. Uniqueness is established as a straight consequence of a representation property. Existence is obtained using shifted spaces, and regular conditional probability distributions. A last part is then consecrated to the link with some Dynkin games and Israeli options.

<div>The rise of artificial intelligence (AI) hinges on the efficient training of modern deep neural networks (DNNs) for non-convex optimization and uncertainty quantification, which boils down to a non-convex Bayesian learning problem. A standard tool to handle the problem is Langevin Monte Carlo, which proposes to approximate the posterior distribution with theoretical guarantees. However, non-convex Bayesian learning in real big data applications can be arbitrarily slow and often fails to capture the uncertainty or informative modes given a limited time. As a result, advanced techniques are still required.</div><div><br></div><div>In this thesis, we start with the replica exchange Langevin Monte Carlo (also known as parallel tempering), which is a Markov jump process that proposes appropriate swaps between exploration and exploitation to achieve accelerations. However, the na\"ive extension of swaps to big data problems leads to a large bias, and the bias-corrected swaps are required. Such a mechanism leads to few effective swaps and insignificant accelerations. To alleviate this issue, we first propose a control variates method to reduce the variance of noisy energy estimators and show a potential to accelerate the exponential convergence. We also present the population-chain replica exchange and propose a generalized deterministic even-odd scheme to track the non-reversibility and obtain an optimal round trip rate. Further approximations are conducted based on stochastic gradient descents, which yield a user-friendly nature for large-scale uncertainty approximation tasks without much tuning costs. </div><div><br></div><div>In the second part of the thesis, we study scalable dynamic importance sampling algorithms based on stochastic approximation. Traditional dynamic importance sampling algorithms have achieved successes in bioinformatics and statistical physics, however, the lack of scalability has greatly limited their extensions to big data applications. To handle this scalability issue, we resolve the vanishing gradient problem and propose two dynamic importance sampling algorithms based on stochastic gradient Langevin dynamics. Theoretically, we establish the stability condition for the underlying ordinary differential equation (ODE) system and guarantee the asymptotic convergence of the latent variable to the desired fixed point. Interestingly, such a result still holds given non-convex energy landscapes. In addition, we also propose a pleasingly parallel version of such algorithms with interacting latent variables. We show that the interacting algorithm can be theoretically more efficient than the single-chain alternative with an equivalent computational budget.</div>

Wir bauen das verallgemeinerte diskrete Modell des zu Grunde liegenden Aktienpreisprozesses, der als eine bessere Annäherung an den Aktienpreisprozess dient als der klassische zufällige Spaziergang. Das verallgemeinerte Multinomial-Modell des Option-Preises in Bezug auf das neue Modell des Aktienpreisprozesses wird erhalten. Das entsprechende asymptotische Verfahren erlaubt, die verallgemeinerte Black-Scholes Formel zu erhalten, die die Formel als einen Begrenzungsfall des verallgemeinerten diskreten Option-Preis Modells bewertet.