Tesi sul tema "Stochastic second order methods"

Pour résoudre les problèmes de machine learning à grande échelle, les méthodes de premier ordre telles que la descente du gradient stochastique et l'ADAM sont les méthodes de choix en raison de leur coût pas cher par itération. Le problème des méthodes du premier ordre est qu'elles peuvent nécessiter un réglage important des paramètres et/ou une connaissance des paramètres du problème. Il existe aujourd'hui un effort considérable pour développer des méthodes du second ordre stochastiques efficaces afin de résoudre des problèmes de machine learning à grande échelle. La motivation est qu'elles demandent moins de réglage des paramètres et qu'elles convergent pour une plus grande variété de modèles et de datasets. Dans la première partie de la thèse, nous avons présenté une approche de principe pour désigner des méthodes de Newton stochastiques à fin de résoudre à la fois des équations non linéaires et des problèmes d'optimisation d'une manière efficace. Notre approche comporte deux étapes. Premièrement, nous pouvons réécrire les équations non linéaires ou le problème d'optimisation sous forme d'équations non linéaires souhaitées. Ensuite, nous appliquons de nouvelles méthodes du second ordre stochastiques pour résoudre ce système d'équations non linéaires. Grâce à notre approche générale, nous présentons de nombreux nouveaux algorithmes spécifiques du second ordre qui peuvent résoudre efficacement les problèmes de machine learning à grande échelle sans nécessiter de connaissance du problème ni de réglage des paramètres. Dans la deuxième partie de la thèse, nous nous concentrons sur les algorithmes d'optimisation appliqués à un domaine spécifique : l'apprentissage par renforcement (RL). Cette partie est indépendante de la première partie de la thèse. Pour atteindre de telles performances dans les problèmes de RL, le policie gradient (PG) et sa variante, le policie gradient naturel (NPG), sont les fondements de plusieurs algorithmes de l'état de l'art (par exemple, TRPO et PPO) utilisés dans le RL profond. Malgré le succès empirique des méthodes de RL et de PG, une compréhension théorique solide du PG de "vanille" a longtemps fait défaut. En utilisant la structure du RL du problème et des techniques modernes de preuve d'optimisation, nous obtenons nouvelles analyses en temps fini de la PG et de la NPG. Grâce à notre analyse, nous apportons également de nouvelles perspectives aux méthodes avec de meilleurs choix d'hyperparamètres
To solve large scale machine learning problems, first-order methods such as stochastic gradient descent and ADAM are the methods of choice because of their low cost per iteration. The issue with first order methods is that they can require extensive parameter tuning, and/or knowledge of the parameters of the problem. There is now a concerted effort to develop efficient stochastic second order methods to solve large scale machine learning problems. The motivation is that they require less parameter tuning and converge for wider variety of models and datasets. In the first part of the thesis, we presented a principled approach for designing stochastic Newton methods for solving both nonlinear equations and optimization problems in an efficient manner. Our approach has two steps. First, we can re-write the nonlinear equations or the optimization problem as desired nonlinear equations. Second, we apply new stochastic second order methods to solve this system of nonlinear equations. Through our general approach, we showcase many specific new second-order algorithms that can solve the large machine learning problems efficiently without requiring knowledge of the problem nor parameter tuning. In the second part of the thesis, we then focus on optimization algorithms applied in a specific domain: reinforcement learning (RL). This part is independent to the first part of the thesis. To achieve such high performance of RL problems, policy gradient (PG) and its variant, natural policy gradient (NPG), are the foundations of the several state of the art algorithms (e.g., TRPO and PPO) used in deep RL. In spite of the empirical success of RL and PG methods, a solid theoretical understanding of even the “vanilla” PG has long been elusive. By leveraging the RL structure of the problem together with modern optimization proof techniques, we derive new finite time analysis of both PG and NPG. Through our analysis, we also bring new insights to the methods with better hyperparameter choices

In this thesis, our research focus on a weak second order stochastic Runge–Kutta method applied to a system of stochastic differential equations known as the Gatheral Model. We approximate numerical solutions to this system and investigate the rate of convergence of our method. Both call and put options are priced using Monte-Carlo simulation to investigate the order of convergence. The numerical results show that our method is consistent with the theoretical order of convergence of the Monte-Carlo simulation. However, in terms of the Runge-Kutta method, we cannot accept the consistency of our method with the theoretical order of convergence without further research.

Dans cette thèse, nous étudions trois types de problèmes stochastiques : les problèmes avec contraintes probabilistes, les problèmes distributionnellement robustes et les problèmes avec recours. Les difficultés des problèmes stochastiques sont essentiellement liées aux problèmes de convexité du domaine des solutions, et du calcul de l’espérance mathématique ou des probabilités qui nécessitent le calcul complexe d’intégrales multiples. A cause de ces difficultés majeures, nous avons résolu les problèmes étudiées à l’aide d’approximations efficaces.Nous avons étudié deux types de problèmes stochastiques avec des contraintes en probabilités, i.e., les problèmes linéaires avec contraintes en probabilité jointes (LLPC) et les problèmes de maximisation de probabilités (MPP). Dans les deux cas, nous avons supposé que les variables aléatoires sont normalement distribués et les vecteurs lignes des matrices aléatoires sont indépendants. Nous avons résolu LLPC, qui est un problème généralement non convexe, à l’aide de deux approximations basée sur les problèmes coniques de second ordre (SOCP). Sous certaines hypothèses faibles, les solutions optimales des deux SOCP sont respectivement les bornes inférieures et supérieures du problème du départ. En ce qui concerne MPP, nous avons étudié une variante du problème du plus court chemin stochastique contraint (SRCSP) qui consiste à maximiser la probabilité de la contrainte de ressources. Pour résoudre ce problème, nous avons proposé un algorithme de Branch and Bound pour calculer la solution optimale. Comme la relaxation linéaire n’est pas convexe, nous avons proposé une approximation convexe efficace. Nous avons par la suite testé nos algorithmes pour tous les problèmes étudiés sur des instances aléatoires. Pour LLPC, notre approche est plus performante que celles de Bonferroni et de Jaganathan. Pour MPP, nos résultats numériques montrent que notre approche est là encore plus performante que l’approximation des contraintes probabilistes individuellement.La deuxième famille de problèmes étudiés est celle relative aux problèmes distributionnellement robustes où une partie seulement de l’information sur les variables aléatoires est connue à savoir les deux premiers moments. Nous avons montré que le problème de sac à dos stochastique (SKP) est un problème semi-défini positif (SDP) après relaxation SDP des contraintes binaires. Bien que ce résultat ne puisse être étendu au cas du problème multi-sac-à-dos (MKP), nous avons proposé deux approximations qui permettent d’obtenir des bornes de bonne qualité pour la plupart des instances testées. Nos résultats numériques montrent que nos approximations sont là encore plus performantes que celles basées sur les inégalités de Bonferroni et celles plus récentes de Zymler. Ces résultats ont aussi montré la robustesse des solutions obtenues face aux fluctuations des distributions de probabilités. Nous avons aussi étudié une variante du problème du plus court chemin stochastique. Nous avons prouvé que ce problème peut se ramener au problème de plus court chemin déterministe sous certaine hypothèses. Pour résoudre ce problème, nous avons proposé une méthode de B&B où les bornes inférieures sont calculées à l’aide de la méthode du gradient projeté stochastique. Des résultats numériques ont montré l’efficacité de notre approche. Enfin, l’ensemble des méthodes que nous avons proposées dans cette thèse peuvent s’appliquer à une large famille de problèmes d’optimisation stochastique avec variables entières
In this thesis, we studied three types of stochastic problems: chance constrained problems, distributionally robust problems as well as the simple recourse problems. For the stochastic programming problems, there are two main difficulties. One is that feasible sets of stochastic problems is not convex in general. The other main challenge arises from the need to calculate conditional expectation or probability both of which are involving multi-dimensional integrations. Due to the two major difficulties, for all three studied problems, we solved them with approximation approaches.We first study two types of chance constrained problems: linear program with joint chance constraints problem (LPPC) as well as maximum probability problem (MPP). For both problems, we assume that the random matrix is normally distributed and its vector rows are independent. We first dealt with LPPC which is generally not convex. We approximate it with two second-order cone programming (SOCP) problems. Furthermore under mild conditions, the optimal values of the two SOCP problems are a lower and upper bounds of the original problem respectively. For the second problem, we studied a variant of stochastic resource constrained shortest path problem (called SRCSP for short), which is to maximize probability of resource constraints. To solve the problem, we proposed to use a branch-and-bound framework to come up with the optimal solution. As its corresponding linear relaxation is generally not convex, we give a convex approximation. Finally, numerical tests on the random instances were conducted for both problems. With respect to LPPC, the numerical results showed that the approach we proposed outperforms Bonferroni and Jagannathan approximations. While for the MPP, the numerical results on generated instances substantiated that the convex approximation outperforms the individual approximation method.Then we study a distributionally robust stochastic quadratic knapsack problems, where we only know part of information about the random variables, such as its first and second moments. We proved that the single knapsack problem (SKP) is a semedefinite problem (SDP) after applying the SDP relaxation scheme to the binary constraints. Despite the fact that it is not the case for the multidimensional knapsack problem (MKP), two good approximations of the relaxed version of the problem are provided which obtain upper and lower bounds that appear numerically close to each other for a range of problem instances. Our numerical experiments also indicated that our proposed lower bounding approximation outperforms the approximations that are based on Bonferroni's inequality and the work by Zymler et al.. Besides, an extensive set of experiments were conducted to illustrate how the conservativeness of the robust solutions does pay off in terms of ensuring the chance constraint is satisfied (or nearly satisfied) under a wide range of distribution fluctuations. Moreover, our approach can be applied to a large number of stochastic optimization problems with binary variables.Finally, a stochastic version of the shortest path problem is studied. We proved that in some cases the stochastic shortest path problem can be greatly simplified by reformulating it as the classic shortest path problem, which can be solved in polynomial time. To solve the general problem, we proposed to use a branch-and-bound framework to search the set of feasible paths. Lower bounds are obtained by solving the corresponding linear relaxation which in turn is done using a Stochastic Projected Gradient algorithm involving an active set method. Meanwhile, numerical examples were conducted to illustrate the effectiveness of the obtained algorithm. Concerning the resolution of the continuous relaxation, our Stochastic Projected Gradient algorithm clearly outperforms Matlab optimization toolbox on large graphs

L'objectif de cette thèse est l'étude de la représentation probabiliste des différentes classes d'EDPSs non-linéaires(semi-linéaires, complètement non-linéaires, réfléchies dans un domaine) en utilisant les équations différentielles doublement stochastiques rétrogrades (EDDSRs). Cette thèse contient quatre parties différentes. Nous traitons dans la première partie les EDDSRs du second ordre (2EDDSRs). Nous montrons l'existence et l'unicité des solutions des EDDSRs en utilisant des techniques de contrôle stochastique quasi- sure. La motivation principale de cette étude est la représentation probabiliste des EDPSs complètement non-linéaires. Dans la deuxième partie, nous étudions les solutions faibles de type Sobolev du problème d'obstacle pour les équations à dérivées partielles inteégro-différentielles (EDPIDs). Plus précisément, nous montrons la formule de Feynman-Kac pour l'EDPIDs par l'intermédiaire des équations différentielles stochastiques rétrogrades réfléchies avec sauts (EDSRRs). Plus précisément, nous établissons l'existence et l'unicité de la solution du problème d'obstacle, qui est considérée comme un couple constitué de la solution et de la mesure de réflexion. L'approche utilisée est basée sur les techniques de flots stochastiques développées dans Bally et Matoussi (2001) mais les preuves sont beaucoup plus techniques. Dans la troisième partie, nous traitons l'existence et l'unicité pour les EDDSRRs dans un domaine convexe D sans aucune condition de régularité sur la frontière. De plus, en utilisant l'approche basée sur les techniques du flot stochastiques nous démontrons l'interprétation probabiliste de la solution faible de type Sobolev d'une classe d'EDPSs réfléchies dans un domaine convexe via les EDDSRRs. Enfin, nous nous intéressons à la résolution numérique des EDDSRs à temps terminal aléatoire. La motivation principale est de donner une représentation probabiliste des solutions de Sobolev d'EDPSs semi-linéaires avec condition de Dirichlet nul au bord. Dans cette partie, nous étudions l'approximation forte de cette classe d'EDDSRs quand le temps terminal aléatoire est le premier temps de sortie d'une EDS d'un domaine cylindrique. Ainsi, nous donnons les bornes pour l'erreur d'approximation en temps discret. Cette partie se conclut par des tests numériques qui démontrent que cette approche est effective
The objective of this thesis is to study the probabilistic representation (Feynman-Kac for- mula) of different classes ofStochastic Nonlinear PDEs (semilinear, fully nonlinear, reflected in a domain) by means of backward doubly stochastic differential equations (BDSDEs). This thesis contains four different parts. We deal in the first part with the second order BDS- DEs (2BDSDEs). We show the existence and uniqueness of solutions of 2BDSDEs using quasi sure stochastic control technics. The main motivation of this study is the probabilistic representation for solution of fully nonlinear SPDEs. First, under regularity assumptions on the coefficients, we give a Feynman-Kac formula for classical solution of fully nonlinear SPDEs and we generalize the work of Soner, Touzi and Zhang (2010-2012) for deterministic fully nonlinear PDE. Then, under weaker assumptions on the coefficients, we prove the probabilistic representation for stochastic viscosity solution of fully nonlinear SPDEs. In the second part, we study the Sobolev solution of obstacle problem for partial integro-differentialequations (PIDEs). Specifically, we show the Feynman-Kac formula for PIDEs via reflected backward stochastic differentialequations with jumps (BSDEs). Specifically, we establish the existence and uniqueness of the solution of the obstacle problem, which is regarded as a pair consisting of the solution and the measure of reflection. The approach is based on stochastic flow technics developed in Bally and Matoussi (2001) but the proofs are more technical. In the third part, we discuss the existence and uniqueness for RBDSDEs in a convex domain D without any regularity condition on the boundary. In addition, using the approach based on the technics of stochastic flow we provide the probabilistic interpretation of Sobolev solution of a class of reflected SPDEs in a convex domain via RBDSDEs. Finally, we are interested in the numerical solution of BDSDEs with random terminal time. The main motivation is to give a probabilistic representation of Sobolev solution of semilinear SPDEs with Dirichlet null condition. In this part, we study the strong approximation of this class of BDSDEs when the random terminal time is the first exit time of an SDE from a cylindrical domain. Thus, we give bounds for the discrete-time approximation error.. We conclude this part with numerical tests showing that this approach is effective

Cette thèse traite des équations différentielles stochastiques rétrogrades réfléchies du second ordre dans une filtration générale . Nous avons traité tout d'abord la réflexion à une barrière inférieure puis nous avons étendu le résultat dans le cas d'une barrière supérieure. Notre contribution consiste à démontrer l'existence et l'unicité de la solution de ces équations dans le cadre d'une filtration générale sous des hypothèses faibles. Nous remplaçons la régularité uniforme par la régularité de type Borel. Le principe de programmation dynamique pour le problème de contrôle stochastique robuste est donc démontré sous les hypothèses faibles c'est à dire sans régularité sur le générateur, la condition terminal et la barrière. Dans le cadre des Équations Différentielles Stochastiques Rétrogrades (EDSRs ) standard, les problèmes de réflexions à barrières inférieures et supérieures sont symétriques. Par contre dans le cadre des EDSRs de second ordre, cette symétrie n'est plus valable à cause des la non linéarité de l'espérance sous laquelle est définie notre problème de contrôle stochastique robuste non dominé. Ensuite nous un schéma d'approximation numérique d'une classe d'EDSR de second ordre réfléchies. En particulier nous montrons la convergence de schéma et nous testons numériquement les résultats obtenus
This thesis deals with the second-order reflected backward stochastic differential equations (2RBSDEs) in general filtration. In the first part , we consider the reflection with a lower obstacle and then extended the result in the case of an upper obstacle . Our main contribution consists in demonstrating the existence and the uniqueness of the solution of these equations defined in the general filtration under weak assumptions. We replace the uniform regularity by the Borel regularity(through analytic measurability). The dynamic programming principle for the robust stochastic control problem is thus demonstrated under weak assumptions, that is to say without regularity on the generator, the terminal condition and the obstacle. In the standard Backward Stochastic Differential Equations (BSDEs) framework, there is a symmetry between lower and upper obstacles reflection problem. On the contrary, in the context of second order BSDEs, this symmetry is no longer satisfy because of the nonlinearity of the expectation under which our robust stochastic non-dominated stochastic control problem is defined. In the second part , we get a numerical approximation scheme of a class of second-order reflected BSDEs. In particular we show the convergence of our scheme and we test numerically the results

This thesis presents an approximate solution of second order relative motion equations. The equations of motion for a Keplerian orbit in spherical coordinates are expanded in Taylor series form using reference conditions consistent with that of a circular orbit. Only terms that are linear or quadratic in state variables are kept in the expansion. A perturbation method is employed to obtain an approximate solution of the resulting nonlinear differential equations. This new solution is compared with the previously known solution of the linear case to show improvement, and with numerical integration of the quadratic differential equation to understand the error incurred by the approximation. In all cases, the comparison is made by computing the difference of the approximate state (analytical or numerical) from numerical integration of the full nonlinear Keplerian equations of motion.
Master of Science

Analytical chemistry can be split into two main types, qualitative and quantitative. Most modern analytical chemistry is quantitative. Popular sensitivity to health issues is aroused by the mountains of government regulations that use science to, for instance, provide public health information to prevent disease caused by harmful exposure to toxic substances. The concept of the minimum amount of an analyte or compound that can be detected or analysed appears in many of these regulations (for example, to discard the presence of traces of toxic substances in foodstuffs) generally as a part of method validation aimed at reliably evaluating the validity of the measurements.

The lowest quantity of a substance that can be distinguished from the absence of that substance (a blank value) is called the detection limit or limit of detection (LOD). Traditionally, in the context of simple measurements where the instrumental signal only depends on the amount of analyte, a multiple of the blank value is taken to calculate the LOD (traditionally, the blank value plus three times the standard deviation of the measurement). However, the increasing complexity of the data that analytical instruments can provide for incoming samples leads to situations in which the LOD cannot be calculated as reliably as before.

Measurements, instruments and mathematical models can be classified according to the type of data they use. Tensorial theory provides a unified language that is useful for describing the chemical measurements, analytical instruments and calibration methods. Instruments that generate two-dimensional arrays of data are second-order instruments. A typical example is a spectrofluorometer, which provides a set of emission spectra obtained at different excitation wavelengths.

The calibration methods used with each type of data have different features and complexity. In this thesis, the most commonly used calibration methods are reviewed, from zero-order (or univariate) to second-order (or multi-linears) calibration models. Second-order calibration models are treated in details since they have been applied in the thesis.

Concretely, the following methods are described:
- PARAFAC (Parallel Factor Analysis)
- ITTFA (Iterative Target Transformation Analysis)
- MCR-ALS (Multivariate Curve Resolution-Alternating Least Squares)
- N-PLS (Multi-linear Partial Least Squares)

Analytical methods should be validated. The validation process typically starts by defining the scope of the analytical procedure, which includes the matrix, target analyte(s), analytical technique and intended purpose. The next step is to identify the performance characteristics that must be validated, which may depend on the purpose of the procedure, and the experiments for determining them. Finally, validation results should be documented, reviewed and maintained (if not, the procedure should be revalidated) as long as the procedure is applied in routine work.

The figures of merit of a chemical analytical process are 'those quantifiable terms which may indicate the extent of quality of the process. They include those terms that are closely related to the method and to the analyte (sensitivity, selectivity, limit of detection, limit of quantification, ...) and those which are concerned with the final results (traceability, uncertainty and representativity) (Inczédy et al., 1998). The aim of this thesis is to develop theoretical and practical strategies for calculating the limit of detection for complex analytical situations. Specifically, I focus on second-order calibration methods, i.e. when a matrix of data is available for each sample.

The methods most often used for making detection decisions are based on statistical hypothesis testing and involve a choice between two hypotheses about the sample. The first hypothesis is the "null hypothesis": the sample is analyte-free. The second hypothesis is the "alternative hypothesis": the sample is not analyte-free. In the hypothesis test there are two possible types of decision errors. An error of the first type occurs when the signal for an analyte-free sample exceeds the critical value, leading one to conclude incorrectly that the sample contains a positive amount of the analyte. This type of error is sometimes called a "false positive". An error of the second type occurs if one concludes that a sample does not contain the analyte when it actually does and it is known as a "false negative". In zero-order calibration, this hypothesis test is applied to the confidence intervals of the calibration model to estimate the LOD as proposed by Hubaux and Vos (A. Hubaux, G. Vos, Anal. Chem. 42: 849-855, 1970).

One strategy for estimating multivariate limits of detection is to transform the multivariate model into a univariate one. This strategy has been applied in this thesis in three practical applications:
1. LOD for PARAFAC (Parallel Factor Analysis).
2. LOD for ITTFA (Iterative Target Transformation Factor Analysis).
3. LOD for MCR-ALS (Multivariate Curve Resolution - Alternating Least Squares)

In addition, the thesis includes a theoretical contribution with the proposal of a sample-dependent LOD in the context of multivariate (PLS) and multi-linear (N-PLS) Partial Least Squares.
La Química Analítica es pot dividir en dos tipus d'anàlisis, l'anàlisi quantitativa i l'anàlisi qualitativa. La gran part de la química analítica moderna és quantitativa i fins i tot els govern fan ús d'aquesta ciència per establir regulacions que controlen, per exemple, nivells d'exposició a substàncies tòxiques que poden afectar la salut pública. El concepte de mínima quantitat d'un analit o component que es pot detectar apareix en moltes d'aquestes regulacions, en general com una part de la validació dels mètodes per tal de garantir la qualitat i la validesa dels resultats.

La mínima quantitat d'una substància que pot ser diferenciada de l'absència d'aquesta substància (el que es coneix com un blanc) s'anomena límit de detecció (limit of detection, LOD). En procediments on es treballa amb mesures analítiques que són degudes només a la quantitat d'analit present a la mostra (situació d'ordre zero) el LOD es pot calcular com un múltiple de la mesura del blanc (tradicionalment, 3 vegades la desviació d'aquesta mesura). Tanmateix, l'evolució dels instruments analítics i la complexitat creixent de les dades que generen, porta a situacions en les que el LOD no es pot calcular fiablement d'una forma tan senzilla. Les mesures, els instruments i els models de calibratge es poden classificar en funció del tipus de dades que utilitzen. La Teoria Tensorial s'ha utilitzat en aquesta tesi per fer aquesta classificació amb un llenguatge útil i unificat. Els instruments que generen dades en dues dimensions s'anomenen instruments de segon ordre i un exemple típic és l'espectrofluorímetre d'excitació-emissió, que proporciona un conjunt d'espectres d'emissió obtinguts a diferents longituds d'ona d'excitació.

Els mètodes de calibratge emprats amb cada tipus de dades tenen diferents característiques i complexitat. En aquesta tesi, es fa una revisió dels models de calibratge més habituals d'ordre zero (univariants), de primer ordre (multivariants) i de segon ordre (multilinears). Els mètodes de segon ordre estan tractats amb més detall donat que són els que s'han emprat en les aplicacions pràctiques portades a terme.

Concretament es descriuen:

- PARAFAC (Parallel Factor Analysis)
- ITTFA (Iterative Target Transformation Analysis)
- MCR-ALS (Multivariate Curve Resolution-Alternating Least Squares)
- N-PLS (Multi-linear Partial Least Squares)

Com s'ha avançat al principi, els mètodes analítics s'han de validar. El procés de validació inclou la definició dels límits d'aplicació del procediment analític (des del tipus de mostres o matrius fins l'analit o components d'interès, la tècnica analítica i l'objectiu del procediment). La següent etapa consisteix en identificar i estimar els paràmetres de qualitat (figures of merit, FOM) que s'han de validar per, finalment, documentar els resultats de la validació i mantenir-los mentre sigui aplicable el procediment descrit.

Algunes FOM dels processos químics de mesura són: sensibilitat, selectivitat, límit de detecció, exactitud, precisió, etc. L'objectiu principal d'aquesta tesi és desenvolupar estratègies teòriques i pràctiques per calcular el límit de detecció per problemes analítics complexos. Concretament, està centrat en els mètodes de calibratge que treballen amb dades de segon ordre.

Els mètodes més emprats per definir criteris de detecció estan basats en proves d'hipòtesis i impliquen una elecció entre dues hipòtesis sobre la mostra. La primera hipòtesi és la hipòtesi nul·la: a la mostra no hi ha analit. La segona hipòtesis és la hipòtesis alternativa: a la mostra hi ha analit. En aquest context, hi ha dos tipus d'errors en la decisió. L'error de primer tipus té lloc quan es determina que la mostra conté analit quan no en té i la probabilitat de cometre l'error de primer tipus s'anomena fals positiu. L'error de segon tipus té lloc quan es determina que la mostra no conté analit quan en realitat si en conté i la probabilitat d'aquest error s'anomena fals negatiu. En calibratges d'ordre zero, aquesta prova d'hipòtesi s'aplica als intervals de confiança de la recta de calibratge per calcular el LOD mitjançant les fórmules d'Hubaux i Vos (A. Hubaux, G. Vos, Anal. Chem. 42: 849-855, 1970)

Una estratègia per a calcular límits de detecció quan es treballa amb dades de segon ordre es transformar el model multivariant en un model univariant. Aquesta estratègia s'ha fet servir en la tesi en tres aplicacions diferents::
1. LOD per PARAFAC (Parallel Factor Analysis).
2. LOD per ITTFA (Iterative Target Transformation Factor Analysis).
3. LOD per MCR-ALS (Multivariate Curve Resolution - Alternating Least Squares)

A més, la tesi inclou una contribució teòrica amb la proposta d'un LOD que és específic per cada mostra, en el context del mètode multivariant PLS i del multilinear N-PLS.

Thesis (MScIng( Mechanical Engineering)--University of Stellenbosch, 2007.
121 Leaves printed single pages, preliminary pages a-l and numbered pages 1-81.
ENGLISH ABSTRACT:In the midst of the current non-renewable energy crises specifically with regard to fossil fuel, various research institutions across the world have turned their focus to renewable and sustainable development. Using our available non.renewable resources as efficiently as possible has been a focal point the past decades and will certainly be as long as these resources exist Various means to utilize the world's abundant and freely available renewable energy has been studied and some even introduced and installed as sustainable energy sources, Electricity generation by means of wind powered turbines, photo-voltaic cells, and tidal and wave energy are but a few examples. Modern photo-voltaic cells are known to have a solar to electricity conversion efficiency of 12% (Van Heerden, 2003) while wind turbines have an approximate wind to electricity conversion efficiency of 50% (Twele et aI., 2002). This low solar to electricity conversion efficiency together with the fact that renewable energy research is a relatively modern development, lead to the investigation into methods capable of higher solar to electricity conversion efficiencies. One such method could be to use the relatively old technology of the Stirling cycle developed in the early 1800's (solar to electricity conversion efficiency in the range of 20.24 % according Van Heerden, 2003). The Stirling cycle provides a method for converting thermal energy to mechanical power which can be used to generate electricity, One of the main advantages of Stirling machines is that they are capable of using any form of heat source ranging from solar to biomass and waste heat. This document provides a discussion of some of the available methods for the analysis of Stirling machines. The six (6) different methods considered include: the method of Beale, West, mean-pressurepower- formula (MPPF), Schmidt, idea! adiabatic and the simple analysis methods. The first three (3) are known to be good back-of-the-envelope methods specifically for application as synthesis tools during initialisation of design procedures, while the latter three (3) are analysis tools finding application during Stirling engine design and analysis procedures. These analysis methods are based on the work done by Berchowitz and Urieli (1984) and form the centre of this document. Sections to follow provide a discussion of the mathematical model as well as the MATlAB implementation thereof. Experimental tests were conducted on the Heinrici engine to provide verification of the simulated resutls. Shortcomings of these analyses methods are also discussed in the sections to follow. Recommendations regarding improvements of the simulation program, possible fields of application for Stirling technology, as well as future fields of study are made in the final chapter of this document. A review of relevanl literature regarding modern applications of Stirling technology and listings of companies currently manufacturing and developing Stirling machines and findings of research done at various other institutions are provided.
AFRIKAANSE OPSOMMING:Die tempo van uitputling van die wereld se nie-hernubare energiebronne die afgelope jare het aanleiding gegee daartoe dal daar loenemend fokus toegespits word op die ontwikkeling van hernubare alternatiewe. Meer doeltreffende benutting van die wereld se nie-hernubare energie is reeds 'n fokus punt, vir navorsers reg oor die wereld, vir die afgelope dekades. Die aarde se oorvloedryke hernubare energie bronne word reeds met verskeie metodes ontgin. Die omskakeling van wind-, son- en gety energie na elektrisieteids is net 'n paar voorbeelde. Die effektiwiteid van sonkrag na elektrisietyds omskakeling van moderne fotovo!la'iese selle is in die orde van 12% (Van Heerden, 2003) terwyl die doeltreffendeid van wind energie na elektrisiteit omskakelling in die orde van 50% (Twele et at, 2002) is. Hierdie relatief lae omskelings doeltreffendeid van sonkrag na elektrisietyd, tesame met die feit dat die hernubare industrie nag relatief jonk is, lei lot die soeke na ander meer doellreffende moontlikhede Die Stirling siklus is nie 'n mod erne beginsel nie, maar die toepassing daarvan veral in die hernubare energie industrie is wei 'n relatiewe nuwe beg rip, veral in teme van die omskakeling van sonkrag na elektriese energie (gemiddelde sonkrag na lektriese energie omskakelings doellreffendeid in die orde van 20-24% is gevind deur Van Heerden, 2003). Die omskakeling van lermiese energie na meganiese energie is sekerlik die hoof uitkomsle van die Stirling siklus, alhoewel dit ook toepassing vind in die verkoefingsindustrie. Die feit dat die Stirling siklus van enige vorm van termiese energie (bv. son. biomassa, asook hilte geproduseer as byproduk tydens sekere prosesse) gebruik kan maak. is een van die redes wat die tegnologie 56 aanloklik maak, spesifiek !.o,v. die hernubare energie sektor. Ses (6) metodes vir die analise van die Stirling siklus word in hierdie dokument bespreek. Dit slui! die volgnde in: Beale-, West-, die gemiddelde-druk-krag-metode (GDKM), Schmidt-, adiabatiese- en die eenvoudige analise melodes. Die eerste drie (3) metodes is handige berekenings metodes Iydens die aanvangs en sinlesefase van Stirling enjin ontwerp, lerwyl die laaste drie (3) meer loegespils is op die volledige ontwerps- en analisefases gedurende die Stirling eniin ontwerps proses. Die drie (3) analise melodes is gebaseer op die werk wat deur Berchowitz en Urieli (1984) gedoen is en maak die kern van die dokument uit. Die wiskundige model, implimentering daarvan in MATlAB, sowel as die eksperimentele verifieering van die resultate word bespreek. Tekortkominge van die analise metodes word ook aangespreek in elke hoofsluk. Moontlikke verbeterings len opsigte van die verskeie aannames word in die finale hoofsluk van die dokumenl aangespreek. Verskeie voorgestelde riglings vir toekomslige navorsings projekle word ook in die finale hoofstuk van die dokument genoem. 'n Kort oorsig van die relevanle lileraluur in verband mel huidige loepassings van die Stirling legnologie, asook die name van maatskappye wal tans hierdie tegnologiee ontwikkel en vervaardig, word genoem.

L'objectif principal de cette thèse est d'étudier quelques problèmes de mathématiques financières dans un marché incomplet avec incertitude sur les modèles. Récemment, la théorie des équations différentielles stochastiques rétrogrades du second ordre (2EDSRs) a été développée par Soner, Touzi et Zhang sur ce sujet. Dans cette thèse, nous adoptons leur point de vue. Cette thèse contient quatre parties dans le domain des 2EDSRs. Nous commençons par généraliser la théorie des 2EDSRs initialement introduite dans le cas de générateurs lipschitziens continus à celui de générateurs à croissance quadratique. Cette nouvelle classe des 2EDSRs nous permettra ensuite d'étudier le problème de maximisation d'utilité robuste dans les modèles non-dominés. Dans la deuxième partie, nous étudions ce problème pour trois fonctions d'utilité. Dans chaque cas, nous donnons une caractérisation de la fonction valeur et d'une stratégie d'investissement optimale via la solution d'une 2EDSR. Dans la troisième partie, nous fournissons également une théorie d'existence et unicité pour des EDSRs réfléchies du second ordre avec obstacles inférieurs et générateurs lipschitziens, nous appliquons ensuite ce résultat à l'étude du problème de valorisation des options américaines dans un modèle financier à volatilité incertaine. Dans la quatrième partie, nous étudions des 2EDSRs avec sauts. En particulier, nous prouvons l'existence d'une unique solution dans un espace approprié. Comme application de ces résultats, nous étudions un problème de maximisation d'utilité exponentielle robuste avec incertitude sur les modèles. L'incertitude affecte à la fois le processus de volatilité, mais également la mesure des sauts
The main objective of this PhD thesis is to study some financial mathematics problems in an incomplete market with model uncertainty. In recent years, the theory of second order backward stochastic differential equations (2BSDEs for short) has been developed by Soner, Touzi and Zhang on this topic. In this thesis, we adopt their point of view. This thesis contains of four key parts related to 2BSDEs. In the first part, we generalize the 2BSDEs theory initially introduced in the case of Lipschitz continuous generators to quadratic growth generators. This new class of 2BSDEs will then allow us to consider the robust utility maximization problem in non-dominated models. In the second part, we study this problem for exponential utility, power utility and logarithmic utility. In each case, we give a characterization of the value function and an optimal investment strategy via the solution to a 2BSDE. In the third part, we provide an existence and uniqueness result for second order reflected BSDEs with lower obstacles and Lipschitz generators, and then we apply this result to study the problem of American contingent claims pricing with uncertain volatility. In the fourth part, we define a notion of 2BSDEs with jumps, for which we prove the existence and uniqueness of solutions in appropriate spaces. We can interpret these equations as standard BSDEs with jumps, under both volatility and jump measure uncertainty. As an application of these results, we shall study a robust exponential utility maximization problem under model uncertainty, where the uncertainty affects both the volatility process and the jump measure

Many variable selection techniques have been developed that focus on first-order linear regression models. In some applications, such as modeling response surfaces, fitting second-order terms can improve predictive accuracy. However, the number of spurious interactions can be large leading to poor results with many methods. We focus on forward selection, describing algorithms that use the natural hierarchy existing in second-order linear regression models to limit spurious interactions. We then develop stopping rules by extending False Selection Rate methodology to these algorithms. In addition, we describe alternative estimation methods for fitting regression models including the LASSO, CART, and MARS. We also propose a general method for controlling multiple-group false selection rates, which we apply to second-order linear regression models. By estimating a separate entry level for first-order and second-order terms, we obtain equal contributions to the false selection rate from each group. We compare the methods via Monte Carlo simulation and apply them to optimizing response surface experimental designs.

This thesis is concerned with the efficient numerical solution of nonlinear partial differential equations (PDEs) of elliptic and parabolic type. Such PDEs arise frequently in models used to describe many physical phenomena, from the diffusion of a toxin in soil to the flow of viscous fluids. The main focus of this research is to better understand the implementation and performance of nonlinear multigrid methods for the solution of elliptic and parabolic PDEs, following their discretisation. For the most part finite element discretisations are considered, but other techniques are also discussed. Following discretisation of a PDE the two most frequently used nonlinear multigrid methods are Newton-Multigrid and the Full Approximation Scheme (FAS). These are both very efficient algorithms, and have the advantage that when they are applied to practical problems, their execution times scale linearly with the size of the problem being solved. Even though this has yet to be proved in theory for most problems, these methods have been widely adopted in practice in order to solve highly complex nonlinear (systems of) PDEs. Many research groups use either Newton-MG or FAS without much consideration as to which should be preferred, since both algorithms perform satisfactorily. In this thesis we address the question as to which method is likely to be more computationally efficient in practice. As part of this investigation the implementation of the algorithms is considered in a framework which allows the direct comparison of the computational effort of the two iterations. As well as this, the convergence properties of the methods are considered, applied to a variety of model problems. Extensive results are presented in the comparison, which are explained by available theory whenever possible. The strength and range of results presented allows us to confidently conclude that for a practical problem, discretised using a finite element discretisation, an improved efficiency and stability of a Newton-MG iteration, compared to an FAS iteration, is likely to be observed. The relative advantage of a Newton-MG method is likely to be larger the more complex the problem being solved becomes.

The main objective of the thesis is evaluating and comparing Second-Order Elastic Analysis Methods defined in two different specifications, AISC 2005 and TS648 (1980). There are many theoretical approaches that can provide exact solution for the problem. However, approximate methods are still needed for design purposes. Simple formulations for code applications were developed, and they are valid as acceptable results can be obtained within admissible error limits. Within the content of the thesis, firstly background information related to second-order effects will be presented. The emphasis will be on the definition of geometric non-linearity, also called as P-&
#948
and P-&
#916
effects. In addition, the approximate methods defined in AISC 2005 (B1 &ndash
B2 Method), and TS648 (1980) will be discussed in detail. Then, example problems will be solved for the demonstration of theoretical formulations for members with and without end translation cases. Also, the results obtained from the structural analysis software, SAP2000, will be compared with the results acquired from the exact and the approximate methods. Finally, conclusions related to the study will be stated.

This thesis is mostly about the analysis of second order linear vibrating systems. The main purpose of this study is to extend methods which have previously been developed for either undamped or proportionally damped or classically damped systems to the general case. These methods are commonly used in aerospace industries. Ground vibration testing of aircraft is performed to identify the dynamic behaviour of the structure. New aircraft materials and joining methods - composite materials and/or novel adhesive bonding approaches in place of riveted or welded joints - cause higher levels of damping that have not been seen before in aircraft structure. Any change occurring in an original structure causes associated changes of the dynamic behaviour of the structure. Analytical finite element analyses and experimental modal testing have become essential tools for engineers. These techniques are used to determine the dynamic characteristics of mechanical structures. In Chapters 3 and 4, structural analysis and modal testing have been carried out an aircraft-like structure. Modal analysis techniques are used to extract modal data which are identified from a single column of the frequency response matrix. The proposed method is presented for fitting modal peaks one by one. This technique overcomes the difficulty due to the conventional methods which require a series of measured FRFs at different points of excitation. New methods presented in this thesis are developed and implemented initially for undamped systems in all cases. These ideas are subsequently extended for generally damped linear systems. The equations of motion of second order damped systems are represented in state space. These methods have been developed based on Lancaster Augmented Matrices (LAMs) and diagonalising structure preserving equivalences (DSPEs). In Chapter 5, new methods are developed for computing the derivatives of the non-zeros of the diagonalised system and the derivatives of the diagonalising SPEs with respect to modifications in the system matrices. These methods have provided a new approach to the evaluation and the understanding of eigenvalue and eigenvector derivatives. This approach resolves the quandary where eigenvalue and eigenvector derivatives become undefined when a pair of complex eigenvalues turns into a pair of real eigenvalues or vice-versa. They also have resolved when any one or more of the system matrices is singular. Numerical examples have illustrated the new methods and they have shown that the method results overcome certain difficulties of conventional methods. In Chapter 6, Möbius transformations are used to address a problem where the mass matrix is singular. Two new transformations are investigated called system spectral transformation SSTNQ and diagonalising spectral/similarity transformation DSTOQ. The transformation SSTNQ maps between matrices of two systems having the same short eigenvectors and their diagonalised system matrices. The transformation DSTOQ maps between two diagonalising SPE‟s having identical eigenvalues. Modal correlation methods are implemented to evaluate and quantify the differences between the output results from these techniques. Different cross orthogonality measures represent a class of methods which are recently performed as modal correlation for damped systems. In Chapter 7, cross orthogonality measures and mutual orthogonality measures are developed for undamped systems. These measures are defined in terms of real matrices - the diagonalising structure preserving equivalences (DSPEs). New methods are well developed for ill-conditioned system such that they work for all occasions and not only for cases where mass matrix is non-singular. Also a measure of the residuals is introduced which does not demand invertibility of diagonalised system matrices. Model updating methods are used in order to update models of systems by matching the output results from analytical system models with the experimentally obtained values. In Chapter 8, both cross-orthogonality measures and mutual-orthogonality measures are developed and used in the model updating of generally damped linear systems. Model updating based on the mutual orthogonality measures exhibits monotonic convergence from every starting position. That is to say, the ball of convergence has an infinite radius whereas updating procedures based on comparing eigenvectors exhibit a finite ball of convergence. Craig Bampton transformations are one of component methods which are used to reduce and decouple large structure systems. In Chapter 9 Craig Bampton transformations are developed for undamped systems and extended for damped second order systems in state space. Craig Bampton transformations are generalised and presented in SPEs forms. The two parts of the Craig Bampton transformations are extended in the full sizes of the substructure. The extended Craig Bampton transformations are modified to format each block of transformed substructure matrices as LAMs matrices format. This thesis generalises and develops the methods mentioned above and illustrates these concepts with an experimental modal test and some examples. The thesis also contains brief information about basic vibration properties of general linear structures and literature review relevant to this project.

Approved for public release; distribution is unlimited
There are numerous numerical methods for solving different types of partial differential equations (PDEs) that describe the physical dynamics of the world. For instance, PDEs are used to understand fluid flow for aerodynamics, wave dynamics for seismic exploration, and orbital mechanics. The goal of these numerical methods is to approximate the solution to a continuous PDE with an accurate discrete representation. The focus of this thesis is to explore a new Discontinuous Galerkin (DG) method for approximating the second order wave equation in complex geometries with curved elements. We begin by briefly highlighting some of the numerical methods used to solve PDEs and discuss the necessary concepts to understand DG methods. These concepts are used to develop a one- and two-dimensional DG method with an upwind flux, boundary conditions, and curved elements. We demonstrate convergence numerically and prove discrete stability of the method through an energy analysis.

This thesis covers topics such as finite difference schemes, mean-square convergence, modelling, and numerical approximations of second order quasi-linear stochastic partial differential equations (SPDE) driven by white noise in less than three space dimensions. The motivation for discussing and expanding these topics lies in their implications in such physical phenomena as signal and information flow, gravitational and electromagnetic fields, large scale weather systems, and macro-computer networks. Chapter 2 delves into the hyperbolic SPDE in one space and one time dimension. This is an important equation to such fields as signal processing, communications, and information theory where singularities propagate throughout space as a function of time. Chapter 3 discusses some concepts and implications of elliptic SPDE's driven by additive noise. These systems are key for understanding steady state phenomena. Chapter 4 presents some numerical work regarding elliptic SPDE's driven by multiplicative and general noise. These SPDE's are open topics in the theoretical literature, hence numerical work provides significant insight into the nature of the process. Chapter 5 presents some numerical work regarding quasi-geostrophic geophysical fluid dynamics involving stochastic noise and demonstrates how these systems can be represented as a combination of elliptic and hyperbolic components.

Dans ce travail, nous développons et examinons deux nouveaux algorithmes d'éclatement du premier ordre pour résoudre des problèmes d'optimisation composites à grande échelle dans des espaces à dimensions infinies. Ces problèmes sont au coeur de nombres de domaines scientifiques et d'ingénierie, en particulier la science des données et l'imagerie. Notre travail est axé sur l'assouplissement des hypothèses de régularité de Lipschitz généralement requises par les algorithmes de fractionnement du premier ordre en remplaçant l'énergie euclidienne par une divergence de Bregman. Ces développements permettent de résoudre des problèmes ayant une géométrie plus exotique que celle du cadre euclidien habituel. Un des algorithmes développés est l'hybridation de l'algorithme de gradient conditionnel, utilisant un oracle de minimisation linéaire à chaque itération, avec méthode du Lagrangien augmenté, permettant ainsi la prise en compte de contraintes affines. L'autre algorithme est un schéma d'éclatement primal-dual incorporant les divergences de Bregman pour le calcul des opérateurs proximaux associés. Pour ces deux algorithmes, nous montrons la convergence des valeurs Lagrangiennes, la convergence faible des itérés vers les solutions ainsi que les taux de convergence. En plus de ces nouveaux algorithmes déterministes, nous introduisons et étudions également leurs extensions stochastiques au travers d'un point de vue d'analyse de stablité aux perturbations. Nos résultats dans cette partie comprennent des résultats de convergence presque sûre pour les mêmes quantités que dans le cadre déterministe, avec des taux de convergence également. Enfin, nous abordons de nouveaux problèmes qui ne sont accessibles qu'à travers les hypothèses relâchées que nos algorithmes permettent. Nous démontrons l'efficacité numérique et illustrons nos résultats théoriques sur des problèmes comme la complétion de matrice parcimonieuse de rang faible, les problèmes inverses sur le simplexe, ou encore les problèmes inverses impliquant la distance de Wasserstein régularisée
In this work we develop and examine two novel first-order splitting algorithms for solving large-scale composite optimization problems in infinite-dimensional spaces. Such problems are ubiquitous in many areas of science and engineering, particularly in data science and imaging sciences. Our work is focused on relaxing the Lipschitz-smoothness assumptions generally required by first-order splitting algorithms by replacing the Euclidean energy with a Bregman divergence. These developments allow one to solve problems having more exotic geometry than that of the usual Euclidean setting. One algorithm is hybridization of the conditional gradient algorithm, making use of a linear minimization oracle at each iteration, with an augmented Lagrangian algorithm, allowing for affine constraints. The other algorithm is a primal-dual splitting algorithm incorporating Bregman divergences for computing the associated proximal operators. For both of these algorithms, our analysis shows convergence of the Lagrangian values, subsequential weak convergence of the iterates to solutions, and rates of convergence. In addition to these novel deterministic algorithms, we introduce and study also the stochastic extensions of these algorithms through a perturbation perspective. Our results in this part include almost sure convergence results for all the same quantities as in the deterministic setting, with rates as well. Finally, we tackle new problems that are only accessible through the relaxed assumptions our algorithms allow. We demonstrate numerical efficiency and verify our theoretical results on problems like low rank, sparse matrix completion, inverse problems on the simplex, and entropically regularized Wasserstein inverse problems

Uno de los ámbitos de mayor interés en el campo de la química analítica es el análisis medioambiental, ya que hay que asegurar y mantener la calidad, ya sea del aire o del agua para que su composición no comporte ningún peligro para la salud de los seres vivos. Los diferentes procesos industriales han contribuido a mejorar la calidad de vida, pero pueden producir subproductos que si son introducidos directa o indirectamente en el agua causan serios problemas de contaminación. De este modo, es necesario el tratamiento de una gran cantidad de desechos industriales, los cuales requieren una continua minimización.

Los avances en el campo del análisis medioambiental se han dirigido al desarrollo de nuevas técnicas que sean de fácil uso, que permitan minimizar la manipulación de la muestra, los costes y el tiempo de análisis, que no precisen del uso de disolventes orgánicos, que puedan ser fácilmente automatizadas y que sean rápidas. Entre las técnicas analíticas que presentan las características citadas se encuentran las técnicas de inyección de flujo.

La instrumentación analítica actual permite generar datos de diferente dimensionalidad: los denominados datos de orden cero, cuando la señal obtenida por el detector para cada muestra es un escalar y este valor se relaciona con la concentración mediante un calibrado univariante; los datos de primer orden son aquellos que corresponden a un vector por muestra, y se relacionan con la concentración mediante una calibración multivariante o de primer orden; y datos de segundo orden, que son aquellos que corresponden a una matriz por muestra y aplican una calibración de segundo orden para relacionar la señal con la concentración de los analitos de interés. Los datos de orden cero son útiles en los casos en los que se tiene una respuesta única y específica para el analito de interés, mientras que los datos de primer orden permiten determinar la concentración de un analito en presencia de interferentes siempre que éstos estén contemplados en las muestras de calibrado. En el caso de tener interferentes no definidos o que éstos no se puedan reproducir en las muestras de calibrado, se emplean datos de segundo orden y calibraciones de segundo orden.

Esta Tesis Doctoral se ha enmarcado en el desarrollo de nuevas metodologías analíticas para la determinación de subproductos de la industria de curtidos de pieles mediante un sistema de inyección de flujo, Análisis por inyección secuencial (SIA) y calibración de datos de segundo orden.

SIA, Sequential Injection Analysis, es un sistema de inyección de flujo, introducido por el profesor Jaromir Ruzicka el 1990, en el que la muestra y los reactivos son aspirados de forma secuencial, se mezclan por difusión en el reactor y se transportan mediante un flujo hacia el detector.

Un sistema SIA acoplado con un detector espectrofotométrico de diodos en fila (DAD) permite obtener datos de segundo orden, de forma que se recoge el espectro completo en un intervalo de longitudes de onda, a diferentes tiempos durante la elución del pico SIA. De esta forma se obtiene una matriz de datos (m x n), donde m son los diferentes tiempos y n las longitudes de onda de medida, para el análisis de cada una de las muestras.

En este trabajo, se presentan diferentes aplicaciones mediante análisis por inyección secuencial y calibración de datos de segundo orden con resolución de curvas multivariante con mínimos cuadrados alternados (MCR-ALS) para la determinación y especiación de cromo y para la determinación de tres colorantes simultáneamente. También se presentan dos revisiones bibliográficas críticas acerca de la determinación de cromo y del análisis multicomponente en sistemas de flujo. Además se desarrolló una aplicación de Cromatografía por Inyección Secuencial (SIC) con resolución de curvas multivariante con mínimos cuadrados alternados para la determinación de fenoles.

El cromo es un elemento ampliamente usado en esta industria. Su determinación tiene gran importancia medioambiental debido a la gran toxicidad de la especie de Cr(VI) como agente cancerígeno, mientras que el Cr(III) es un elemento esencial. En esta Tesis se presentan cuatro trabajos referidos a la determinación de cromo. En el primero, Use of multivariate curve resolution for determination of chromium in tanning samples using sequential injection analysis, Analytical and Bioanalytical Chemistry 382 (2005) 328-334 se basa en los fundamentos para permitir la determinación de cromo con el sistema SIA-MCR-ALS. Dependiendo de la capacidad de reacción del Cr(III) se puede diseñar una secuencia analítica en el sistema SIA de modo que se obtenga un sistema en evolución para obtener una matriz de datos como señal analítica. Cr(III) fue oxidado a Cr(VI) fuera del sistema SIA para aumentar la sensibilidad del Cr(III). De este modo, en el sistema SIA se produjo un gradiente de pH para provocar una conversión entre dos especies de Cr(VI), cromato y dicromato. El segundo artículo, Factorial design for optimizing chromium determination in tanning wastewater, Microchemical Journal 83 (2006) 98-104 consiste, tal y como el título sugiere, en la optimización del método desarrollado previamente para incorporar en una única etapa las dos etapas principales (oxidación de Cr(III) y evolución del Cr(VI)). El tercer trabajo, Chromium speciation using sequential injection analysis and multivariate curve resolution, Analytica Chimica Acta 571 (2006) 129-135 consiste en la determinación simultánea de Cr(III) y Cr(VI) debido a que la especiación de cromo es de gran interés, sobretodo en el campo medioambiental. Se llevó a cabo una derivatización previa de Cr(III), formando un complejo con ácido etilendiaminotetraacético (EDTA) para aumentar la sensibilidad del Cr(III). El sistema en evolución en el sistema SIA se generó mediante un gradiente de pH. El cuarto trabajo, Chromium determination and speciation since 2000, Trends in Analytical Chemistry, 25 (2006) 1006-1015 consiste en un trabajo bibliográfico que describe las posibilidades de determinación y especiación de cromo.

Otros analitos de interés en este ámbito son los colorantes de los baños residuales, debido a su elevada toxicidad y a su baja biodegradabilidad. Por tanto, tiene gran relevancia el estudio de la proporción de colorantes en las muestras y su efecto al largo del tiempo. Tanto el cromo como los colorantes son compuestos de interés en otros ámbitos como por ejemplo, alimentos, artes gráficas, etc. Se presentan cuatro trabajos científicos en este ámbito con dos objetivos prácticos. Por un lado, controlar la cantidad de colorante que permanece en solución después del proceso de teñido, de modo que se puedan optimizar algunas de las etapas del proceso, y por otro lado, estudiar diferentes estrategias para reducir el porcentaje de colorantes en aguas residuales. El primer trabajo Sequential injection analysis with second-order treatment for the determination of dyes in the exhaustion process of tanning effluents, Talanta 71 (2007) 1393-1398 consiste en el desarrollo de un método analítico basado en SIA y calibración de datos de segundo orden (MCR-ALS) para determinar simultáneamente tres colorantes en muestras de deshecho de la industria de curtidos. Este método se aplicó a muestras curtidas con sales de cromo o con especies vegetales. El segundo trabajo Matrix effect in second-order data. Determination of dyes in a tanning process using vegetable tanning agents, Analytica Chimica Acta 600 (2007) 233-239 conlleva la aplicación del método a muestras curtidas con especies vegetales, ya que éstas son especies altamente absorbentes y provocan el conocido efecto matriz, por lo que se propusieron estrategias para tratar de corregir estos efectos. Para alcanzar el segundo objetivo se consideró la adsorción de colorantes en carbón activo como una estrategia atractiva para eliminar los colorantes. El tercer trabajo, Kinetic and adsorption study of acid dye removal using activated carbon, Chemosphere 69 (2007) 1151-1158 estudia la adsorción y los parámetros cinéticos, tanto para los colorantes individuales como para las mezclas. El cuarto trabajo, Experimental designs for optimizing and fitting the adsorption of dyes onto activated carbon, submitted, pretende obtener una superficie de respuesta de forma que, dependiendo de las concentraciones de colorantes, se puedan establecer las condiciones experimentales para obtener una predeterminada adsorción de colorantes en carbón activo.

En la última etapa de la Tesis se exploraron diferentes posibilidades para aumentar la capacidad de los análisis simultáneos mediante sistemas de flujo y métodos de calibración de segundo orden. De este estudio se obtuvieron dos trabajos. El primero, Multicomponent analysis in flow systems, Trends in Analytical Chemistry, 26 (2007) 767-774 muestra una visión general de las determinaciones de múltiples analitos en sistemas de flujo. El segundo trabajo, Coupling of sequential injection chromatography with multivariate curve resolution alternating least squares for enhancement of peak capacity, Analytical Chemistry, 10.1021/ac071202h es el resultado de la combinación de dos estrategias propuestas para el análisis multicomponente, la cromatografía de inyección secuencial (SIC) y la calibración de datos de segundo orden con MCR-ALS. Este estudio se llevó a cabo en colaboración con el Grupo de Química Analítica, Automatización y Medioambiente de la Universidad de las Islas Baleares (UIB), en donde se realizó la parte experimental.
Environmental analysis is a field with high interest in the analytical chemistry community. It is very important to assure and maintain the quality, from air as well as from water, to avoid that its composition cause any risk for the living organisms. Different industrial processes have contributed to improve health quality, but they can produce intermediates that if are introduced direct or indirectly to the waters can cause important problems of pollution. In that way, it is necessary to treat a big amount of industrial wastes, which should be continuously minimized.

Advances in the environmental field have been focusing on the development of new techniques of easy use, with low sample manipulation, low costs and short analysis times that could be easily automatised and fast. Flow analysis techniques can be found in this classification.

The current analytical instruments can generate data with different dimensionality. The analytical signal can be a scalar (a single absorbance measure), a vector (single absorbance measurement along time), or a data matrix (spectrum recorded along time) per each analysed sample. These data have been classified, as zero-order data when the signal is a scalar, first-order data when is a vector, and second-order data when the signal is a matrix. Zero-order data are useful for cases in which we have a unique and specific response for the analyte of interest, meanwhile first-order data allow quantification of an analyte in the presence of interferents but they should be contained in the calibration samples. When the interferents are not known and they could not be present in the calibration samples, second-order data and second-order calibrations are used.

This doctoral thesis has been carried out for developing new analytical methodologies for determining subproducts of the tanning industry using sequential injection analysis (SIA) and second-order calibration.

SIA, Sequential Injection Analysis, is a flow injection system introduced in 1990 by Professor Jaromir Ruzicka, where the sample and reagents are introduced sequentially into the system, are mixed by diffusion process in the reactor and then they are pumped through the detector.
A SIA system using a diode array spectrophotometer can generate second-order data, it means that the analytical signal is a data matrix for each sample on which we have absorbances in a wavelength interval in one axis and in a time interval in the other axis.

In this project, we present different practical applications using sequential injection analysis and second-order calibration using multivariate curve resolution alternating least squares (MCR-ALS) for determining and speciating chromium and for determining three acid dyes simultaneously. Furthermore we present two bibliographic overviews about chromium determination and multicomponent analysis in flow systems. We also developed an application of sequential injection chromatography (SIC) using multivariate curve resolution alternating least squares (MCR-ALS) for determining phenolic derivatives.

Chromium is an element widely used in the industry. Its determination presents high environmental interest because the toxicity of the Cr(VI) species as a carcinogenic agent, meanwhile Cr(III) is an essential element. In this Thesis we present four papers relative to chromium determination. The first paper Use of multivariate curve resolution for determination of chromium in tanning samples using sequential injection analysis, Analytical and Bioanalytical Chemistry 382 (2005) 328-334 is based on the basis to permit quantification of chromium with the system SIA-MCR-ALS. Depending on the capacity of reaction of Cr(III), different analytical sequences in the SIA system can be designed. Cr(III) was oxidized into Cr(VI) outside of the SIA system to increase Cr(III) sensibility. In that way, in the SIA system we induced a pH gradient to see the conversion of the two species of Cr(VI), chromate and dichromate. The second paper Factorial design for optimizing chromium determination in tanning wastewater, Microchemical Journal 83 (2006) 98-104 presents an automatic system for total chromium determination using different experimental designs to optimize the overall process. The third paper, Chromium speciation using sequential injection analysis and multivariate curve resolution, Analytica Chimica Acta 571 (2006) 129-135 is a paper in which the two main species of chromium, Cr(III) and Cr(VI), are determined simultaneously in a single analysis in the presence of interferents. The fourth paper, Chromium determinations and speciation since 2000, Trends in Analytical Chemistry, 25 (2006) 1006-1015 is a bibliographic study that describes the possibilities available for chromium determination and speciation.

Other analytes of interest in this field are dyes from wastewater due to its high toxicity and its low biodegradability. That is why studying the proportion of dyes in samples and its effect through time presents high relevance. Chromium and dyes are compounds of high interest in other application fields, such as, foods, printing and graphic design, etc. We present four scientific papers in this field with two practical objectives. On the one hand, control the amount of dyes that remains in solution after the dyeing process has been done. On the other hand, study different strategies to reduce the percentage of dyes in wastewater. The first paper, Sequential injection analysis with second-order treatment for the determination of dyes in the exhaustion process of tanning effluents Talanta 71 (2007) 1393-1398, describes the developed method for determining three acid dyes in a single step. This method was applied to water samples of leathers tanned with chromium salts and with vegetal agents. The second paper Matrix effect in second-order data. Determination of dyes in a tanning process using vegetable tanning agents Analytica Chimica Acta 600 (2007) 233-239 presents strategies to use when a sample presents matrix effects using second-order data. To solve the second objective, we studied the behaviour of dyes on activated carbon and present the results in the third paper Kinetic and adsorption study of acid dye removal using activated carbon Chemosphere 69 (2007) 1151-1158 where we studied the adsorption and the kinetic parameters of dyes being alone in solution or in a mixture of them. The fourth paper, Experimental designs for optimizing and fitting the adsorption of dyes onto activated carbon Submitted, we described a sequential methodology to obtain a response surface and optimize the adsorption process as a way of eliminating dyes in wastewater samples from the tanning industry.

In the last period of the Thesis, we explored different strategies to increase the capacity of simultaneous analysis using flow systems and second-order calibration. From this study, we obtained two papers. The first, Multicomponent analysis in flow systems, Trends in Analytical Chemistry, 26 (2007) 767-774, shows a general vision of multiple determinations in flow systems. The second paper, Coupling of sequential injection chromatography with multivariate curve resolution alternating least squares for enhancement of peak capacity, Analytical Chemistry 79 (2007) 7767-7774 shows a combination of two strategies proposed for multicomponent analysis, sequential injection chromatography (SIC) and second-order calibration with MCR-ALS. This study was carried out in collaboration with the Analytical Chemistry, Automation and Environment group of the University of the Balearic Islands.

In this dissertation, we consider modified inexact Newton methods applied to second order nonlinear problems. In the implementation of Newton's method applied to problems with a large number of degrees of freedom, it is often necessary to solve the linear Jacobian system iteratively. Although a general theory for the convergence of modified inexact Newton's methods has been developed, its application to nonlinear problems from nonlinear PDE's is far from complete. The case where the nonlinear operator is a zeroth order perturbation of a fixed linear operator was considered in the paper written by Brown et al.. The goal of this dissertation is to show that one can develop modified inexact Newton's methods which converge at a rate independent of the number of unknowns for problems with higher order nonlinearities. To do this, we are required to first, set up the problem on a scale of Hilbert spaces, and second, to devise a special iterative technique which converges in a higher order Sobolev norm, i.e., H1+alpha(omega) \ H1 0(omega) with 0 < alpha < 1/2. We show that the linear system solved in Newton's method can be replaced with one iterative step provided that the initial iterate is close enough. The closeness criteria can be taken independent of the mesh size. In addition, we have the same convergence rates of the method in the norm of H1 0(omega) using the discrete Sobolev inequalities.

We derive and analyse two families of multistep collocation methods for periodic initial-value problems of the form y" = f(x, y); y((^x)o) = yo, y(^1)(xo) = zo involving ordinary differential equations of second order in which the first derivative does not appear explicitly. A survey of recent results and proposed numerical methods is given in chapter 2. Chapter 3 is devoted to the analysis of a family of implicit Chebyshev methods proposed by Panovsky k Richardson. We show that for each non-negative integer r, there are two methods of order 2r from this family which possess non-vanishing intervals of periodicity. The equivalence of these methods with one-step collocation methods is also established, and these methods are shown to be neither P-stable nor symplectic. In chapters 4 and 5, two families of multistep collocation methods are derived, and their order and stability properties are investigated. A detailed analysis of the two-step symmetric methods from each class is also given. The multistep Runge-Kutta-Nystrom methods of chapter 4 are found to be difficult to analyse, and the specific examples considered are found to perform poorly in the areas of both accuracy and stability. By contrast, the two-step symmetric hybrid methods of chapter 5 are shown to have excellent stability properties, in particular we show that all two-step 27V-point methods of this type possess non-vanishing intervals of periodicity, and we give conditions under which these methods are almost P-stable. P-stable and efficient methods from this family are obtained and demonstrated in numerical experiments. A simple, cheap and effective error estimator for these methods is also given.

2010 - 2011
The aim of this research is the construction and the analysis of new families of numerical methods for the integration of special second order Ordinary Differential Equations (ODEs). The modeling of continuous time dynamical systems using second order ODEs is widely used in many elds of applications, as celestial mechanics, seismology, molecular dynamics, or in the semidiscretisation of partial differential equations (which leads to high dimensional systems and stiffness). Although the numerical treatment of this problem has been widely discussed in the literature, the interest in this area is still vivid, because such equations generally exhibit typical problems (e.g. stiffness, metastability, periodicity, high oscillations), which must efficiently be overcome by using suitable numerical integrators. The purpose of this research is twofold: on the one hand to construct a general family of numerical methods for special second order ODEs of the type y00 = f(y(t)), in order to provide an unifying approach for the analysis of the properties of consistency, zero-stability and convergence; on the other hand to derive special purpose methods, that follow the oscillatory or periodic behaviour of the solution of the problem...[edited by author]
X n. s.

Consider a time-dependent optimal control problem, where the state evolution is described by an initial value problem. There are a variety of numerical methods to solve these problems. The so-called indirect approach is considered detailed in this thesis. The indirect methods solve decoupled boundary value problems resulting from the necessary conditions for the optimal control problem. The so-called Pantoja method describes a computationally efficient stage-wise construction of the Newton direction for the discrete-time optimal control problem. There are many relationships between multiple shooting techniques and Pantoja method, which are investigated in this thesis. In this context, the equivalence of Pantoja method and multiple shooting method of Riccati type is shown. Moreover, Pantoja method is extended to the case where the state equations are discretised using one of implicit numerical methods. Furthermore, the concept of symplecticness and Hamiltonian systems is introduced. In this regard, a suitable numerical method is presented, which can be applied to unconstrained optimal control problems. It is proved that this method is a symplectic one. The iterative solution of optimal control problems in ordinary differential equations by Pantoja or Riccati equivalent methods leads to a succession of triple sweeps through the discretised time interval. The second (adjoint) sweep relies on information from the first (original) sweep, and the third (final) sweep depends on both of them. Typically, the steps on the adjoint sweep involve more operations and require more storage than the other two. The key difficulty is given by the enormous amount of memory required for the implementation of these methods if all states throughout forward and adjoint sweeps are stored. One of goals of this thesis is to present checkpointing techniques for memory reduced implementation of these methods. For this purpose, the well known aspect of checkpointing has to be extended to a `nested checkpointing` for multiple transversals. The proposed nested reversal schedules drastically reduce the required spatial complexity. The schedules are designed to minimise the overall execution time given a certain total amount of storage for the checkpoints. The proposed scheduling schemes are applied to the memory reduced implementation of the optimal control problem of laser surface hardening and other optimal control problems
Es wird ein Problem der optimalen Steuerung betrachtet. Die dazugehoerigen Zustandsgleichungen sind mit einer Anfangswertaufgabe definiert. Es existieren zahlreiche numerische Methoden, um Probleme der optimalen Steuerung zu loesen. Der so genannte indirekte Ansatz wird in diesen Thesen detailliert betrachtet. Die indirekten Methoden loesen das aus den Notwendigkeitsbedingungen resultierende Randwertproblem. Das so genannte Pantoja Verfahren beschreibt eine zeiteffiziente schrittweise Berechnung der Newton Richtung fuer diskrete Probleme der optimalen Steuerung. Es gibt mehrere Beziehungen zwischen den unterschiedlichen Mehrzielmethoden und dem Pantoja Verfahren, die in diesen Thesen detailliert zu untersuchen sind. In diesem Zusammenhang wird die aequivalence zwischen dem Pantoja Verfahren und der Mehrzielmethode vom Riccati Typ gezeigt. Ausserdem wird das herkoemlige Pantoja Verfahren dahingehend erweitert, dass die Zustandsgleichungen mit Hilfe einer impliziten numerischen Methode diskretisiert sind. Weiterhin wird das Symplektische Konzept eingefuehrt. In diesem Zusammenhang wird eine geeignete numerische Methode praesentiert, die fuer ein unrestringiertes Problem der optimalen Steuerung angewendet werden kann. In diesen Thesen wird bewiesen, dass diese Methode symplectisch ist. Das iterative Loesen eines Problems der optimalen Steuerung in gewoenlichen Differentialgleichungen mit Hilfe von Pantoja oder Riccati aequivalenten Verfahren fuehrt auf eine Aufeinanderfolge der Durchlaeufetripeln in einem diskretisierten Zeitintervall. Der zweite (adjungierte) Lauf haengt von der Information des ersten (primalen) Laufes, und der dritte (finale) Lauf haeng von den beiden vorherigen ab. Ueblicherweise beinhalten Schritte und Zustaende des adjungierten Laufes wesentlich mehr Operationen und benoetigen auch wesentlich mehr Speicherplatzkapazitaet als Schritte und Zustaende der anderen zwei Durchlaeufe. Das Grundproblem besteht in einer enormen Speicherplatzkapazitaet, die fuer die Implementierung dieser Methoden benutzt wird, falls alle Zustaende des primalen und des adjungierten Durchlaufes zu speichern sind. Ein Ziel dieser Thesen besteht darin, Checkpointing Strategien zu praesentieren, um diese Methoden speichereffizient zu implementieren. Diese geschachtelten Umkehrschemata sind so konstruiert, dass fuer einen gegebenen Speicherplatz die gesamte Laufzeit zur Abarbeitung des Umkehrschemas minimiert wird. Die aufgestellten Umkehrschemata wurden fuer eine speichereffiziente Implementierung von Problemen der optimalen Steuerung angewendet. Insbesondere betrifft dies das Problem einer Oberflaechenabhaertung mit Laserbehandlung

In this thesis we analyse the high-order in time discontinuous Galerkin nite element method (DGFEM) for second-order in time linear abstract wave equations. Our abstract approximation analysis is a generalisation of the approach introduced by Claes Johnson (in Comput. Methods Appl. Mech. Engrg., 107:117-129, 1993), writing the second order problem as a system of fi rst order problems. We consider abstract spatial (time independent) operators, highorder in time basis functions when discretising in time; we also prove approximation results in case of linear constraints, e.g. non-homogeneous boundary data. We take the two steps approximation approach i.e. using high-order in time DGFEM; the discretisation approach in time introduced by D Schötzau (PhD thesis, Swiss Federal institute of technology, Zürich, 1999) to fi rst obtain the semidiscrete scheme and then conformal spatial discretisation to obtain the fully-discrete formulation. We have shown solvability, unconditional stability and conditional a priori error estimates within our abstract framework for the fully discretized problem. The skew-symmetric spatial forms arising in our abstract framework for the semi- and fully-discrete schemes do not full ll the underlying assumptions in D. Schötzau's work. But the semi-discrete and fully discrete forms satisfy an Inf-sup condition, essential for our proofs; in this sense our approach is also a generalisation of D. Schötzau's work. All estimates are given in a norm in space and time which is weaker than the Hilbert norm belonging to our abstract function spaces, a typical complication in evolution problems. To the best of the author's knowledge, with the approximation approach we used, these stability and a priori error estimates with their abstract structure have not been shown before for the abstract variational formulation used in this thesis. Finally we apply our abstract framework to the acoustic and an elasto-dynamic linear equations with non-homogeneous Dirichlet boundary data.

La thèse porte sur les schémas numériques pour les problèmes de décisions Markoviennes (MDPs), les équations aux dérivées partielles (EDPs), les équations différentielles stochastiques rétrogrades (ED- SRs), ainsi que les équations différentielles stochastiques rétrogrades réfléchies (EDSRs réfléchies). La thèse se divise en trois parties.La première partie porte sur des méthodes numériques pour résoudre les MDPs, à base de quan- tification et de régression locale ou globale. Un problème de market-making est proposé: il est résolu théoriquement en le réécrivant comme un MDP; et numériquement en utilisant le nouvel algorithme. Dans un second temps, une méthode de Markovian embedding est proposée pour réduire des prob- lèmes de type McKean-Vlasov avec information partielle à des MDPs. Cette méthode est mise en œuvre sur trois différents problèmes de type McKean-Vlasov avec information partielle, qui sont par la suite numériquement résolus en utilisant des méthodes numériques à base de régression et de quantification.Dans la seconde partie, on propose de nouveaux algorithmes pour résoudre les MDPs en grande dimension. Ces derniers reposent sur les réseaux de neurones, qui ont prouvé en pratique être les meilleurs pour apprendre des fonctions en grande dimension. La consistance des algorithmes proposés est prouvée, et ces derniers sont testés sur de nombreux problèmes de contrôle stochastique, ce qui permet d’illustrer leurs performances.Dans la troisième partie, on s’intéresse à des méthodes basées sur les réseaux de neurones pour résoudre les EDPs, EDSRs et EDSRs réfléchies. La convergence des algorithmes proposés est prouvée; et ces derniers sont comparés à d’autres algorithmes récents de la littérature sur quelques exemples, ce qui permet d’illustrer leurs très bonnes performances
The present thesis deals with numerical schemes to solve Markov Decision Problems (MDPs), partial differential equations (PDEs), quasi-variational inequalities (QVIs), backward stochastic differential equations (BSDEs) and reflected backward stochastic differential equations (RBSDEs). The thesis is divided into three parts.The first part focuses on methods based on quantization, local regression and global regression to solve MDPs. Firstly, we present a new algorithm, named Qknn, and study its consistency. A time-continuous control problem of market-making is then presented, which is theoretically solved by reducing the problem to a MDP, and whose optimal control is accurately approximated by Qknn. Then, a method based on Markovian embedding is presented to reduce McKean-Vlasov control prob- lem with partial information to standard MDP. This method is applied to three different McKean- Vlasov control problems with partial information. The method and high accuracy of Qknn is validated by comparing the performance of the latter with some finite difference-based algorithms and some global regression-based algorithm such as regress-now and regress-later.In the second part of the thesis, we propose new algorithms to solve MDPs in high-dimension. Neural networks, combined with gradient-descent methods, have been empirically proved to be the best at learning complex functions in high-dimension, thus, leading us to base our new algorithms on them. We derived the theoretical rates of convergence of the proposed new algorithms, and tested them on several relevant applications.In the third part of the thesis, we propose a numerical scheme for PDEs, QVIs, BSDEs, and RBSDEs. We analyze the performance of our new algorithms, and compare them to other ones available in the literature (including the recent one proposed in [EHJ17]) on several tests, which illustrates the efficiency of our methods to estimate complex solutions in high-dimension.Keywords: Deep learning, neural networks, Stochastic control, Markov Decision Process, non- linear PDEs, QVIs, optimal stopping problem BSDEs, RBSDEs, McKean-Vlasov control, perfor- mance iteration, value iteration, hybrid iteration, global regression, local regression, regress-later, quantization, limit order book, pure-jump controlled process, algorithmic-trading, market-making, high-dimension

Thesis (Ph.D)--Industrial and Systems Engineering, Georgia Institute of Technology, 2009.
Committee Chair: Arkadi Nemirovski; Committee Co-Chair: Alexander Shapiro; Committee Co-Chair: Renato D. C. Monteiro; Committee Member: Anatoli Jouditski; Committee Member: Shabbir Ahmed. Part of the SMARTech Electronic Thesis and Dissertation Collection.

This dissertation aims to develop efficient algorithms with improved scalability and stability properties for large-scale optimization and optimization under uncertainty, and to bridge some of the gaps between modern optimization theories and recent applications emerging in the Big Data environment. To this end, the dissertation is dedicated to two important subjects -- i) Large-scale Convex Composite Optimization and ii) Error-in-Measurement Optimization. In spite of the different natures of these two topics, the common denominator, to be presented, lies in their accommodation for systematic use of saddle point techniques for mathematical modeling and numerical processing. The main body can be split into three parts. In the first part, we consider a broad class of variational inequalities with composite structures, allowing to cover the saddle point/variational analogies of the classical convex composite minimization (i.e. summation of a smooth convex function and a simple nonsmooth convex function). We develop novel composite versions of the state-of-the-art Mirror Descent and Mirror Prox algorithms aimed at solving such type of problems. We demonstrate that the algorithms inherit the favorable efficiency estimate of their prototypes when solving structured variational inequalities. Moreover, we develop several variants of the composite Mirror Prox algorithm along with their corresponding complexity bounds, allowing the algorithm to handle the case of imprecise prox mapping as well as the case when the operator is represented by an unbiased stochastic oracle. In the second part, we investigate four general types of large-scale convex composite optimization problems, including (a) multi-term composite minimization, (b) linearly constrained composite minimization, (c) norm-regularized nonsmooth minimization, and (d) maximum likelihood Poisson imaging. We demonstrate that the composite Mirror Prox, when integrated with saddle point techniques and other algorithmic tools, can solve all these optimization problems with the best known so far rates of convergences. Our main related contributions are as follows. Firstly, regards to problems of type (a), we develop an optimal algorithm by integrating the composite Mirror Prox with a saddle point reformulation based on exact penalty. Secondly, regards to problems of type (b), we develop a novel algorithm reducing the problem to solving a ``small series'' of saddle point subproblems and achieving an optimal, up to log factors, complexity bound. Thirdly, regards to problems of type (c), we develop a Semi-Proximal Mirror-Prox algorithm by leveraging the saddle point representation and linear minimization over problems' domain and attain optimality both in the numbers of calls to the first order oracle representing the objective and calls to the linear minimization oracle representing problem's domain. Lastly, regards to problem (d), we show that the composite Mirror Prox when applied to the saddle point reformulation circumvents the difficulty with non-Lipschitz continuity of the objective and exhibits better convergence rate than the typical rate for nonsmooth optimization. We conduct extensive numerical experiments and illustrate the practical potential of our algorithms in a wide spectrum of applications in machine learning and image processing. In the third part, we examine error-in-measurement optimization, referring to decision-making problems with data subject to measurement errors; such problems arise naturally in a number of important applications, such as privacy learning, signal processing, and portfolio selection. Due to the postulated observation scheme and specific structure of the problem, straightforward application of standard stochastic optimization techniques such as Stochastic Approximation (SA) and Sample Average Approximation (SAA) are out of question. Our goal is to develop computationally efficient and, hopefully, not too conservative data-driven techniques applicable to a broad scope of problems and allowing for theoretical performance guarantees. We present two such approaches -- one depending on a fully algorithmic calculus of saddle point representations of convex-concave functions and the other depending on a general approximation scheme of convex stochastic programming. Both approaches allow us to convert the problem of interests to a form amenable for SA or SAA. The latter developments are primarily focused on two important applications -- affine signal processing and indirect support vector machines.

Asset and Liability Management (ALM) models have become well established decision tools for pension funds. ALMs are commonly modelled as multi-stage, in which a large terminal wealth is required, while at intermediate time periods, constraints on the funding ratio, that is, the ratio of assets to liabilities, are imposed. Underfunding occurs when the funding ratio is too low; a target value for funding ratios is pre-specified by the decision maker. The risk of underfunding has been usually modelled by employing established risk measures; this controls one single aspect of the funding ratio distributions. For example, controlling the expected shortfall below the target has limited power in controlling shortfall under worst-case scenarios. We propose ALM models in which the risk of underfunding is modelled based on the concept of Second Order Stochastic Dominance (SSD). This is a criterion of ranking random variables - in our case funding ratios - that takes the entire distributions of interest into account and works under the widely accepted assumptions of decision makers being rational and risk averse. In the proposed SSD models, investment decisions are taken such that the resulting short-term distribution of the funding ratio is non-dominated with respect to SSD, while a constraint is imposed on the expected terminal wealth. This is done by considering progressively larger tails of the funding ratio distribution and considering target levels for them; a target distribution is thus implied. Different target distributions lead to different SSD efficient solutions. Improved distributions of funding ratios may be thus achieved, compared to the existing risk models for ALM. This is the first contribution of this thesis. Interesting results are obtained in the special case when the target distribution is deterministic, specified by one single outcome. In this case, we can obtain equivalent risk minimisation models, with risk defined as expected shortfall or as worst case loss. This represents the second contribution. The third contribution is a framework for scenario generation based on the "Birth, Immigration, Death, Emigration" (BIDE) population model and the Empirical copula; the scenarios are used to evaluate the proposed models and their special cases both in-sample and out-of-sample. As an application, we consider the planning problem of a large DB pension fund in Saudi Arabia.

The objective of this thesis, which in clearly inspired by an industrial framework, is to try and narrow the gap between theoretical neutron modelling and application in the context of nuclear reactor design. This thesis is divided into three main chapters, preceded by a general overview. This structure reflects the three main topics which were chosen for this research project. The first topic develops the Spectral Element Method (SEM) approach and its use in conjunction with transport approximations. As it is documented in the specialized numerical analysis handbooks and in previous works by the author, the method has an excellent convergence rate which outperforms most classical schemes, but it has also some important drawbacks which sometimes seem to discourage its use for linear transport problems applied to nontrivial benchmarks. In order to elaborate the methodology of the specific problems encountered in reactor physics, three aspects are addressed looking for improvements. The first topic analyzed is related to the convergence order, whose value is less straightforward to define a priori by means of functional analysis than other numerical schemes. The adjective “spectral” refers in fact to the maximum order claimed, exponential with respect to the average size of the mesh. A comprehensive set of convergence tests is carried out applying SEM to a few transport models and with the aid of manufactured solutions, thus isolating the numerical effects from the deviations which are due only to modelling approximations. The hypothesis of grid conformity is also relaxed, replacing the classical Galerkin variational formulation with the Discontinuous Galerkin theory, characterized by a more flexible treatment of the mesh interfaces; this scheme allows local grid refinement and opens the way, in perspective, to mesh adaption. Finally, a simple and sufficiently flexible technique to deform the boundaries of each mesh is introduced and applied, in order to adapt the grid to curved geometries. In this way, the advantages of SEM can be applied to a vast class of common problems like lattice calculations. Moreover, thanks to a change of the basis functions used in SEM, it is possible to obtain elements with three sides (straight or deformed), that are a typical war horse of the Finite Element approach. The second topic is essentially devoted to the most “industrial” part of the thesis, developed entirely during the stay of the author in the AREVA NP headquarters in Paris. In AREVA, and in all other nuclear engineering enterprises, neutron diffusion is still the preferred neutronic model for full-core studies. Better approximations are reserved for library preparation, fuel studies and code validation, none of these being typically too much time or budget-constrained. Today needs start to require a certain level of improvement also in full-core analyses, trying to fitly model localized dis-homogeneities and reduce the penalizing engineering margins which are taken as provisions. On the other hand, a change in the model does not mean only an effort to write a new code, but has huge follow-ups due to the validation processes required by the authorities. Second-order transport may support the foreseen methodology update because it can be implemented re-using diffusion routines as the computational engine. The AN method, a second-order approximation of the transport equation, has been introduced in some studies, and its effect is discussed. Moreover, some effort has been reserved to the introduction of linear anisotropy in the model. The last topic deals with ray effects; they are a known issue of the discrete ordinate approach (SN methods) which is responsible for a reduction in the accuracy of the solution, especially in penetration problems with low scattering, like several shielding calculations performed for operator safety concerns. Ray effects are here characterized from a formal point of view in both static and time dependent situations. Then, quantitative indicators are defined to help with the interpretation of the SN results. Based on these studies, some mitigation measures are proposed and their efficacy is discussed.

L'approche de la biologie des systèmes vise à intégrer les méthodologies appliquées dans la conception et l'analyse des systèmes technologiques complexes, au sein de la biologie afin de comprendre les principes de fonctionnement globaux des systèmes biologiques. La thèse s'inscrit dans le cadre de la biologie des systèmes et en particulier dans la prolongation d'une méthode issue de ce cadre : la méthode Resource Blance Analysis (RBA). Nous visons dans cette thèse à augmenter le pouvoir prédictif de la méthode via un travail de modélisation tout en gardant un bon compromis entre représentativité des modèles issus de ce cadre et leur résolution numérique efficace. La thèse se décompose en deux grandes parties : la première vise à intégrer les aspects thermodynamiques et cinétiques inhérents aux réseaux métaboliques. La deuxième vise à comprendre l'impact de l'aspect stochastique de la production des enzymes sur le croissance de la bactérie. Des méthodes numériques ont été élaborées pour la résolution des modèles ainsi établis dans les deux cas déterministe et stochastique
In order to understand the global functioning principals of biological systems, system bio- logy approach aims to integrate the methodologies used in the conception and the analysis of complex technological systems, within the biology. This PhD thesis fits into the system biology framework and in particular the extension of the already existing method Resource Balance Analysis (RBA). We aim in this PhD thesis to improve the predictive power of this method by introducing more complex model. However, this new model should respect a good trade-off between the representativity of the model and its efficient numerical computation. This PhD thesis is decomposed into two major parts. The first part aims the integration of the metabolic network inherent thermodynamical and kinetic aspects. The second part aims the comprehension of the impact of enzyme production stochastic aspect on the bacteria growth. Numerical methods are elaborated to solve the obtained models in both deterministic and stochastic cases

Modelling and simulation are essential to modern research in cell biology. This thesis follows a journey starting from the construction of new stochastic methods for discrete biochemical systems to using them to simulate a population of interacting haematopoietic stem cell lineages. The first part of this thesis is on discrete stochastic methods. We develop two new methods, the stochastic extrapolation framework and the Stochastic Bulirsch-Stoer methods. These are based on the Richardson extrapolation technique, which is widely used in ordinary differential equation solvers. We believed that it would also be useful in the stochastic regime, and this turned out to be true. The stochastic extrapolation framework is a scheme that admits any stochastic method with a fixed stepsize and known global error expansion. It can improve the weak order of the moments of these methods by cancelling the leading terms in the global error. Using numerical simulations, we demonstrate that this is the case up to second order, and postulate that this also follows for higher order. Our simulations show that extrapolation can greatly improve the accuracy of a numerical method. The Stochastic Bulirsch-Stoer method is another highly accurate stochastic solver. Furthermore, using numerical simulations we find that it is able to better retain its high accuracy for larger timesteps than competing methods, meaning it remains accurate even when simulation time is speeded up. This is a useful property for simulating the complex systems that researchers are often interested in today. The second part of the thesis is concerned with modelling a haematopoietic stem cell system, which consists of many interacting niche lineages. We use a vectorised tau-leap method to examine the differences between a deterministic and a stochastic model of the system, and investigate how coupling niche lineages affects the dynamics of the system at the homeostatic state as well as after a perturbation. We find that larger coupling allows the system to find the optimal steady state blood cell levels. In addition, when the perturbation is applied randomly to the entire system, larger coupling also results in smaller post-perturbation cell fluctuations compared to non-coupled cells. In brief, this thesis contains four main sets of contributions: two new high-accuracy discrete stochastic methods that have been numerically tested, an improvement that can be used with any leaping method that introduces vectorisation as well as how to use a common stepsize adapting scheme, and an investigation of the effects of coupling lineages in a heterogeneous population of haematopoietic stem cell niche lineages.

In this thesis a new approach for solving a certain class of anomalous diffusion equations was developed. The theory and algorithms arising from this work will pave the way for more efficient and more accurate solutions of these equations, with applications to science, health and industry. The method of finite volumes was applied to discretise the spatial derivatives, and this was shown to outperform existing methods in several key respects. The stability and convergence of the new method were rigorously established.

Orientador: Ronei Jesus Poppi
Dissertação (mestrado) - Universidade Estadual de Campinas, Instituto de Quimica
Made available in DSpace on 2018-08-08T23:32:47Z (GMT). No. of bitstreams: 1 Carneiro_RenatoLajarim_M.pdf: 4176371 bytes, checksum: cbe2edc08ad07ea0e4607e69fc38aec5 (MD5) Previous issue date: 2007
Resumo: Esse trabalho teve por objetivo desenvolver um programa em MatLab baseado no Algoritmo Genético (GA) para aplicar e verificar as principais vantagens deste na seleção de variáveis para métodos de calibração de segunda ordem (BLLS-RBL, PARAFAC e N-PLS). Para esta finalidade foram utilizados três conjuntos de dados: 1. Determinação de pesticidas e um metabólito em vinho tinto por HPLC-DAD em três situações distintas. Nestas três situações foram observadas sobreposições dos interferentes sobre os compostos de interesse. Estes compostos eram os pesticidas carbaril (CBL), tiofanato metílico (TIO), simazina (SIM) e dimetoato (DMT) e o metabólito ftalimida (PTA). 2. Quantificação das vitaminas B2 (riboflavina) e B6 (piridoxina) por espectrofluorimetria de excitação/emissão em formulações infantis comerciais, sendo três leites em pó e dois suplementos alimentares. 3. Análise dos fármacos ácido ascórbico (AA) e ácido acetilsalicílico (AAS) em formulações farmacêuticas por FIA com gradiente de pH e detecção por arranjo de diodos, onde a variação de pH causa alteração na estrutura das moléculas dos fármacos mudando seus espectros na região do ultravioleta. A performance dos modelos, com e sem seleção de variáveis, foi comparada através de seus erros, expressados como a raiz quadrada da média dos quadrados dos erros de previsão (RMSEP), e os erros relativos de previsão (REP). Resultados melhores foram claramente observados quando o GA foi utilizado para a seleção de variáveis nos métodos de calibração de segunda ordem.
Abstract: The aim of this work was to develop a program in MatLab using Genetic Algorithm (GA) to apply and to verify the main advantages of variables selection for second-order calibration methods (BLLS-RBL, PARAFAC and N-PLS). For this purpose three data sets had been used: 1. Determination of pesticides and a metabolite in red wines using HPLC-DAD in three distinct situations, where overlappings of the interferentes on interest compounds are observed. These composites were the pesticides carbaryl (CBL), methyl thiophanate (TIO), simazine (SIM) and dimethoate (DMT) and the metabolite phthalimide (PTA). 2. Quantification of the B2 (riboflavine) and (pyridoxine) B6 vitamins for spectrofluorimetry of excitation-emission in commercial infantile products, being three powder milk and two supplement foods. 3. Analysis of ascorbic acid (AA) and acetylsalicylic acid (AAS) in pharmaceutical tablets by FIA with pH gradient and detection for diode array, where the variation of pH causes alterations in the structure of molecules of analites shifting its spectra in the region of the ultraviolet. The performance of the models, with and without selection of variable, was compared through its errors, expressed as the root mean square error of prediction (RMSEP), and the relative errors of prediction (REP). The best results were obtained when the GA was used for the selection of variable in second-order calibration methods.
Mestrado
Quimica Analitica
Mestre em Química

Orientador: Jose Roberto Sanches Mantovani
Resumo: Neste trabalho são propostos modelos matemáticos determinístico e estocástico de programação cônica de segunda ordem em coordenadas retangulares para o problema de fluxo de potência ótimo de geração e controle de potência reativa no sistemas elétricos de potência, considerando as minimização dos custos de geração de energia, perdas ativas da rede e emissão de poluentes no meio ambiente. Os modelos contemplam as principais características físicas e econômicas do problema estudado, assim como os limites operacionais do sistema elétrico. Os modelos são programados em linguagem AMPL e suas soluções são obtidas através do solver comercial CPLEX. Os sistemas testes IEEE30, IEEE118 e ACTIVSg200 são utilizados nas simulações computacionais dos modelos propostos. Os resultados obtidos pelo modelo determinístico desenvolvido são validados através de comparações com os resultados fornecidos pelo software MATPOWER , onde ambos consideram apenas a existência de gerações termoelétricas. No modelo estocástico utiliza-se a técnica de geração de cenários e considera-se um período de um ano (8760 horas), e geradores que utilizam fontes de geração renováveis e não renováveis.
Abstract: In this work we propose deterministic and stochastic mathematical models of second order conical programming in rectangular coordinates for the optimal power flow problem of reactive power generation and control in electric power systems, considering the minimization of energy generation costs, losses networks and emission of pollutants into the environment. The models contemplate the main physical and economic characteristics of the studied problem, as well as the operational limits of the electric system. The models are programmed in AMPL language and their solutions are obtained through the commercial solver CPLEX. The IEEE30, IEEE118 and ACTIVSg200 test systems are used in the computer simulations of the proposed models. The results obtained by the deterministic model developed are validated through comparisons with the results provided by the software MATPOWERR , where both consider only the existence of thermoelectric generations. The stochastic model uses the scenario generation technique and considers a period of one year (8760 hours), and generators using renewable and non-renewable generation sources.
Mestre

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

Les modèles de Markov d’ordre 2 (HMM2) sont des modèles stochastiques qui ont démontré leur efficacité dans l’exploration de séquences génomiques. Cette thèse explore l’intérêt de modèles de différents types (M1M2, M2M2, M2M0) ainsi que leur couplage à des méthodes combinatoires pour segmenter les génomes bactériens sans connaissances a priori du contenu génétique. Ces approches ont été appliquées à deux modèles bactériens afin d’en valider la robustesse : Streptomyces coelicolor et Streptococcus thermophilus. Ces espèces bactériennes présentent des caractéristiques génomiques très distinctes (composition, taille du génome) en lien avec leur écosystème spécifique : le sol pour les S. coelicolor et le milieu lait pour S. thermophilus
Second-order Hidden Markov Models (HMM2) are stochastic processes with a high efficiency in exploring bacterial genome sequences. Different types of HMM2 (M1M2, M2M2, M2M0) combined to combinatorial methods were developed in a new approach to discriminate genomic regions without a priori knowledge on their genetic content. This approach was applied on two bacterial models in order to validate its achievements: Streptomyces coelicolor and Streptococcus thermophilus. These bacterial species exhibit distinct genomic traits (base composition, global genome size) in relation with their ecological niche: soil for S. coelicolor and dairy products for S. thermophilus. In S. coelicolor, a first HMM2 architecture allowed the detection of short discrete DNA heterogeneities (5-16 nucleotides in size), mostly localized in intergenic regions. The application of the method on a biologically known gene set, the SigR regulon (involved in oxidative stress response), proved the efficiency in identifying bacterial promoters. S. coelicolor shows a complex regulatory network (up to 12% of the genes may be involved in gene regulation) with more than 60 sigma factors, involved in initiation of transcription. A classification method coupled to a searching algorithm (i.e. R’MES) was developed to automatically extract the box1-spacer-box2 composite DNA motifs, structure corresponding to the typical bacterial promoter -35/-10 boxes. Among the 814 DNA motifs described for the whole S. coelicolor genome, those of sigma factors (B, WhiG) could be retrieved from the crude data. We could show that this method could be generalized by applying it successfully in a preliminary attempt to the genome of Bacillus subtilis

Cette thèse est composée de deux parties indépendantes : la première partie traite des équations différentielles stochastiques dans le cadre de la G-espérance, tandis que la deuxième partie présente les résultats obtenus pour les équations différentielles stochastiques du seconde ordre. Dans un premier temps, on considère les intégrales stochastiques par rapport à un processus croissant, et on donne une extension de la formule d'Itô dans le cadre de la G-espérance. Ensuite, on étudie une classe d'équations différentielles stochastiques réfléchies unidimensionnelles dirigées par un G-mouvement brownien. Dans la suite, en utilisant une méthode de localisation, on prouve l'existence et l'unicité de solutions pour les équations différentielles stochastiques dirigées par un G-mouvement brownien, dont les coefficients sont localement lipschitziens. Enfin, dans le même cadre, on discute des problèmes de réflexion multidimensionnelle et on fournit quelques résultats de convergence. Dans un deuxième temps, on étudie une classe d'équations différentielles stochastiques rétrogrades du seconde ordre à croissance quadratique. Le but de ce travail est de généraliser le résultat obtenu par Possamaï et Zhou en 2012. On montre aussi l'existence et l'unicité des solutions pour ces équations, mais sous des hypothèses plus faibles. De plus, ce résultat théorique est appliqué aux problèmes de maximisation robuste de l'utilité du portefeuille en finance
This thesis consists of two relatively independent parts : the first part concerns stochastic differential equations in the framework of the G-expectation, while the second part deals with a class of second order backward stochastic differential equations. In the first part, we first consider stochastic integrals with respect to an increasing process and give an extension of Itô's formula in the G-framework. Then, we study a class of scalar valued reflected stochastic differential equations driven by G-Brownian motion. Subsequently, we prove the existence and the uniqueness of solutions for some locally Lipschitz stochastic differential equations driven by G-Brownian motion. At the end of this part, we consider multidimensional reflected problems in the G-framework, and some convergence results are obtained. In the second part, we study the wellposedness of a class of second order backward stochastic differential equations (2BSDEs) under a quadratic growth condition on their coefficients. The aim of this part is to generalize a wellposedness result for quadratic 2BSDEs by Possamaï and Zhou in 2012. In this thesis, we work under some usual assumptions and deduce the existence and uniqueness theorem as well. Moreover, this theoretical result for quadratic 2BSDEs is applied to solve some robust utility maximization problems in finance

Orientador: José Roberto Sanches Mantovani
Resumo: Neste trabalho propõem-se formulações matemáticas e metodologias para resolver o problema de planejamento da expansão e operação de sistemas de distribuição de energia elétrica de longo prazo com instalação de geração distribuída despachável, renovável e dispositivos armazenadores de energia, considerando as incertezas nos parâmetros e variáveis envolvidas no comportamento do sistema. No modelo de otimização desenvolvido considera- se uma formulação com espaço de busca convexo como um problema de programação cônica inteira de segunda ordem. Como primeira metodologia de solução para o modelo matemático proposto, usam-se solvers de otimização comerciais através de linguagem de programação matemática. Em segundo lugar é proposta a técnica de otimização meta-heurística VND combinada com um solver de otimização para resolver o modelo de otimização desenvolvido. Os algoritmos e modelos matemáticos de otimização usados para resolver o planejamento de sistemas de distribuição são implementados em AMPL e testados em sistemas presentes na literatura. Finalmente são comparadas as metodologias segundo a solução obtida e desempenho em tempo computacional.
Abstract: This work proposes mathematical formulations and methodologies to solve the long-term electric power distribution system operation and expansion planning with distributed renewable energy sources and energy storage devices, considering the uncertainties in the involved parameters and variables in the system behavior. In the developed optimization model, a convex formulation is considered as integer second-order conic programming problem. The first solution methodology for the proposed mathematical model, the commercial optimization solvers that uses mathematical modelling language is used. In the second way, the VND meta-heuristic optimization technique is proposed combined with the optimization solver to analyze the obtained solutions of the search through optimal neighborhoods. The mathematical optimization model and the proposed algorithm used to solver the planning of distribution systems are implemented in AMPL and tested in literature’s systems. Finally, the methodologies according to the obtained solution and computational time performance are compared.
Doutor