Mémoire numérisé par la Division de la gestion de documents et des archives de l'Université de Montréal

Mémoire numérisé par la Division de la gestion de documents et des archives de l'Université de Montréal

[pt] A habilidade de determinar o momento ótimo para investir proporciona uma vantagem estratégica para as empresas em um ambiente competitivo. Se por um lado as incertezas exógenas podem ser fontes inibidoras do investimento, sob pressão dos competidores, essas empresas podem auferir uma vantagem competitiva caso invistam primeiro. Esta tese sintetiza os conceitos aplicados às metodologias de Opções Reais e da Teoria dos Jogos para propor uma metodologia de avaliação que contribui para a análise financeira de investimentos no mercado imobiliário. O primeiro modelo da tese é um modelo econométrico que considera as inter-relações na formação dos preços que compõem a dinâmica do mercado imobiliário do Rio de Janeiro, através de um modelo de autorregressão vetorial (VAR) identificado por grafos acíclicos direcionados. O objetivo deste trabalho é o de determinar a estratégia ótima de exercício da opção de investimento no equilíbrio de Nash considerando a incerteza na demanda por imóveis em uma região do Rio de Janeiro. Para tanto, o segundo modelo da tese adota uma versão modificada da metodologia de Grenadier (2002) com uma especificação mais adequada e robusta das incertezas para a função de demanda estocástica, em que foi inserida uma elasticidade do fator estocástico. Os parâmetros do modelo foram estimados com ferramentas econométricas a partir de dados reais do mercado imobiliário carioca. Foram realizadas simulações de Monte Carlo da demanda de forma a comparar os oligopólios em termos de níveis de investimento e quantidade produzida. Os resultados obtidos nos modelos de Jogos de Opções Reais desenvolvidos são intuitivos no sentido de que quanto maior a quantidade de concorrentes, menor o nível de demanda (gatilho) exigido para o investimento em novas unidades e quanto maior a volatilidade da demanda, maior o nível de demanda para ser ótimo o investimento. / [en] The ability to determine the optimal investment timing provides a strategic advantage for companies working in a competitive environment. While on the one hand, exogenous uncertainties may inhibit investment, under pressure from competitors, these firms can gain a competitive advantage if they are first movers. This thesis summarizes the concepts applied to Real Options and Game Theory methodologies to propose an evaluation methodology that contributes to the financial analysis of investments in the real estate market. The first model of the thesis is an econometric model that considers the interrelationships in the formation of prices that compose the real estate market dynamics of Rio de Janeiro, through a vector autoregression (VAR) model identified by directed acyclic graphs. The objective of this work is to determine the optimal strategy to exercise the investment option in the Nash equilibrium considering the uncertainty in the real estate demand in a region of Rio de Janeiro. To this end, the second model of the thesis adopted a modified version of the Grenadier (2002) methodology with a more adequate and robust specification of the uncertainties for the stochastic demand function, in which a stochastic factor elasticity was inserted. The parameters of the model were estimated using econometric tools based on real data from the real estate market in Rio de Janeiro. Monte Carlo simulations of demand were performed to compare the oligopolies in terms of levels of investment and quantity produced. The quantitative results obtained are intuitive in the sense that the larger the number of competitors, the lower the level of demand (threshold) required for investment in new units, whereas the greater the volatility of demand, the greater the demand threshold for the investment to be optimal.

The paradigm of network coding was introduced as a means to conserve bandwidth (or equivalently increase throughput) in information flow networks. Network coding makes use of the fact that unlike physical commodities, information can be replicated and coded together at the nodes of the network. As a result, routing can be strictly suboptimal in many classes of information flow networks compared to network coding. Network-error correction is the art of designing network codes such that the sinks of the network will be able to decode the required information in the presence of errors in the edges of the network, known as network-errors. The network coding problem on a network with given sink demands can be considered to have the following three major subproblems, which naturally also extend to the study of network-error correcting codes, as they can be viewed as a special class of network codes (a) Existence of a network code that satisfies the demands (b) Efficient construction of such a network code (c) Minimum alphabet size for the existence of such a network code. This thesis primarily considers linear network coding and error correction and in- vestigates solutions to these issues for certain classes of network coding and error correction problems in acyclic networks. Our contributions are broadly summarised as follows. (1) We propose the use of convolutional codes for multicast network-error correc- tion. Depending upon the number of network-errors required to be corrected in the network, convolutional codes are designed at the source of the multicast network so that these errors can be corrected at the sinks of the networks as long as they are separated by certain number of time instants (for which we give a bound). In con- trast to block codes for network-error correction which require large field sizes, using convolutional codes enables the field size of the network code to be small. We discuss the performance of such networks under the BSC edge error model. (2)Existing construction algorithms of block network-error correcting codes require a rather large field size, which grows with the size of the network and the number of sinks, and thereby can be prohibitive in large networks. In our work, we give an algorithm which, starting from a given network-error correcting code, can obtain an- other network code using a small field, with the same error correcting capability as the original code. The major step in our algorithm is to find a least degree irreducible poly- nomial which is coprime to another large degree polynomial. We utilize the algebraic properties of finite fields to implement this step so that it becomes much faster than the brute-force method. A recently proposed algorithm for network coding using small fields can be seen as a special case of our algorithm for the case of no network-errors. (3)Matroids are discrete mathematical objects which generalize the notion of linear independence of sets of vectors. It has been observed recently that matroids and network coding share a deep connection, and several important results of network coding has been obtained using these connections from matroid theory. In our work, we establish that matroids with certain special properties correspond to networks with error detecting and correcting properties. We call such networks as matroidal error detecting (or equivalently, correcting) networks. We show that networks have scalar linear network-error detecting (or correcting) codes if and only if there are associated with representable matroids with some special properties. We also use these ideas to construct matroidal error correcting networks along with their associated matroids. In the case of representable matroids, these algorithms give rise to scalar linear network- error correcting codes on such networks. Finally we also show that linear network coding is not sufficient for the general network-error detection (correction) problem with arbitrary demands. (4)Problems related to network coding for acyclic, instantaneous networks have been extensively dealt with in the past. In contrast, not much attention has been paid to networks with delays. In our work, we elaborate on the existence, construction and minimum field size issues of network codes for networks with integer delays. We show that the delays associated with the edges of the network cannot be ignored, and in fact turn out to be advantageous, disadvantageous or immaterial, depending on the topology of the network and the network coding problem considered. In the process, we also show multicast network codes which involve only delaying the symbols arriving at the nodes of the networks and coding the delayed symbols over a binary field, thereby making coding operations at the nodes less complex. (5) In the usual network coding framework, for a given set of network demands over an arbitrary acyclic network with integer delays assumed for the links, the out- put symbols at the sink nodes, at any given time instant, is a Fq-linear combination of the input symbols generated at different time instants where Fq denotes the field over which the network operates. Therefore the sinks have to use sufficient memory elements in order to decode simultaneously for the entire stream of demanded infor- mation symbols. We propose a scheme using an ν-point finite-field discrete fourier transform (DFT) which converts the output symbols at the sink nodes at any given time instant, into a Fq-linear combination of the input symbols generated during the same time instant without making use of memory at the intermediate nodes. We call this as transforming the acyclic network with delay into ν-instantaneous networks (ν is sufficiently large). We show that under certain conditions, there exists a network code satisfying sink demands in the usual (non-transform) approach if and only if there exists a network code satisfying sink demands in the transform approach.

Multiple users compete for a common resource like bandwidth to communicate data in interference networks. Existing approaches in dealing with interference limit the rate of communication due to paucity of shared resources. This limitation in the rate gets more glaring as the number of users in the network increases. For example, existing wireless systems either choose to orthogonalize the users (for example, Frequency Division Multiple Access (FDMA) systems or Code Division Multiple Access (CDMA) systems) or treat interference as Gaussian noise at the receivers. It is well known that these approaches are sub-optimal in general. Orthogonalization of users limit the number of available interference-free channels (known as degrees of freedom, abbreviated as DoF) and treating interference as noise means that the receiver cannot make use of the structure in the interfering signals. This motivates the need to analyze alternate transmit and decoding schemes in interference networks. This thesis mainly analyzes transmit schemes that use linear precoding for various configurations of interference networks with some practical constraints imposed by the use of finite input constellations, propagation delays, and channel state availability at the transmitters. The main contributions of this thesis are listed below. Achievable rates using precoding with finite constellation inputs in Gaussian Interference Channels (GIC) is analyzed. A metric for finding the approximate angle of rotation to maximally enlarge the Constellation Constrained (CC) capacity of two-user Gaussian Strong Interference Channel (GSIC) is proposed. Even as the Gaussian alphabet FDMA rate curve touches the capacity curve of the GSIC, with both the users using the same finite constellation, we show that the CC FDMA rate curve lies strictly inside the CC capacity curve at high powers. For a K-user MIMO GIC, a set of necessary and sufficient conditions on the precoders under which the mutual information between between relevant transmit-receive pairs saturate like in the single user case is derived. Gradient-ascent based algorithms to optimize the sum-rate achieved by precoding with finite constellation inputs and treating interference as noise are proposed. For a class of Gaussian interference networks with general message demands, identified as symmetrically connected interference networks, the expected sumspectral efficiency (in bits/sec/Hz) is shown to grow linearly with the number of transmitters at finite SNR, using a time-domain Interference Alignment (IA) scheme in the presence of line of sight (LOS) channels. For a 2×2 MIMO X-Network with M antennas at each node, we identify spacetime block codes that could be coupled with an appropriate precoding scheme to achieve the maximum possible sum-DoF of 4M 3 , for M = 3, 4. The proposed schemes are shown to achieve a diversity gain of M with SNR-independent finite constellation inputs. The proposed schemes have lower CSIT requirements compared to existing schemes. This thesis also makes an attempt to guarantee a minimum throughput when the zero-interference conditions cannot be satisfied in a wireline network with three unicast sessions with delays, using Precoding Based Network Alignment (PBNA). Three different PBNA schemes namely PBNA with time-varying local encoding coefficients (LECs), PBNA using transform approach and time-invariant LECs, and PBNA using transform approach and block time-varying LECs are proposed and their feasibility conditions analyzed.

Dans cette thèse, nous nous intéressons à differentes colorations des sommets d’un graphe et aux homomorphismes de graphes. Nous nous intéressons plus spécialement aux graphes planaires et aux graphes peu denses. Nous considérons la coloration propre des sommets, la coloration acyclique, la coloration étoilée, lak-forêt-coloration, la coloration fractionnaire et la version par liste de la plupart de ces concepts.Dans le Chapitre 2, nous cherchons des conditions suffisantes de 3-liste colorabilité des graphes planaires. Ces conditions sont exprimées en termes de sous-graphes interdits et nos résultats impliquent plusieurs résultats connus.La notion de la coloration acyclique par liste des graphes planaires a été introduite par Borodin, Fon-Der Flaass, Kostochka, Raspaud, et Sopena. Ils ont conjecturé que tout graphe planaire est acycliquement 5-liste coloriable. Dans le Chapitre 3, on obtient des conditions suffisantes pour qu’un graphe planaire admette une k-coloration acyclique par liste avec k 2 f3; 4; 5g.Dans le Chapitre 4, nous montrons que tout graphe subcubique est 6-étoilé coloriable.D’autre part, Fertin, Raspaud et Reed ont montré que le graphe de Wagner ne peut pas être 5-étoilé-coloriable. Ce fait implique que notre résultat est optimal. De plus, nous obtenons des nouvelles bornes supérieures sur la choisissabilité étoilé d’un graphe planaire subcubique de maille donnée.Une k-forêt-coloration d’un graphe G est une application ¼ de l’ensemble des sommets V (G) de G dans l’ensemble de couleurs 1; 2; ¢ ¢ ¢ ; k telle que chaque classede couleur induit une forêt. Le sommet-arboricité de G est le plus petit entier ktel que G a k-forêt-coloration. Dans le Chapitre 5, nous prouvons une conjecture de Raspaud et Wang affirmant que tout graphe planaire sans triangles intersectants admet une sommet-arboricité au plus 2.Enfin, au Chapitre 6, nous nous concentrons sur le problème d’homomorphisme des graphes peu denses dans le graphe de Petersen. Plus précisément, nous prouvons que tout graphe sans triangles ayant un degré moyen maximum moins de 5=2 admet un homomorphisme dans le graphe de Petersen. En outre, nous montrons que la borne sur le degré moyen maximum est la meilleure possible. / In this thesis, we are interested in various vertex coloring and homomorphism problems of graphs with special emphasis on planar graphs and sparsegraphs. We consider proper vertex coloring, acyclic coloring, star coloring, forestcoloring, fractional coloring and the list version of most of these concepts.In Chapter 2, we consider the problem of finding sufficient conditions for a planargraph to be 3-choosable. These conditions are expressed in terms of forbiddensubgraphs and our results extend several known results.The notion of acyclic list coloring of planar graphs was introduced by Borodin,Fon-Der Flaass, Kostochka, Raspaud, and Sopena. They conjectured that everyplanar graph is acyclically 5-choosable. In Chapter 3, we obtain some sufficientconditions for planar graphs to be acyclically k-choosable with k 2 f3; 4; 5g.In Chapter 4, we prove that every subcubic graph is 6-star-colorable. On theother hand, Fertin, Raspaud and Reed showed that the Wagner graph cannot be5-star-colorable. This fact implies that our result is best possible. Moreover, weobtain new upper bounds on star choosability of planar subcubic graphs with givengirth.A k-forest-coloring of a graph G is a mapping ¼ from V (G) to the set f1; ¢ ¢ ¢ ; kgsuch that each color class induces a forest. The vertex-arboricity of G is the smallestinteger k such that G has a k-forest-coloring. In Chapter 5, we prove a conjecture ofRaspaud and Wang asserting that every planar graph without intersecting triangleshas vertex-arboricity at most 2.Finally, in Chapter 6, we focus on the homomorphism problems of sparse graphsto the Petersen graph. More precisely, we prove that every triangle-free graph withmaximum average degree less than 5=2 admits a homomorphism to the Petersengraph. Moreover, we show that the bound on the maximum average degree in ourresult is best possible.

Dans cette thèse, nous nous intéressons à différentes notions de colorations sous contraintes. Nous nous intéressons plus spécialement à la coloration acyclique, à la coloration forte d'arêtes et à la coloration d'arêtes sommets adjacents distinguants.Dans le Chapitre 2, nous avons étudié la coloration acyclique. Tout d'abord nous avons cherché à borner le nombre chromatique acyclique pour la classe des graphes de degré maximum borné. Ensuite nous nous sommes attardés sur la coloration acyclique par listes. La notion de coloration acyclique par liste des graphes planaires a été introduite par Borodin, Fon-Der Flaass, Kostochka, Raspaud et Sopena. Ils ont conjecturé que tout graphe planaire est acycliquement 5-liste coloriable. De notre côté, nous avons proposé des conditions suffisantes de 3-liste coloration acyclique des graphes planaires. Dans le Chapitre 3, nous avons étudié la coloration forte d'arêtes des graphes subcubiques en majorant l'indice chromatique fort en fonction du degré moyen maximum. Nous nous sommes également intéressés à la coloration forte d'arêtes des graphes subcubiques sans cycles de longueurs données et nous avons également obtenu une majoration optimale de l'indice chromatique fort pour la famille des graphes planaires extérieurs. Nous avons aussi présenté différents résultats de complexité pour la classe des graphes planaires subcubiques. Enfin, au Chapitre 4, nous avons abordé la coloration d'arêtes sommets adjacents distinguants en déterminant les majorations de l'indice avd-chromatique en fonction du degré moyen maximum. Notre travail s'inscrit dans la continuité de celui effectué par Wang et Wang en 2010. Plus précisément, nous nous sommes focalisés sur la famille des graphes de degré maximum au moins 5. / In this thesis, we are interested in various coloring of graphs under constraints. We study acyclic coloring, strong edge coloring and adjacent vertex-distinguishing edge coloring.In Chapter 2, we consider acyclic coloring and we bound the acyclic chromatic number by a function of the maximum degree of the graph. We also study acyclic list coloring. The notion of acyclic list coloring of planar graphs was introduced by Borodin, Fon-Der Flaass, Kostochka, Raspaud, and Sopena. They conjectured that every planar graph is acyclically 5-choosable. We obtain some sufficient conditions for planar graphs to be acyclically 3-choosable.In Chapter 3, we study strong edge coloring of graphs. We prove some upper bounds of the strong chromatic index of subcubic graphs as a function of the maximum average degree. We also obtain a tight upper bound for the minimum number of colors in a strong edge coloring of outerplanar graphs as a function of the maximum degree. We also prove that the strong edge k-colouring problem, when k=4,5,6, is NP-complete for subcubic planar bipartite graphs with some girth condition. Finally, in Chapter 4, we focus on adjacent vertex-distinguishing edge coloring, or avd-coloring, of graphs. We bound the avd-chromatic number of graphs by a function of the maximum average degree. This work completes a result of Wang and Wang in 2010.

Introdução: O número de estudos relacionados à Síndrome Metabólica (SM) vem aumentando nos últimos anos, muitas vezes motivados pelo aumento do número de casos de sobrepeso/obesidade e diabetes Tipo II levando ao desenvolvimento de doenças cardiovasculares e, como consequência, infarto agudo do miocárdio e AVC, dentre outros desfechos desfavoráveis. A SM é uma doença multifatorial composta de cinco características, porém, para que um indivíduo seja diagnosticado com ela, possuir pelo menos três dessas características torna-se condição suficiente. Essas cinco características são: Obesidade visceral, caracterizada pelo aumento da circunferência da cintura, Glicemia de jejum elevada, Triglicérides aumentado, HDL-colesterol reduzido, Pressão Arterial aumentada. Objetivo: Estabelecer a rede de associações entre os fenótipos que compõem a Síndrome Metabólica através do aprendizado de estruturas de dependência, decompor a rede em componentes de correlação genética e ambiental e avaliar o efeito de ajustes por covariáveis e por variantes genéticas exclusivamente relacionadas à cada um dos fenótipos da rede. Material e Métodos: A amostra do estudo corresponderá a 79 famílias da cidade mineira de Baependi, composta por 1666 indivíduos. O aprendizado de estruturas de redes será feito por meio da Teoria de Grafos e Modelos de Equações Estruturais envolvendo o modelo linear misto poligênico para determinar as relações de dependência entre os fenótipos que compõem a Síndrome Metabólica / Introduction: The number of studies related to Metabolic Syndrome (MetS) has been increasing in the last years, encouraged by the increase on the overweight / obesity and Type II Diabetes cases, leading to the development of cardiovascular disease and, therefore, acute myocardial infarction and stroke, and others unfavorable outcomes. MetS is a multifactorial disease containing five characteristics, however, for an individual to be diagnosed with MetS, he/she may have at least three of them. These characteristics are: Truncal Obesity, characterized by increasing on the waist circumference, increasing on Fasting Blood Glucose, increasing on Triglycerides, decreasing on HDL cholesterol and increasing on Blood Pressure. Aims: Establish the best association network between MetS phenotypes through structured dependency learning between phenotypes considering genetic variants exclusively related to each phenotype. Materials and Methods: The study sample is composed of 79 families, 1666 individuals of a city in a rural area of Brazil, called Beapendi. Structured learning will use graph theory and Structural Equations Models to establish the dependency relations between MetS phenotypes

Les analogues de nucléosides ont reçu une attention particulière en raison de leurs importantes applications anticancéreuses et antivirales. Dans cette thèse, de nouveaux analogues nucléosidiques de type 1′,2′-cis et 1′,2′-trans ribo portant un centre stéréogénique quaternaire fonctionnalisé en position C3′ ont été synthétisés par des réactions de N-glycosylation stéréosélectives, qui ont été contrôlées en installant différents types de groupes protecteurs sur le C2′ substituant hydroxyle. Le précurseur acyclique critique de 2,4-syn diol a été obtenu par réduction diastéréosélective d’une β-hydroxycétone en utilisant la délivrance d'hydrure intermoléculaire. Une approche pour une séparation facile des 2,4-syn et 2,4-anti diols par protection/déprotection acétonide a été établie, de sorte que le 2,4-syn diol pur puisse être rapidement accessible par oxydation allylique successive et protection acétonide. Une stratégie alternative a également été développée pour la préparation d'analogues nucléosidiques en C1′-β de type ribo portant un centre quaternaire C3′ avec un groupe hydroxyle C5′ libre. Dans cette stratégie, les diacétates de type ribo ont servi de donneur de glycosyle qui ont été synthétisés à partir d’une époxydation diastéréosélective d’un précurseur de glycal. La réaction énantiosélective consécutive de Mukaiyama aldol et le transfert d'allyle intramoléculaire de radicaux libres catalysé par photoredox ont été établis et développés dans notre laboratoire pour installer le centre stéréogénique quaternaire. Des nucléosides 5′-triphosphates portant soit une purine soit une pyrimidine ont ensuite été synthétisés et sont testés contre le cancer et les infections virales. De plus, l'analogue L-1′,2′-cis-4′-thionucléoside portant un centre quaternaire stéréogénique fonctionnalisé en position C3′ avec un substituant hydroxyle en C2′ a été synthétisé par une stratégie acyclique avec 1′,2′-syn thioaminal précurseur, qui a subi une cyclisation intramoléculaire de type SN2 de type S1′→C4′. Le 1′,2′-syn thioaminal a été synthétisé par une addition de nucléobase diastéréosélective sur un dithioacétal. / Nucleoside analogues have received extensive attention due to their important anticancer and antiviral applications. In this thesis, novel 1′,2′-cis and trans ribo-like nucleoside analogues bearing an all-carbon C3′ quaternary stereogenic center were synthesized using stereoselective N-glycosylation reactions, which were controlled by installing different types of protecting groups on the C2′ hydroxyl substituent. The critical acyclic 2,4-syn diol precursor was obtained by diastereoselective reduction of a β-hydroxy ketone using intermolecular hydride delivery. An approach for easy separation of the 2,4-syn and 2,4-anti diols through acetonide protection/deprotection was established to rapidly access the pure 2,4-syn diol through successive allylic oxidation and acetonide protection. An alternative strategy was also developed for the preparation of ribo-like C1′-β nucleoside analogues bearing an all-carbon C3′ quaternary center with a free C5′ hydroxyl group. In this strategy, ribo-like diacetates served as the glycosyl donors which were synthesized from a diastereoselective epoxidation of a glycal precursor. A consecutive enantioselective Mukaiyama aldol reaction followed by a photoredox catalyzed free radical intramolecular allyl transfer were established and developed in our lab to install the all-carbon quaternary stereogenic center. Nucleoside 5′-triphosphates bearing either a purine or a pyrimidine nucleobase were then synthesized and are currently being tested against cancer and viral infections. In addition, L-1′,2′-cis-4′-thionucleoside analogues bearing an all-carbon C3′ stereogenic quaternary center along with a C2′ hydroxyl substituent were synthesized using an acyclic strategy from a 1′,2′-syn thioaminal precursor followed by a S1′→C4′ intramolecular SN2-like cyclization. The 1′,2′-syn thioaminal was synthesized by a diastereoselective nucleobase addition onto a dithioacetal.

Um Erkenntnisse aus großen Mengen wissenschaftlicher Rohdaten zu gewinnen, sind komplexe Datenanalysen erforderlich. Scientific Workflows sind ein Ansatz zur Umsetzung solcher Datenanalysen. Um Skalierbarkeit zu erreichen, setzen die meisten Workflow-Management-Systeme auf bereits existierende Lösungen zur Verwaltung verteilter Ressourcen, etwa Batch-Scheduling-Systeme. Die Abschätzung der Ressourcen, die zur Ausführung einzelner Arbeitsschritte benötigt werden, wird dabei immer noch an die Nutzer:innen delegiert. Dies schränkt die Leistung und Benutzerfreundlichkeit von Workflow-Management-Systemen ein, da den Nutzer:innen oft die Zeit, das Fachwissen oder die Anreize fehlen, den Ressourcenverbrauch genau abzuschätzen. Diese Arbeit untersucht, wie die Ressourcennutzung während der Ausführung von Workflows automatisch erlernt werden kann. Im Gegensatz zu früheren Arbeiten werden Scheduling und Vorhersage von Ressourcenverbrauch in einem engeren Zusammenhang betrachtet. Dies bringt verschiedene Herausforderungen mit sich, wie die Quantifizierung der Auswirkungen von Vorhersagefehlern auf die Systemleistung. Die wichtigsten Beiträge dieser Arbeit sind: 1. Eine Literaturübersicht aktueller Ansätze zur Vorhersage von Spitzenspeicherverbrauch mittels maschinellen Lernens im Kontext von Batch-Scheduling-Systemen. 2. Ein Scheduling-Verfahren, das statistische Methoden verwendet, um vorherzusagen, welche Scheduling-Entscheidungen verbessert werden können. 3. Ein Ansatz zur Nutzung von zur Laufzeit gemessenem Spitzenspeicherverbrauch in Vorhersagemodellen, die die fortwährende Optimierung der Ressourcenallokation erlauben. Umfangreiche Simulationsexperimente geben Einblicke in Schlüsseleigenschaften von Scheduling-Heuristiken und Vorhersagemodellen. 4. Ein Vorhersagemodell, das die asymmetrischen Kosten überschätzten und unterschätzten Speicherverbrauchs berücksichtigt, sowie die Folgekosten von Vorhersagefehlern einbezieht. / Scientific experiments produce data at unprecedented volumes and resolutions. For the extraction of insights from large sets of raw data, complex analysis workflows are necessary. Scientific workflows enable such data analyses at scale. To achieve scalability, most workflow management systems are designed as an additional layer on top of distributed resource managers, such as batch schedulers or distributed data processing frameworks. However, like distributed resource managers, they do not automatically determine the amount of resources required for executing individual tasks in a workflow. The status quo is that workflow management systems delegate the challenge of estimating resource usage to the user. This limits the performance and ease-of-use of scientific workflow management systems, as users often lack the time, expertise, or incentives to estimate resource usage accurately. This thesis is an investigation of how to learn and predict resource usage during workflow execution. In contrast to prior work, an integrated perspective on prediction and scheduling is taken, which introduces various challenges, such as quantifying the effects of prediction errors on system performance. The main contributions are: 1. A survey of peak memory usage prediction in batch processing environments. It provides an overview of prior machine learning approaches, commonly used features, evaluation metrics, and data sets. 2. A static workflow scheduling method that uses statistical methods to predict which scheduling decisions can be improved. 3. A feedback-based approach to scheduling and predictive resource allocation, which is extensively evaluated using simulation. The results provide insights into the desirable characteristics of scheduling heuristics and prediction models. 4. A prediction model that reduces memory wastage. The design takes into account the asymmetric costs of overestimation and underestimation, as well as follow up costs of prediction errors.