A labeled embedding of a planar graph G is a pair (G, g) consisting of a planar drawing G of G and a function g assigning labels (colors) to the faces of G. We study the problem of Embedding Restriction Satisfiability (ERS) that investi- gates whether a given graph has a labeled embedding satisfying a provided set of conditions. ERS is a relatively new problem, so not much is known about it. Nevertheless, it has great potential. It generalizes several problems looking for a particular drawing of a planar graph, such as the problem of Partially Embedded Planarity. Therefore, ERS may become a focal point in the area of graph draw- ing. In this thesis, we examine the computational complexity of ERS. We show that ERS is NP-complete. After that, we look at the complexity of some specific classes of its instances. We try to locate the boundary between the NP-complete and the polynomial variants of the problem. 1

Dans la première partie, on propose un système de type pour le lambda-calcul, dans le style du calcul des séquents, nomme « Soft Type Assignment » (STA) qui est inspiré par la logique linéaire « soft ». STA a la propriété de réduction du sujet et est correct et complète pour les calculs en temps polynomial. Par la suite on propose un déduction naturelle, STA_N. Ce système est simple mais il a le désavantage que les variables dans le sujet peuvent être explicitement renommées. Pour résoudre ce problème, on propose le système STA_M, où les contextes sont des multi-ensembles, donc les règles pour renommer les variables peuvent être interdit. L’inférence de type pour STA_M ne semble pas décidable. On propose un algorithme qui pour chaque lambda-terme rend l’ensemble de contraintes que doivent être satisfait pour que le terme soit type. Pi est correct et complet. Ensuite on étend le lambda-calcul par des constantes booléennes et on propose le système STA_B. La particularité de STA_B est que la règle du conditionnel utilise les contextes de façon additive. Chaque programme de STA_B peut être exécuté, par une machine abstraite, en espace polynomial. De plus le système est aussi complet pour PSPACE. Dans la deuxième partie, on propose une restriction de PCF, nommée SlPCF. Ce langage est équipé avec une sémantique opérationnelle qui mélange l’appelle par nom et l’appelle par valeur et peut être interprèté en mode standard dans les espaces cohérents linéaires. SlPCF est complet pour les fonctions récursives, mais il n’est pas complet et donc il n’est pas fully abstract pour les espaces cohérents linéaires / In the first part, we propose, inspired by Soft Linear Logic, a type assignment system for lambda-calculus in sequent calculus style, named Soft Type Assignment (STA). STA enjoys the subject reduction property. and is correct and complete for polynomial time computations. Then, we propose a natural deduction named STA_N. While simple, STA_N has the disadvantage of allowing the explicit renaming of variables in the subject. To overcome to this problem, we propose another natural deduction system, named STA_M, where contexts are multisets, hence rules renaming variables can be avoided. The type inference for STA_M seems in general undecidable. We propose an algorithm Pi returning, for every lambda-term, a set of constraints that need to be satisfied in order to type the term. Pi is correct and complete. We extend the lambda-calculus by basic boolean constants and we propose the system STA_B. The peculiarity of STA_B is that the conditional rule treats the contexts in an additive way. Every STA_B program can be executed, through an abstract machine, in polynomial space. Moreover, STA_B is also complete for PSPACE. In the second part we propose a restriction of PCF, named SlPCF. The language is naturally equipped with an operational semantics mixing call-by-name and call-by-value parameter passing and it can be interpreted in linear coherence space in a standard way. SlPCF is recursive complete, but it is not complete, and thus not fully abstract, with respect to linear coherence spaces

Nesta tese apresentamos a abordagem da Mecânica Estatística para o clássico problema de otimização denominado problema da partição numérica (PPN), que é definido como: Dada uma seqüência de N números reais positivos {a1, a2, a3,....aN}, o problema consiste em particioná-los em dois conjuntos complementares, A e Ac, tais que o valor absoluto da diferença da soma dos ais nos dois conjuntos seja minimizada. No caso em que os aj\'s são variáveis aleatórias estatisticamente independentes distribuídas uniformemente no intervalo unitário, este problema NP-completo equivale ao problema de encontrar o estado fundamental de um modelo de Ising antiferromagnético aleatório de alcance infinito. Conseqüentemente, a análise probabilística do PPN pode ser realizada com as ferramentas da Mecânica Estatística de sistemas desordenados. Neste trabalho empregamos a aproximação recozida (annealed) para derivar uma expressão analítica para o limitante inferior do valor médio da diferença para partições tanto com vínculo de cardinalidade quanto sem vínculo para grandes valores de N. Além disso, calculamos analiticamente a fração de estados metaestáveis, isto é, estados que possuem a menor energia mediante todos os vizinhos (estados que diferem pela troca de um único spin). Concluímos a análise da abordagem direta, cujas instâncias . / In this thesis we present a statistical mechanics approach to a classical optimization problem called the number partitioning problem (NPP), which is stated as follows. Given a sequence of N positive real numbers , the number partitioning problem consists of partitioning them into two sets A and its complementary set Ac such that the absolute value of the difference of the sums of aj over the two sets is minimized. In each case in which the aj\'s are statistically independent random variables uniformly distributed in the unit interval, this NP-complete problem is equivalent to the problem of finding the ground state of an infinite range, random antiferromagnetic Ising model. Hence the probabilistic analysis of the NPP can be carried out within the framework of the standard statistical mechanics of disordered systems. In this vein we employ the annealed approximation to derive analytical lower bounds to the average value of the difference for the best-constrained and unconstrained partitions in the large N limit. Furthermore, we calculate analytically the fraction of metastable states, i.e. states that are stable against all single spin flips. We conclude the analysis of the so-called direct approach, in which the instances {ai} are fixed and the partitions are variable, with the analytical study of the linear programming relaxation of this NP-complete integer programming. In the second part of this thesis we propose and explore an inverse approach to the NPP, in which the optimal partitions are fixed and the instances are variable. Specifically, using the replica framework we study analytically the instance space of the number partitioning problem. We show that, regardless of the distribution of the instance entries, there is an upper bound &#945cN to the number of perfect random partitions (i.e. partitions for which that difference is zero). In particular, in the case where the two sets have the same cardinality (balanced partitions) we find &#945c =1/2. Moreover, in the case of unbalanced partitions, we show that perfect random partitions exist only if the difference between the cardinalities of the two sets scales like m N-1/2}.

Geometrically representable graphs are extensively studied area of research in contempo- rary literature due to their structural characterizations and efficient algorithms. The most frequently studied class of such graphs is the class of interval graphs. In this thesis we focus on two problems, generalizing the problem of recognition, for classes related to interval graphs. In the first part, we are concerned with adjusted interval graphs. This class has been studied as the right digraph analogue of interval graphs. For interval graphs, there are polynomial algorithms to extend a partial representation by given intervals into a full interval representation. We will introduce a similar problem - the partial ordering extension - and we will provide a polynomial algorithm to extend a partial ordering of adjusted interval digraphs. In the second part, we show two NP-completeness results regarding the simultaneous representation problem, introduced by Lubiw and Jampani. The simultaneous representation problem for a given class of intersection graphs asks if some k graphs can be represented so that every vertex is represented by the same object in each representation. We prove that it is NP-complete to decide this for the class of interval and circular-arc graphs in the case when k is a part of the input and graphs...

<p>In this thesis we focus resource allocation problems and electronic markets in particular. The main application area of ours is electricity markets. We present a number of algorithms and include practical experience.</p><p>There is an ongoing restructuring of power markets in Europe and elsewhere, this implies that an industry that previously has been viewed as a natural monopoly becomes exposed to competition. In the thesis we move a step further suggesting that end users should take active part in the trade on power markets such as <i>(i)</i> day-ahead markets and <i>(ii) </i>markets handling close to real-time balancing of power grids. Our ideas and results can be utilised <i>(a) </i>to increase the efficiency of these markets and <i>(b) </i>to handle strained situations when power systems operate at their limits. For this we utilise information and communication technology available today and develop electronic market mechanisms designed for large numbers of participants typically distributed over a power grid.</p><p>The papers of the thesis cover resource allocation with separable objective functions, a market mechanism that accepts actors with discontinuous demand, and mechanisms that allow actors to express combinatorial dependencies between traded commodities on multi-commodity markets. Further we present results from field tests and simulations.</p>

m-Estimation represents a broad class of estimators, including least-squares and maximum likelihood, and is a widely used tool for statistical inference. Its successful application however, often requires negotiating physical resources for desired levels of accuracy. These limiting factors, which we abstractly refer as costs, may be computational, such as time-limited cluster access for parameter learning, or they may be financial, such as purchasing human-labeled training data under a fixed budget. This thesis explores these accuracy- cost tradeoffs by proposing a family of estimators that maximizes a stochastic variation of the traditional m-estimator. Such "stochastic m-estimators" (SMEs) are constructed by stitching together different m-estimators, at random. Each such instantiation resolves the accuracy-cost tradeoff differently, and taken together they span a continuous spectrum of accuracy-cost tradeoff resolutions. We prove the consistency of the estimators and provide formulas for their asymptotic variance and statistical robustness. We also assess their cost for two concerns typical to machine learning: computational complexity and labeling expense. For the sake of concreteness, we discuss experimental results in the context of a variety of discriminative and generative Markov random fields, including Boltzmann machines, conditional random fields, model mixtures, etc. The theoretical and experimental studies demonstrate the effectiveness of the estimators when computational resources are insufficient or when obtaining additional labeled samples is necessary. We also demonstrate that in some cases the stochastic m-estimator is associated with robustness thereby increasing its statistical accuracy and representing a win-win.

In this thesis we focus resource allocation problems and electronic markets in particular. The main application area of ours is electricity markets. We present a number of algorithms and include practical experience. There is an ongoing restructuring of power markets in Europe and elsewhere, this implies that an industry that previously has been viewed as a natural monopoly becomes exposed to competition. In the thesis we move a step further suggesting that end users should take active part in the trade on power markets such as (i) day-ahead markets and (ii) markets handling close to real-time balancing of power grids. Our ideas and results can be utilised (a) to increase the efficiency of these markets and (b) to handle strained situations when power systems operate at their limits. For this we utilise information and communication technology available today and develop electronic market mechanisms designed for large numbers of participants typically distributed over a power grid. The papers of the thesis cover resource allocation with separable objective functions, a market mechanism that accepts actors with discontinuous demand, and mechanisms that allow actors to express combinatorial dependencies between traded commodities on multi-commodity markets. Further we present results from field tests and simulations.

Cette thèse porte sur l'étude des propriétés structurelles de graphes dont la compréhension permet de concevoir des algorithmes efficaces pour résoudre des problèmes d'optimisation. Nous nous intéressons plus particulièrement aux méthodes de décomposition des graphes, aux jeux de poursuites et à la notion de convexité. Le jeu de Processus a été défini comme un modèle de la reconfiguration de routage. Souvent, ces jeux où une équipe de chercheurs doit effacer un graphe non orienté sont reliés aux décompositions de graphes. Dans les digraphes, nous montrons que le jeu de Processus est monotone et nous définissons une nouvelle décomposition de graphes que lui est équivalente. Ensuite, nous étudions d'autres décompositions de graphes. Nous proposons un algorithme FPT-unifiée pour calculer plusieurs paramètres de largeur de graphes. En particulier, ceci est le premier FPT-algorithme pour la largeur arborescente q-branché et spéciale d'un graphe. Nous étudions ensuite un autre jeu qui modélise les problèmes de pré-chargement. Nous introduisons la variante en ligne du jeu de surveillance. Nous étudions l'écart entre le jeu de surveillance classique et ses versions connecté et en ligne, en fournissant de nouvelles bornes. Nous définissons ensuite un cadre général pour l'étude des jeux poursuite-évasion. Cette méthode nous permet de donner les premiers résultats d'approximation pour certains de ces jeux. Finalement, nous étudions un autre paramètre lié à la convexité des graphes et à la propagation d'infection dans les réseaux, le nombre enveloppe. Nous fournissons plusieurs résultats de complexité en fonction des structures des graphes et en utilisant des décompositions de graphes.

Let there be m parallel machines and n jobs to be scheduled non-preemptively. A job j scheduled on machine i takes p_{i,j} time units to complete, where 1 ≤ i ≤ m and 1 ≤ j ≤ n. For a given schedule, the makespan is the completion time of a machine that finishes last. The goal is to produce a schedule of all n jobs with minimum makespan. This is known as the makespan problem on unrelated parallel machines (UPMs), denoted as R||C_{max}. In this thesis, we focus on subclasses of R||C_{max}. Our research consists of two components. First, a survey of theoretic results for R||C_{max} with a focus on approximation algorithms is presented. Second, we present exact polynomial-time algorithms and approximation algorithms for some subclasses of R||C_{max}. For instance, we present k-approximation algorithms on par with or better than the best known for certain subclasses of R||C_{max}.

This thesis considers spectral approaches to finding maximum cocliques in graphs. We focus on the relation between the eigenspaces of a graph and the size and location of its maximum cocliques. Our main result concerns the computational problem of finding the size of a maximum coclique in a graph. This problem is known to be NP-Hard for general graphs. Recently, Codenotti et al. showed that computing the size of a maximum coclique is still NP-Hard if we restrict to the class of circulant graphs. We take an alternative approach to this result using quotient graphs and coding theory. We apply our method to show that computing the size of a maximum coclique is NP-Hard for the class of Cayley graphs for the groups $\mathbb{Z}_p^n$ where $p$ is any fixed prime. Cocliques are closely related to equitable partitions of a graph, and to parallel faces of the eigenpolytopes of a graph. We develop this connection and give a relation between the existence of quadratic polynomials that vanish on the vertices of an eigenpolytope of a graph, and the existence of elements in the null space of the Veronese matrix. This gives a us a tool for finding equitable partitions of a graph, and proving the non-existence of equitable partitions. For distance-regular graphs we exploit the algebraic structure of association schemes to derive an explicit formula for the rank of the Veronese matrix. We apply this machinery to show that there are strongly regular graphs whose $\tau$-eigenpolytopes are not prismoids. We also present several partial results on cocliques and graph spectra. We develop a linear programming approach to the problem of finding weightings of the adjacency matrix of a graph that meets the inertia bound with equality, and apply our technique to various families of Cayley graphs. Towards characterizing the maximum cocliques of the folded-cube graphs, we find a class of large facets of the least eigenpolytope of a folded cube, and show how they correspond to the structure of the graph. Finally, we consider equitable partitions with additional structural constraints, namely that both parts are convex subgraphs. We show that Latin square graphs cannot be partitioned into a coclique and a convex subgraph.