Networks are a mathematical abstraction of the interactions between a set of entities, with extensive applications in social science, epidemiology, bioinformatics, and cybersecurity, among others. There are many fundamental problems when analyzing network data, such as anomaly detection, dense subgraph mining, motif finding, information diffusion, and epidemic spread. A common underlying task in all these problems is finding an "interesting subgraph"; that is, finding a part of the graph---usually small relative to the whole---that optimizes a score function and has some property of interest, such as connectivity or a minimum density. Finding subgraphs that satisfy common constraints of interest, such as the ones above, is computationally hard in general, and state-of-the-art algorithms for many problems in network analysis are heuristic in nature. These methods are fast and usually easy to implement. However, they come with no theoretical guarantees on the quality of the solution, which makes it difficult to assess how the discovered subgraphs compare to an optimal solution, which in turn affects the data mining task at hand. For instance, in anomaly detection, solutions with low anomaly score lead to sub-optimal detection power. On the other end of the spectrum, there have been significant advances on approximation algorithms for these challenging graph problems in the theoretical computer science community. However, these algorithms tend to be slow, difficult to implement, and they do not scale to the large datasets that are common nowadays. The goal of this dissertation is developing scalable algorithms with theoretical guarantees for various network analysis problems, where the underlying task is to find subgraphs with constraints. We find interesting subgraphs with guarantees by adapting techniques from parameterized complexity, convex optimization, and submodularity optimization. These techniques are well-known in the algorithm design literature, but they lead to slow and impractical algorithms. One unifying theme in the problems that we study is that our methods are scalable without sacrificing the theoretical guarantees of these algorithm design techniques. We accomplish this combination of scalability and rigorous bounds by exploiting properties of the problems we are trying to optimize, decomposing or compressing the input graph to a manageable size, and parallelization. We consider problems on network analysis for both static and dynamic network models. And we illustrate the power of our methods in applications, such as public health, sensor data analysis, and event detection using social media data. / Ph. D.

Alliances are used to denote agreements between members of a group with similar interests. Alliances can occur between nations, biological sequences, business cartels, and other entities. The notion of alliances in graphs was first introduced by Kristiansen, Hedetniemi, and Hedetniemi in . A defensive alliance in a graph G = (V, E) is a non empty set S ⊆ V where, for all x ∈ S, |N[x] ∩ S| ≥ |N[x] − S|. Consequently, every vertex that is a member of a defensive alliance has at least as many vertices defending it as there are vertices attacking it. Alliances can be used to model a variety of applications such as classification problems, communities in the web distributed protocols, etc [Sha01, FLG00, SX07]. In [GK98, GK00], Gerber and Kobler introduced the problem of partitioning a graph into strong defensive alliances for the first time as the "Satisfactory Graph Partitioning (SGP)" problem. In his dissertation , Shafique used the problem of partitioning a graph into alliances to model problems in data clustering. Decision problems for several types of alliances and alliance partitions have been shown to be NP-complete. However, because of their applicability, it is of interest to study methods to overcome the complexity of these problems. In this thesis, we will present a variety of algorithms for finding alliances in different families of graphs with a running time that is polynomial in terms of the size of the input, and allowing exponential running time as a function of a chosen parameter. This study is guided by the theory of parameterized complexity introduced by Rod Downey and Michael Fellows in [DF99]. In addition to parameterized algorithms for alliance related problems, we study the partition of series-parallel graphs into alliances. The class of series-parallel graphs is a special class in graph theory since many problems known to be NP-complete on general graphs have been shown to have polynomial time algorithms on series-parallel graphs [ZLL04, Hoj95, DS99, HHL87, TNS82]. For example, the problem of finding a minimum defensive alliance has been shown to have a linear time algorithm when restricted to series-parallel graphs . Series-parallel graphs have also been to focus of study in a wide range of applications including CMOS layout and scheduling problems [ML86, Oud97]. Our motivation is driven by clustering properties that can be modeled with alliances. We observe that partitioning series-parallel graphs into alliances of roughly the same size can be used to partition task graphs to minimize the communication between processors and balance the workload of each processor. We present a characterization of series-parallel graphs that allow a partition into defensive alliances and a subclass of series-parallel graphs with a satisfactory partitions.

Pour mieux saisir les liens complexes entre génotype et phénotype, une méthode utilisée consiste à étudier les relations entre différents éléments biologiques (entre les protéines, entre les métabolites...). Celles-ci forment ce qui est appelé un réseau biologique, que l'on représente algorithmiquement par un graphe. Nous nous intéressons principalement dans cette thèse au problème de la recherche d'un motif (multi-ensemble de couleurs) dans un graphe coloré, représentant un réseau biologique. De tels motifs correspondent généralement à un ensemble d'éléments conservés au cours de l'évolution et participant à une même fonction biologique. Nous continuons l'étude algorithmique de ce problème et de ses variantes (qui admettent plus de souplesse biologique), en distinguant les instances difficiles algorithmiquement et en étudiant différentes possibilités pour contourner cette difficulté (complexité paramétrée, réduction d'instance, approximation...). Nous proposons également un greffon intégré au logiciel Cytoscape pour résoudre efficacement ce problème, que nous testons sur des données réelles.Nous nous intéressons également à différents problèmes de génomique comparative. La démarche scientifique adoptée reste la même: depuis une formalisation d'un problème biologique, déterminer ses instances difficiles algorithmiquement et proposer des solutions pour contourner cette difficulté (ou prouver que de telles solutions sont impossibles à trouver sous des hypothèses fortes) / To investigate the complex links between genotype and phenotype, one can study the relations between different biological entities. It forms a biological network, represented by a graph. In this thesis, we are interested in the occurrence of a motif (a multi-set of colors) in a vertex-colored graph, representing a biological network. Such motifs usually correspond to a set of elements realizing a same function, and which may have been evolutionarily preserved. We follow the algorithmic study of this problem, by establishing hard instances and studying possibilities to cope with the hardness (parameterized complexity, preprocessing, approximation...). We also develop a plugin for Cytoscape, in order to solve efficiently this problem and to test it on real data.We are also interested in different problems related to comparative genomics. The scientific method is the same: studying problems arising from biology, specifying the hard instances and giving solutions to cope with the hardness (or proving such solutions are unlikely)

A combinatorial game is a game in which all players have perfect information and there is no element of chance; some well-known examples include othello, checkers, and chess. When people play combinatorial games they develop strategies, which can be viewed as a function which takes as input a game position and returns a move to make from that position. A strategy is winning if it guarantees the player victory despite whatever legal moves any opponent may make in response. The classical complexity of deciding whether a winning strategy exists for a given position in some combinatorial game has been well-studied both in general and for many specific combinatorial games. The vast majority of these problems are, depending on the specific properties of the game or class of games being studied, complete for either PSPACE or EXP. In the parameterized complexity setting, Downey and Fellows initiated a study of "short" (or k-move) winning strategy problems. This can be seen as a generalization of "mate-in-k" chess problems, in which the goal is to find a strategy which checkmates your opponent within k moves regardless of how he responds. In their monograph on parameterized complexity, Downey and Fellows suggested that AW[*] was the "natural home" of short winning strategy problems, but there has been little work in this field since then. In this thesis, we study the parameterized complexity of finding short winning strategies in combinatorial games. We consider both the general and several specific cases. In the general case we show that many short games are as hard classically as their original variants, and that finding a short winning strategy is hard for AW[P] when the rules are implemented as succinct circuits. For specific short games, we show that endgame problems for checkers and othello are in FPT, that alternating hitting set, hex, and the non-endgame problem for othello are in AW[*], and that short chess is AW[*]-complete. We also consider pursuit-evasion parameterized by the number of cops. We show that two variants of pursuit-evasion are AW[*]-hard, and that the short versions of these problems are AW[*]-complete.

Title: Parameterized Complexity Author: Ondřej Suchý Department: Department of Applied Mathematics Advisor: Prof. RNDr. Jan Kratochvíl, CSc. Advisor's e-mail address: honza@kam.mff.cuni.cz Abstract: This thesis deals with the parameterized complexity of NP-hard graph problems. We explore the complexity of the problems in various scenarios, with respect to miscellaneous parameters and their combina- tions. Our aim is rather to classify in this multivariate manner whether the particular parameters make the problem fixed-parameter tractable or intractable than to present the algorithm achieving the best running time. In the questions we study typically the first-choice parameter is unsuccessful, in which case we propose to use less standard ones. The first family of problems investigated provides a common general- ization of many well known and studied domination and independence problems. Here we suggest using the dual parameterization and show that, in contrast to the standard solution-size, it can confine the in- evitable combinatorial explosion. Further studied problems are ana- logues of the Steiner problem in directed graphs. Here the parameter- ization by the number of terminals to be connected seems to be previ- ously unexplored in the directed setting. Unfortunately, the problems are shown to be...

Dans un graphe orienté contenant un nœud appelé racine, un sous ensemble de nœuds appelés terminaux et une pondération sur les arcs, le problème de l’arborescence de Steiner (DST) consiste en la recherche d’une arborescence de poids minimum contenant pour chaque terminal un chemin de la racine vers ce terminal. Ce problème est NP-Complet. Cette thèse se penche sur l’étude de l’approximabilité de ce problème. Sauf si P=NP, il n’existe pas pour ce problème d’approximation de rapport constant ou logarithmique en k, oú k est le nombre de terminaux. Le plus petit rapport d’approximation connu est O (k") où " est un réel strictement positif. Dans la première partie, nous donnons trois algorithmes d’approximation : un algorithme glouton efficace qui associe deux techniques d’approximations connues pour DST, un algorithme dans le cas des graphes structurés en paliers qui étudie l’approximabilité du problème quand les terminaux sont éloignés de la racine, et un algorithme exponentiel qui combine un algorithme d’approximation et un algorithme exact, dont le rapport d’approximation et la complexité temporelle sont paramétrés par le nombre de terminaux couverts par l’algorithme exact. Dans la seconde partie, nous étudions deux problèmes issus de DST auquel est ajoutée une contrainte sur les nœuds de branchement. Cette contrainte réduit le nombre de solutions réalisables et peut faciliter la recherche d’une solution optimale parmi ce sous-ensemble de solutions. En fonction de la contrainte, nous étudions la possibilité de la trouver en temps polynomial et quel est le rapport d’approximation entre cette solution et la solution du problème non contraint / The directed Steiner tree problem (DST) asks, considering a directed weighted graph, a node r called root and a set of nodes X called terminals, for a minimum cost directed tree rooted in r spanning X. DST is an NP-complete problem. We are interested in the search for polynomial approximations for DST. Unless P = NP, DST can not be approximated neither within a constant ratio nor a logarithmic ratio with respected to k, where k is the number of terminals. The smallest known approximation ratio is O(kԑ)$ where ԑ is a positive real.In the first part, we provide three new approximation algorithms : a practical greedy algorithm merging two of the main approximation techniques for DST, an algorithm for the case where the graph is layered and where the distance between the terminals and the root is high, and an exponential approximation algorithm combining an approximation algorithm and an exact algorithm, parameterized with the number of terminals the exact algorithm must cover.In the last part we study DST with two branching constraints. With this technique, we are able to reduce the number of feasible solutions, and possibly facilitate the search for an optimal solution of the constraint problem. We study how it is possible to build such a solution in polynomial time and if this solution is a good approximation of an optimal solution of the non-constraint problem

This thesis consists of three parts devoted to graph labeling, hereditary graph classes, and parameterized complexity. Packing coloring, originally Broadcasting Chromatic number, assigns natural numbers to vertices such that vertices with the same label are in distance at least the value of the label. This problem is motivated by the assignment of frequencies to the transmitters. We improve hardness on chordal graphs. We proof that packing coloring on chordal graphs with diameter 3 is very hard to approximate. Moreover, we discuss several positive results on interval graphs and on related structural graph parameters. Hereditary graph classes are preserved under vertex deletion. We study graphs that do not contain an induced subgraph H. We prove that 3-coloring is polynomial-time solvable for (P3 + P4)-free and (P2 + P5)-free graphs and thus we have solved the last open cases for the problem on H-free graphs where H has up to 7 vertices. Fair problems are a modification of graph deletion problems, where, instead of minimizing the size of the solution, the aim is to minimize the maximum number of neighbors in the deleted set. We show that those problems can be solved in FPT time for an MSO1 formula parameterized by the size of the formula and the twin cover of the graph. Moreover, we define a basic...

Nous étudions des problèmes NP-difficiles portant sur les graphes contenant des blocages.Nous traitons les problèmes de coupes du point de vue de la complexité paramétrée. La taille p de la coupe est le paramètre. Étant donné un ensemble de sources {s1,...,sk} et une cible t, nous proposons un algorithme qui construit une coupe de taille au plus p séparant au moins r sources de t. Nous nommons ce problème NP-complet Partial One-Target Cut. Notre algorithme est FPT. Nous prouvons également que la variante de Partial One-Target Cut, où la coupe est composée de noeuds, est W[1]-difficile. Notre seconde contribution est la construction d'un algorithme qui compte les coupes minimums entre deux ensembles S et T en temps $2^{O(plog p)}n^{O(1)}$.Nous présentons ensuite plusieurs résultats sur le ratio de compétitivité des stratégies déterministes et randomisées pour le problème du voyageur canadien.Nous prouvons que les stratégies randomisées n'utilisant pas de mémoire ne peuvent pas améliorer le ratio 2k+1. Nous apportons également des éléments concernant les bornes inférieures de compétitivité de l'ensemble des stratégies randomisées. Puis, nous étudions la compétitivité en distance d'un groupe de voyageurs avec et sans communication. Enfin, nous nous penchons sur la compétitivité des stratégies déterministes pour certaines familles de graphes. Deux stratégies, avec un ratio inférieur à 2k+1 sont proposées: une pour les graphes cordaux avec poids uniformes et l'autre pour les graphes où la taille de la plus grande coupe minimale séparant s et t est au plus k. / We study NP-hard problems on graphs with blockages seen as models of networks which are exposed to risk of failures.We treat cut problems via the parameterized complexity framework. The cutset size p is taken as a parameter. Given a set of sources {s1,...,sk} and a target $t, we propose an algorithm which builds a small edge cut of size p separating at least r sources from t. This NP-complete problem is called Partial One-Target Cut. It belongs to the family of multiterminal cut problems. Our algorithm is fixed-parameter tractable (FPT) as its execution takes $2^{O(p^2)}n^{O(1)}$. We prove that the vertex version of this problem, which imposes cuts to contain vertices instead of edges, is W[1]-hard. Then, we design an FPT algorithm which counts the minimum vertex (S,T)-cuts of an undirected graph in time $2^{O(plog p)}n^{O(1)}$.We provide numerous results on the competitive ratio of both deterministic and randomized strategies for the Canadian Traveller Problem. The optimal ratio obtained for the deterministic strategies on general graphs is 2k+1, where k is a given upper bound on the number of blockages. We show that randomized strategies which do not use memory cannot improve the bound 2k+1. In addition, we discuss the tightness of lower bounds on the competitiveness of randomized strategies. The distance competitive ratio for a group of travellers possibly equipped with telecommunication devices is studied. Eventually, a strategy dedicated to equal-weight chordal graphs is proposed while another one is built for graphs with small maximum (s,t)-cuts. Both strategies outperform the ratio 2k+1.

Dans cette thèse, nous abordons des problèmes NP-difficiles à l'aide de techniques combinatoires, en se focalisant sur le domaine de la complexité paramétrée. Les principaux problèmes que nous considérons sont les problèmes de Multicoupe et d'Arbre Orienté Couvrant avec Beaucoup de Feuilles. La Multicoupe est une généralisation naturelle du très classique problème de coupe, et consiste à séparer un ensemble donné de paires de sommets en supprimant le moins d'arêtes possible dans un graphe. Le problème d'Arbre Orienté Couvrant avec Beaucoup de Feuilles consiste à trouver un arbre couvrant avec le plus de feuilles possible dans un graphe dirigé. Les résultats principaux de cette thèse sont les suivants. Nous montrons que le problème de Multicoupe paramétré par la taille de la solution est FPT (soluble à paramètre fixé), c'est-à-dire que l'existence d'une multicoupe de taille $k$ dans un graphe à $n$ sommets peut être décidée en temps $f(k)*poly(n)$. Nous montrons que Multicoupe dans les arbres admet un noyau polynomial, c'est-à-dire est réductible aux instances de taille polynomiale en $k$. Nous donnons un algorithme en temps $O^*(3.72^k)$ pour le problème d'Arbre Orienté Couvrant avec Beaucoup de Feuilles et le premier algorithme exponentiel exact non trivial (c'est-à-dire meilleur que $2^n$). Nous fournissons aussi un noyau quadratique et une approximation à facteur constant. Ces résultats algorithmiques sont basés sur des résultats combinatoires et des propriétés structurelles qui concernent, entre autres, les décompositions arborescentes, les mineurs, des règles de réduction et les $s-t$ numberings. Nous présentons des résultats combinatoires hors du domaine de la complexité paramétrée: une caractérisation des graphes de cercle Helly comme les graphes de cercle sans diamant induit, et une caractérisation partielle des classes de graphes 2-bel-ordonnées. / This thesis tackles NP-hard problems with combinatorial techniques, focusing on the framework of Fixed-Parameter Tractability. The main problems considered here are Multicut and Maximum Leaf Out-branching. Multicut is a natural generalisation of the cut problem, and consists in simultaneously separating prescribed pairs of vertices by removing as few edges as possible in a graph. Maximum Leaf Out-branching consists in finding a spanning directed tree with as many leaves as possible in a directed graph. The main results of this thesis are the following. We show that Multicut is FPT when parameterized by the solution size, i.e. deciding the existence of a multicut of size $k$ in a graph with $n$ vertices can be done in time $f(k)*poly(n)$. We show that Multicut In Trees admits a polynomial kernel, i.e. can be reduced to instances of size polynomial in $k$. We give an $O^*(3.72^k)$ algorithm for Maximum Leaf Out-branching and the first non-trivial (better than $2^n$) exact algorithm. We also provide a quadratic kernel and a constant factor approximation algorithm. These algorithmic results are based on combinatorial results and structural properties, involving tree decompositions, minors, reduction rules and $s-t$ numberings, among others. We present results obtained with combinatorial techniques outside the scope of parameterized complexity: a characterization of Helly circle graphs as the diamond-free circle graphs, and a partial characterisation of 2-well-quasi-ordered classes of graphs.

La première partie de cette thèse s'intéresse au groupage de trafic dans les réseaux de télécommunications. La notion de groupage de trafic correspond à l'agrégation de flux de faible débit dans des conduits de plus gros débit. Cependant, à chaque insertion ou extraction de trafic sur une longueur d'onde il faut placer dans le noeud du réseau un multiplexeur à insertion/extraction (ADM). De plus il faut un ADM pour chaque longueur d'onde utilisée dans le noeud, ce qui représente un coût d'équipements important. Les objectifs du groupage de trafic sont d'une part le partage efficace de la bande passante et d'autre part la réduction du coût des équipements de routage. Nous présentons des résultats d'inapproximabilité, des algorithmes d'approximation, un nouveau modèle qui permet au réseau de pouvoir router n'importe quel graphe de requêtes de degré borné, ainsi que des solutions optimales pour deux scénarios avec trafic all-to-all: l'anneau bidirectionnel et l'anneau unidirectionnel avec un facteur de groupage qui change de manière dynamique. La deuxième partie de la thèse s'intéresse aux problèmes consistant à trouver des sous-graphes avec contraintes sur le degré. Cette classe de problèmes est plus générale que le groupage de trafic, qui est un cas particulier. Il s'agit de trouver des sous-graphes d'un graphe donné avec contraintes sur le degré, tout en optimisant un paramètre du graphe (très souvent, le nombre de sommets ou d'arêtes). Nous présentons des algorithmes d'approximation, des résultats d'inapproximabilité, des études sur la complexité paramétrique, des algorithmes exacts pour les graphes planaires, ainsi qu'une méthodologie générale qui permet de résoudre efficacement cette classe de problèmes (et de manière plus générale, la classe de problèmes tels qu'une solution peut être codé avec une partition d'un sous-ensemble des sommets) pour les graphes plongés dans une surface. Finalement, plusieurs annexes présentent des résultats sur des problèmes connexes.