Orientador: Eduardo Candido Xavier / Dissertação (mestrado) - Universidade Estadual de Campinas, Instituto de Computação / Made available in DSpace on 2018-08-18T10:12:53Z (GMT). No. of bitstreams: 1 Peixoto_RobsonRobertoSouza_M.pdf: 1268588 bytes, checksum: ff8a093aa133696dcd5bbe31bc4d6e78 (MD5) Previous issue date: 2011 / Resumo: Nesta dissertação estudamos algoritmos para resolver problemas de escalonamento de tarefas em grades computacionais. Dado um conjunto de tarefas submetidas a uma grade computacional, deve-se definir em quais recursos essas tarefas serão executadas. Algoritmos de escalonamento são empregados com o objetivo de minimizar o tempo necessário para executar todas as tarefas (makespan) que foram submetidas. Nosso foco é estudar os atuais algoritmos de escalonamento usados em grades computacionais e comparar estes algoritmos. Nesta dissertação apresentamos algoritmos onlines, aproximados e heurísticas para o problema. Como resultados novos, provamos fatores de aproximação para o algoritmo RR quando utilizado para resolver os problemas R; sit|Tj|Cmax, R; sit|Tj|TPCC, R; sit|Tj = L| Cmax e R; sit|Tj = L|TPCC é justo. Por fim, definimos uma interface que adiciona replicação de tarefas a qualquer algoritmo de escalonamento, onde nós mostramos a aproximação desta interface, e apresentamos uma comparação via simulação dos algoritmos sem e com replicação. Nossas simulações mostram que, com a utilização de replicação, houve a redução no makespan de até 80% para o algoritmo Min-min. Nas nossas análises também fazemos uso da métrica RTPCC que calcula exatamente a quantidade de instruções que foram usadas para executar todas as tarefas / Abstract: In this dissertation, we studied algorithms to solve task scheduling problems in computational grids. Given a task set that was submitted to a computational grid, the problem is to define in which resources these tasks will be executed and the order they will be executed. Scheduling algorithms are used in order to minimize the time required to execute all tasks (makespan). We studied the most recent scheduling algorithms proposed to be used in computational grids, and then compare them using simulations. In this dissertation we also present approximate algorithms and new heuristics for the problem. As new results, we proved approximation factors to the RR algorithm when applied to solve the problems R; sit|Tj|Cmax, R; sit|Tj|TPCC, R; sit|Tj = L| Cmax and R; sit|Tj = L|TPCC. Finally, we defined an interface that adds task replication capability to any scheduling algorithm. We then show approximation results for algorithms using this interface, and present a comparison of well know algorithms with and without replication. This comparison is done via simulation. Our simulations show that, with replication, there was up to 80% of reduction in the makespan to some algorithms like the Min-min / Mestrado / Teoria da Computação / Mestre em Ciência da Computação

The research in this thesis is focused on the problem of Block Sorting, which has applications in Computational Biology and in Optical Character Recognition (OCR). A block in a permutation is a maximal sequence of consecutive elements that are also consecutive in the identity permutation. BLOCK SORTING is the process of transforming an arbitrary permutation to the identity permutation through a sequence of block moves. Given an arbitrary permutation π and an integer m, the Block Sorting Problem, or the problem of deciding whether the transformation can be accomplished in at most m block moves has been shown to be NP-hard. After being known to be 3-approximable for over a decade, block sorting has been researched extensively and now there are several 2-approximation algorithms for its solution. This work introduces new structures on a permutation, which are called runs and ordered pairs, and are used to develop two new approximation algorithms. Both the new algorithms are 2-approximation algorithms, yielding the approximation ratio equal to the current best. This work also includes an analysis of both the new algorithms showing they are 2-approximation algorithms.

A degree-constrained subgraph construction (DCS) problem aims to find an optimal spanning subgraph (w.r.t an objective function) subject to certain degree constraints on the vertices. DCS generalizes many combinatorial optimization problems such as Matchings and Edge Covers and has many practical and real-world applications. This thesis focuses on DCS problems where there are only upper and lower bounds on the degrees, known as b-matching and b-edge cover problems, respectively. We explore linear and submodular functions as the objective functions of the subgraph construction.<br><br>The contributions of this thesis involve both the design of new approximation algorithms for these DCS problems, and also their applications to real-world contexts.<br>We designed, developed, and implemented several approximation algorithms for DCS problems. Although some of these problems can be solved exactly in polynomial time, often these algorithms are expensive, tedious to implement, and have little to no concurrency. On the contrary, many of the approximation algorithms developed here run in nearly linear time, are simple to implement, and are concurrent. Using the local dominance framework, we developed the first parallel algorithm submodular b-matching. For weighted b-edge cover, we improved the classic Greedy algorithm using the lazy evaluation technique. We also propose and analyze several approximation algorithms using the primal-dual linear programming framework and reductions to matching. We evaluate the practical performance of these algorithms through extensive experimental results.<br><br>The second contribution of the thesis is to utilize the novel algorithms in real-world applications. We employ submodular b-matching to generate a balanced task assignment for processors to build Fock matrices in the NWChemEx quantum chemistry software. Our load-balanced assignment results in a four-fold speedup per iteration of the Fock matrix computation and scales to 14,000 cores of the Summit supercomputer at Oak Ridge National Laboratory. Using approximate b-edge cover, we propose the first shared-memory and distributed-memory parallel algorithms for the adaptive anonymity problem. Minimum weighted b-edge cover and maximum weight b-matching are shown to be applicable to constructing graphs from datasets for machine learning tasks. We provide a mathematical optimization framework connecting the graph construction problem to the DCS problem.

Considérons un graphe G = (V,E,w) non orienté dont les sommets sont pondérés et un entier k. Le problème à étudier consiste à la construction des algorithmes afin de déterminer le nombre minimum de nœuds qu’il faut enlever au graphe G pour que toutes les composantes connexes restantes contiennent chacune au plus k-sommets. Ce problème nous l’appelons problème de k-Séparateur et on désigne par k-séparateur le sous-ensemble recherché. Il est une généralisation du Vertex Cover qui correspond au cas k = 1 (nombre minimum de sommets intersectant toutes les arêtes du graphe) / Let G be a vertex-weighted undirected graph. We aim to compute a minimum weight subset of vertices whose removal leads to a graph where the size of each connected component is less than or equal to a given positive number k. If k = 1 we get the classical vertex cover problem. Many formulations are proposed for the problem. The linear relaxations of these formulations are theoretically compared. A polyhedral study is proposed (valid inequalities, facets, separation algorithms). It is shown that the problem can be solved in polynomial time for many special cases including the path, the cycle and the tree cases and also for graphs not containing some special induced sub-graphs. Some (k + 1)-approximation algorithms are also exhibited. Most of the algorithms are implemented and compared. The k-separator problem has many applications. If vertex weights are equal to 1, the size of a minimum k-separator can be used to evaluate the robustness of a graph or a network. Another application consists in partitioning a graph/network into different sub-graphs with respect to different criteria. For example, in the context of social networks, many approaches are proposed to detect communities. By solving a minimum k-separator problem, we get different connected components that may represent communities. The k-separator vertices represent persons making connections between communities. The k-separator problem can then be seen as a special partitioning/clustering graph problem

[pt] As árvores de decisão são estruturas comumente utilizadas em cenários nos quais modelos explicáveis de Aprendizado de Máquina são desejados, por serem visualmente intuitivas. Na literatura existente, a busca por explicabilidade em árvores envolve a minimização de métricas como altura e número de nós. Nesse contexto, definimos uma métrica de explicabilidade, chamada de explanation size, que reflete o número de atributos necessários para explicar a classificação dos exemplos. Apresentamos também um algoritmo, intitulado SER-DT, que obtém uma aproximação O(log n) (ótima se P diferente NP) para a minimização da altura no pior caso ou caso médio, assim como do explanation size no pior caso ou caso médio. Em uma série de experimentos, comparamos a implementação de SER-DT com algoritmos conhecidos da área, como CART e EC2, além de testarmos o impacto de parâmetros e estratégias de poda nesses algoritmos. SER-DT mostrou-se competitivo em acurácia com os algoritmos citados, mas gerou árvores muito mais explicáveis. / [en] Decision trees are commonly used structures in scenarios where explainable Machine Learning models are desired, as they are visually intuitive. In the existing literature, the search for explainability in trees involves minimizing metrics such as depth and number of nodes. In this context, we define an explainability metric, called explanation size, which reflects the number of attributes needed to explain the classification of examples. We also present an algorithm, called SER-DT, which obtains an O(log n) approximation (optimal if P different NP) for the minimization of depth in the worst/average case, as well as of explanation size in the worst/average case. In a series of experiments, we compared the SER-DT implementation with well-known algorithms in the field, such as CART and EC2 in addition to testing the impact of parameters and pruning strategies on these algorithms. SER-DT proved to be competitive in terms of accuracy with the aforementioned algorithms, but generated much more explainable trees.

No description available.

Dans cette thèse, nous étudions la complexité algorithmique de problèmes d'optimisation impliquant un processus de diffusion dans un graphe. Plus précisément, nous nous intéressons tout d'abord au problème de sélection d'un ensemble cible. Ce problème consiste à trouver le plus petit ensemble de sommets d'un graphe à “activer” au départ tel que tous les autres sommets soient activés après un nombre fini d'étapes de propagation. Si nous modifions ce processus en permettant de “protéger” un sommet à chaque étape, nous obtenons le problème du pompier dont le but est de minimiser le nombre total de sommets activés en protégeant certains sommets. Dans ce travail, nous introduisons et étudions une version généralisée de ce problème dans laquelle plus d'un sommet peut être protégé à chaque étape. Nous proposons plusieurs résultats de complexité pour ces problèmes à la fois du point de vue de l'approximation mais également de la complexité paramétrée selon des paramètres standards ainsi que des paramètres liés à la structure du graphe. / In this thesis, we investigate the computational complexity of optimization problems involving a “diffusion process” in a graph. More specifically, we are first interested to the target set selection problem. This problem consists of finding the smallest set of initially “activated” vertices of a graph such that all the other vertices become activated after a finite number of propagation steps. If we modify this process by allowing the possibility of ``protecting'' a vertex at each step, we end up with the firefighter problem that asks for minimizing the total number of activated vertices by protecting some particular vertices. In fact, we introduce and study a generalized version of this problem where more than one vertex can be protected at each step. We propose several complexity results for these problems from an approximation point of view and a parameterized complexity perspective according to standard parameterizations as well as parameters related to the graph structure.

Let G be a connected graph and k be a positive integer. A vertex subset D of G is a k-hop connected dominating set if the subgraph of G induced by D is connected, and for every vertex v in G, there is a vertex u in D such that the distance between v and u in G is at most k. We study the problem of finding a minimum k-hop connected dominating set of a graph (Mink-CDS). We prove that Mink-CDS is NP-hard on planar bipartite graphs of maximum degree 4. We also prove that Mink-CDS is APX-complete on bipartite graphs of maximum degree 4. We present inapproximability thresholds for Mink-CDS on bipar- tite and on (1, 2)-split graphs. Interestingly, one of these thresholds is a parameter of the input graph which is not a function of its number of vertices. We also discuss the complex- ity of computing this graph parameter. On the positive side, we show an approximation algorithm for Mink-CDS. When k = 1, we present two new approximation algorithms for the weighted version of the problem, one of them restricted to graphs with a poly- nomially bounded number of minimal separators. Finally, also for the weighted variant of the problem where k = 1, we discuss an integer linear programming formulation and conduct a polyhedral study of its associated polytope. / Seja G um grafo conexo e k um inteiro positivo. Um subconjunto D de vértices de G é um conjunto dominante conexo de k-saltos se o subgrafo de G induzido por D é conexo e se, para todo vértice v em G, existe um vértice u em D a uma distância não maior do que k de v. Estudamos neste trabalho o problema de se encontrar um conjunto dominante conexo de k-saltos com cardinalidade mínima (Mink-CDS). Provamos que Mink-CDS é NP-difícil em grafos planares bipartidos com grau máximo 4. Mostramos que Mink-CDS é APX-completo em grafos bipartidos com grau máximo 4. Apresentamos limiares de inaproximabilidade para Mink-CDS para grafos bipartidos e (1, 2)-split, sendo que um desses é expresso em função de um parâmetro independente da ordem do grafo. Também discutimos a complexidade computacional do problema de se computar tal parâmetro. No lado positivo, propomos um algoritmo de aproximação para Mink-CDS cuja razão de aproximação é melhor do que a que se conhecia para esse problema. Finalmente, quando k = 1, apresentamos dois novos algoritmos de aproximação para a versão do problema com pesos nos vértices, sendo que um deles restrito a classes de grafos com um número polinomial de separadores minimais. Além disso, discutimos uma formulação de programação linear inteira para essa versão do problema e provamos resultados poliédricos a respeito de algumas das desigualdades que constituem o politopo associado à formulação.

Exploramos o seguinte problema: dadas duas sequências X e Y sobre um alfabeto finito, encontre uma subsequência comum máxima de X e Y sem símbolos repetidos. Estudamos a estrutura deste problema, particularmente do ponto de vista de grafos e de combinatória poliédrica. Desenvolvemos algoritmos de aproximação e heurísticas para este problema. O enfoque deste trabalho está na construção de um algoritmo baseado na técnica branch-and-cut, aproveitando-nos de um algoritmo de separação eficiente e de heurísticas e técnicas para encontrarmos uma solução ótima mais cedo. Também estudamos um problema mais fácil no qual este problema é baseado: dadas duas sequências X e Y sobre um alfabeto finito, encontre uma subsequência comum máxima de X e Y. Exploramos este problema do ponto de vista de combinatória poliédrica e descrevemos vários algoritmos conhecidos para resolvê-lo. / We explore the following problem: given two sequences X and Y over a finite alphabet, find a longest common subsequence of X and Y without repeated symbols. We study the structure of this problem, particularly from the point of view of graphs and polyhedral combinatorics. We develop approximation algorithms and heuristics for this problem. The focus of this work is in the construction of an algorithm based on the branch-and-cut technique, taking advantage of an efficient separation algorithm and of heuristics and techniques to find an optimal solution earlier. We also study an easier problem on which this problem is based: given two sequences X and Y over a finite alphabet, find a longest common subsequence of X and Y. We explore this problem from the point of view of polyhedral combinatorics and describe several known algorithms to solve it.

No description available.