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. / 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 being connected. Finding subgraphs that satisfy common constraints of interest is computationally hard, and existing techniques for many problems of this kind are heuristic in nature. Heuristics 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 progress on these challenging graph problems in the theoretical computer science community. However, these techniques 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. One unifying theme in the problems that we study is that our methods are scalable without sacrificing theoretical guarantees. 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.

Recent advances in wireless communications along with developments in low-power circuit design and micro-electro mechanical systems (MEMS) have heralded the advent of compact and inexpensive wireless micro-sensor devices. A large network of such sensor nodes capable of communicating with each other provides significant new capabilities for automatically collecting and analyzing data from physical environments. A notable feature of these networks is that more nodes than are strictly necessary may be deployed to cover a given region. This permits the system to provide reliable information, tolerate many types of faults, and prolong the effective service time. Like most wireless systems, achieving low power consumption is a key consideration in the design of these networks. This thesis presents algorithms for managing power at the distributed system level, rather than just at the individual node level. These distributed algorithms allocate work based on user requests to the individual sensor nodes that comprise the network. The primary goal of the algorithms is to provide a robust and scalable approach for tasking nodes that prolongs the effective life of the network. Theoretical analysis and simulation results are presented to characterize the behavior of these algorithms. Results obtained from simulation experiments indicate that the algorithms can achieve a significant increase in the life of the network. In some cases this may be by an order of magnitude. The algorithms are also shown to ensure a good quality of sensor coverage while improving the network life. Finally, they are shown to be robust to faults and scale to large numbers of nodes. / Master of Science

[pt] Neste trabalho propomos uma arquitetura para a verificação formal de protocolos e algoritmos distribuídos. Esta pode ser vista como uma camada mais abstrata sobre o processo tradicional de verificação formal, onde temos a especificação e propriedade a serem verificadas, o verificador e o resultado retornado por este. O objetivo é simplificar o processo de especificação e verificação formal de protocolos e algoritmos distribuídos através de um ambiente mais dedicado. A parte principal desta arquitetura é a linguagem de especificação LEP, que contém construções de domínio-especifíco para simplificar a especificação destes sistemas. Outra característica desta linguagem é separar as especificações da topologia e do protocolo propriamente dito. Acreditamos que esta separação é válida pois torna mais clara a intenção das partes e ainda permite, por exemplo, o reuso de uma topologia entre diferentes especificações de protocolos. Assim, visamos oferecer uma linguagem cujos exemplos de especificações devem se assemelhar às descrições de algoritmos encontradas nos livros didáticos. Além disso, de forma a se ter a entrada e a saída dos verificadores formais de forma a obter a saída no nível de abstração de LEP. / [en] In this work we propose an architecture for the formal verification of protocols and distribued algoritms. This can be see as a more abstract layer over the ordinary process of formal verification, where we have just the specification of the protocol and properties to be verified, and the formal tool. Our goal is to simplifu the specification and formal verification of protocols and distributed algorithms through a dedicated environment. The core of the architecture is its input specification language (Lep), which provides domain-specific constructions for simplifying the specification of those systems. With LEP the specification of the protocol and the specification of the topology to be referred to protocol are given separetely. We feel that this division improves the legibility of both and allows the reuse of the specification of a topology among distinct protocols. Using this approach we try to offer a language whose specifications should be similar to the descriptions of the algorithms found on the didactic books. Moreover, in order to have the input and output of the architecture compatible, we also propose a way of processing the result of the formal verification tool. Then we could have the result on the abstract level of LEP.

Submitted by Geyciane Santos (geyciane_thamires@hotmail.com) on 2015-06-22T15:27:44Z No. of bitstreams: 1 Tese- Éfren Lopes de Souza.pdf: 7677745 bytes, checksum: 8fe25c4dfc5ccdc0ef44afb8837bd0e0 (MD5) / Approved for entry into archive by Divisão de Documentação/BC Biblioteca Central (ddbc@ufam.edu.br) on 2015-06-24T19:29:19Z (GMT) No. of bitstreams: 1 Tese- Éfren Lopes de Souza.pdf: 7677745 bytes, checksum: 8fe25c4dfc5ccdc0ef44afb8837bd0e0 (MD5) / Approved for entry into archive by Divisão de Documentação/BC Biblioteca Central (ddbc@ufam.edu.br) on 2015-06-24T20:11:47Z (GMT) No. of bitstreams: 1 Tese- Éfren Lopes de Souza.pdf: 7677745 bytes, checksum: 8fe25c4dfc5ccdc0ef44afb8837bd0e0 (MD5) / Made available in DSpace on 2015-06-24T20:11:47Z (GMT). No. of bitstreams: 1 Tese- Éfren Lopes de Souza.pdf: 7677745 bytes, checksum: 8fe25c4dfc5ccdc0ef44afb8837bd0e0 (MD5) Previous issue date: 2014-03-28 / CAPES - Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / Target tracking in Wireless Sensor Networks (WSNs) is an application in which the nodes cooperate to estimate the position of one or more objects of interest. In this context, the contributions of this work are fourfold. First, a survey the state-of-the-art about target tracking algorithms, in which we identified three formulations of tracking problem, and we classified them according to their characteristics. Furthermore, we divided the target tracking process in components to make the general understanding easier. Second, we propose and evaluate the PRATIQUE algorithm for tracking animals in forests. In this case, the nodes are organized into a grid to make feasible the use of sensor nodes in this kind of area in such a way that each cell of the grid is a region that can be occupied by the target. The algorithm estimates the cell where the target is, and uses predictions and hybrid clustering to reduce the communication cost and ensure the tracking accuracy. The results of the simulations show that prediction errors are approximately one cell. The third contribution is the TATI algorithm, this algorithm guides a tracker to approach the target. The sensor network is organized into faces to make the cooperation among the nodes easier, and reduce the path between the tracker and the target. The results show that energy consumption is reduced by 15%, and the tracker stays about 10m closer to the target, compared to the baseline. The fourth contribution is a scheme for performing localization and tracking tasks simultaneously in such a way that errors of range-based localization algorithms are reduced. This algorithm takes advantage of the messages sent to track the target to filter the noise in the distance estimation, reducing localization errors while tracking. The results show that the localization errors can be reduced by up to 70%. / Rastreamento de alvos em Redes de Sensores Sem Fio (RSSFs) é um tipo de aplicação em que os nós cooperam para estimar a posição de um ou mais objetos de interesse. Nesse contexto, este trabalho possui quatro contribuições. A primeira contribuição é um levantamento bibliográfico do estado-da-arte, em que identificamos três diferentes formulações de rastreamento e as classificamos de acordo com suas características. Além disso, dividimos o processo de rastreamento em componentes para facilitar o entendimento geral. A segunda contribuição é a elaboração e avaliação do algoritmo PRATIQUE para rastrear animais em florestas. Nesse caso, os nós são organizados em grade para viabilizar a utilização dos nós sensores nesse tipo de área, de forma que cada célula da grade é uma região que pode ser ocupada pelo alvo. O algoritmo estima a célula em que o alvo está, e usa previsão e um esquema híbrido de agrupamento para reduzir o custo de comunicação e garantir a precisão do rastreamento. Os resultados das simulações mostram que os erros de previsão são de aproximadamente uma célula. A terceira contribuição é o algoritmo TATI, esse algoritmo guia um objeto que visa alcançar o alvo. A rede é estruturada em faces para facilitar a cooperação entre os nós e reduzir o caminho entre o objeto guiado e o alvo. Os resultados mostram que o consumo de energia é reduzido em 15% e o objeto guiado fica cerca de 10m mais próximo do alvo, se comparado com a abordagem relacionada. A quarta contribuição é um esquema para executar as tarefas de localização e rastreamento simultaneamente para reduzir os erros dos algoritmos de localização baseados em alcance. As mensagens enviadas para rastrear o alvo são aproveitadas para filtrar os ruídos presentes nas estimativas de distância, reduzindo o erro de localização enquanto o rastreamento ocorre. Os resultados mostram que os erros de localização podem ser reduzidos em até 70%.

The subject of this thesis is the development of localization algorithms for target localization in wireless sensor networks using received signal strength (RSS) measurements or Quantized RSS (QRSS) measurements. In chapter 3 of the thesis, target localization using RSS measurements is investigated. Many existing works on RSS localization assumes that the shadowing components are uncorrelated. However, here, shadowing is assumed to be spatially correlated. It can be shown that localization accuracy can be improved with the consideration of correlation between pairs of RSS measurements. By linearizing the corresponding Maximum Likelihood (ML) objective function, a weighted least squares (WLS) algorithm is formulated to obtain the target location. An iterative technique based on Newtons method is utilized to give a solution. Numerical simulations show that the proposed algorithms achieves better performance than existing algorithms with reasonable complexity. In chapter 4, target localization with an unknown path loss model parameter is investigated. Most published work estimates location and these parameters jointly using iterative methods with a good initialization of path loss exponent (PLE). To avoid finding an initialization, a global optimization algorithm, particle swarm optimization (PSO) is employed to optimize the ML objective function. By combining PSO with a consensus algorithm, the centralized estimation problem is extended to a distributed version so that can be implemented in distributed WSN. Although suboptimal, the distributed approach is very suitable for implementation in real sensor networks, as it is scalable, robust against changing of network topology and requires only local communication. Numerical simulations show that the accuracy of centralized PSO can attain the Cramer Rao Lower Bound (CRLB). Also, as expected, there is some degradation in performance of the distributed PSO with respect to the centralized PSO. In chapter 5, a distributed gradient algorithm for RSS based target localization using only quantized data is proposed. The ML of the Quantized RSS is derived and PSO is used to provide an initial estimate for the gradient algorithm. A practical quantization threshold designer is presented for RSS data. To derive a distributed algorithm using only the quantized signal, the local estimate at each node is also quantized. The RSS measurements and the local estimate at each sensor node are quantized in different ways. By using a quantization elimination scheme, a quantized distributed gradient method is proposed. In the distributed algorithm, the quantization noise in the local estimate is gradually eliminated with each iteration. Simulations show that the performance of the centralized algorithm can reach the CRLB. The proposed distributed algorithm using a small number of bits can achieve the performance of the distributed gradient algorithm using unquantized data.

In this thesis we look at mobile agent solutions to black hole search and related problems. Mobile agents are computational entities that are autonomous, mobile, and can interact with their environment and each other. The black hole search problem is for a team of these agents to work together to map or explore a graph-like network environment where some elements of the network are dangerous to the agents. Most research into black hole search has focussed on finding a single dangerous node: a black hole. We look at the problem of finding multiple black holes and, in the case of dangerous graph exploration, multiple black links as well. We look at the dangerous graph exploration problem in the network model. The network model is based on a normal static computer network modelled as a simple graph. We give an optimal solution to the dangerous graph exploration problem using agents that start scattered on nodes throughout the network. We then make the problem more difficult by allowing an adversary to delete links during the execution of the algorithm and provide a solution using scattered agents. In the last decade or two, types of networks have emerged, such as ad hoc wireless networks, that are by their nature dynamic. These networks change quickly over time and can make distributed computations difficult. We look at black hole search in one type of dynamic network described by the subway model, which we base on urban subway systems. The model allows us to look at the cost of opportunistic movement by requiring the agents to move using carriers that follow routes among the network's sites, some of which are black holes. We show that there are basic limitations on any solution to black hole search in the subway model and prove lower bounds on any solution's complexity. We then provide two optimal solutions that differ in the agents' starting locations and how they communicate with one another. Our results provide a small window into the cost of deterministic distributed computing in networks that have dynamic elements, but which are not fully random.

We developed a diffusive load balancing scheme that equalizes the available capacities of nodes in a peer-to-peer (P2P) system. These nodes may have different resource capacities, geographic locations, or availabilities (i.e., length of time being part of the peer-to-peer system). The services on these nodes may have different service times and arrival rates of requests. Using the diffusive scheme, the system is able to maintain similar response times for its services. Our scheme is a modification of the diffusive load balancing algorithms proposed for parallel computing systems. This scheme is able to handle services with heterogeneous resource requirements and P2P nodes with heterogeneous capacities. We also adapted the diffusive scheme to clustered peer-to-peer system, where a load balancing operation may move services or nodes between clusters. After a literature survey of this field, this thesis investigates the following issues using analytical reasoning and extensive simulation studies. The load balancing operations equalize the available capacities of the nodes in a neighborhood to their averages. As a result, the available capacities of all nodes in the P2P system converge to a global average. We found that this convergence is faster when the scheme uses neighborhoods defined by the structure of the structured P2P overlay network rather than using randomly selected neighbors. For a system with churn (i.e. nodes joining and leaving), the load balancing operations maintain the standard deviation of the available capacities of nodes within a bound. This bound depends on the amount of churn and the frequency of load balancing operations, as well as on the capacities of the nodes. However, the sizes of the services have little impact on this bound. In a clustered peer-to-peer system, the size of the bound largely depends on the average cluster size. When nodes are moved among clusters for load balancing, the numbers of cluster splits and merges are reduced. This may reduce the maintenance cost of the overlay network.

In this thesis we look at mobile agent solutions to black hole search and related problems. Mobile agents are computational entities that are autonomous, mobile, and can interact with their environment and each other. The black hole search problem is for a team of these agents to work together to map or explore a graph-like network environment where some elements of the network are dangerous to the agents. Most research into black hole search has focussed on finding a single dangerous node: a black hole. We look at the problem of finding multiple black holes and, in the case of dangerous graph exploration, multiple black links as well. We look at the dangerous graph exploration problem in the network model. The network model is based on a normal static computer network modelled as a simple graph. We give an optimal solution to the dangerous graph exploration problem using agents that start scattered on nodes throughout the network. We then make the problem more difficult by allowing an adversary to delete links during the execution of the algorithm and provide a solution using scattered agents. In the last decade or two, types of networks have emerged, such as ad hoc wireless networks, that are by their nature dynamic. These networks change quickly over time and can make distributed computations difficult. We look at black hole search in one type of dynamic network described by the subway model, which we base on urban subway systems. The model allows us to look at the cost of opportunistic movement by requiring the agents to move using carriers that follow routes among the network's sites, some of which are black holes. We show that there are basic limitations on any solution to black hole search in the subway model and prove lower bounds on any solution's complexity. We then provide two optimal solutions that differ in the agents' starting locations and how they communicate with one another. Our results provide a small window into the cost of deterministic distributed computing in networks that have dynamic elements, but which are not fully random.

The Steiner Tree problem is one of the most popular graph problems and has many application areas. The rectilinear version of this problem, introduced by Hanan, has taken special attention since it addresses a fundamental matter in &ldquo / Physical Design&rdquo / phase of the Very Large Scale Integrated (VLSI) Computer Aided Design (CAD) process. The Rectilinear Steiner Tree Problem asks for a minimum length tree that interconnects a given set of points by only horizontal and vertical line segments, enabling the use of extra points. There are various exact algorithms. However the problem is NP-complete hence approximation algorithms have to be used especially for large instances. In this thesis work, first a survey on heuristics for the Rectilinear Steiner Tree Problem is conducted and then two recently developed successful algorithms, BGA by Kahng et. al. and RST by Zhou have been studied and investigated deeply. Both algorithms have subproblems, most of which have individual backgrounds in literature. After an analysis of BGA and RST, the thesis proposes a modification on RST, which leads to a faster and non-recursive version. The efficiency of the modified algorithm has been validated by computational tests using both random and VLSI benchmark instances. A partially parallelized version of the modified algorithm is also proposed for distributed computing environments. It is implemented using MPI (message passing interface) middleware and the results of comparative tests conducted on a cluster of workstations are presented. The proposed distributed algorithm has also proved to be promising especially for large problem instances.

Distributed iterative algorithms are of great importance, as they are known to provide low-complexity and approximate solutions to what are otherwise high-dimensional intractable optimization problems. The theory of message-passing based algorithms is fairly well developed in the coding, machine learning and statistical physics literatures. Even though several applications of message-passing algorithms have already been identified, this work aims at establishing that a plethora of other applications exist where it can be of great importance. In particular, the goal of this work is to develop and demonstrate applications of this class of algorithms in network communications and computational biology. In the domain of communications, message-passing based algorithms provide distributed ways of inferring the optimal solution without the aid of a central agent for various optimization problems that happen in the resource allocation of communication networks. Our main framework is Affinity Propagation (AP), originally developed for clustering problems. We reinterpret this framework to unify the development of distributed algorithms for discrete resource allocation problems. Also, we consider a network-coded communication network, where continuous rate allocation is studied. We formulate an optimization problem with a linear cost function, and then utilize a Belief Propagation (BP) approach to determine a decentralized rate allocation strategy. Next, we move to the domain of computational biology, where graphical representations and computational biology play a major role. First, we consider the motif finding problem with several DNA sequences. In effect, this is a sequence matching problem, which can be modeled using various graphical representations and also solved using low-complexity algorithms based on message-passing techniques. In addition, we address the application of message-passing algorithms for a DNA sequencing problem where the one dimensional structure of a single DNA sequence is identified. We reinterpret the problem as being equivalent to the decoding of a nonlinear code. Based on the iterative decoding framework, we develop an appropriate graphical model which enables us to derive a message-passing algorithm to improve the performance of the DNA sequencing problem. Although this work consists of disparate application domains of communications, networks and computational biology, graphical models and distributed message-passing algorithms form a common underlying theme. / text