Theses and Dissertations

2016

ABSTRACTS

Departamento de Informática 
Pontifícia Universidade Católica do Rio de Janeiro - PUC-Rio
Rio de Janeiro - Brazil
 

This file contains the list of the MSc. Dissertations and PhD. Thesis presented to the Departmento de Informática, Pontifícia Universidade Católica do Janeiro - PUC-Rio, Brazil, in 2016.  They are all available in print format and, according to the authors' preference, some of them are freely available for download, while others are freely available for download to the PUC-Rio community exclusively(*). 

For any requests, questions, or suggestions, please contact:
Rosane Castilho bib-di@inf.puc-rio.br

Last update: 21/JULY/2017
 

INDEX

[In construction; sometimes, digital versions may not be available yet]

[16_MSc_dominguez]
Alain DOMINGUEZ FUENTES. Sintonia fina automática com índices parciais. [Title in English: Database self-tunning with partial indexes]. M.Sc Diss. Port. Presentation: 29/03/2016. 82 p. Advisor: Sergio Lifschitz.

Abstract: Partial indexes are access structures on the physical level of the databases. They are indexes that allow the definition of a subset of tuples in a table through a conditional expression. This dissertation studies the identification and subsequent automatic creation of partial indexes that can contribute in improving the performance of a database system. We propose an algorithm that examines, for each relevant query, the indexable attributes set, for which the creation of a partial index could influence the query optimizer to generate plans that are more efficient. We perform data mining on indexable attributes patterns to obtain correlated attributes according to their frequency in queries within the particular workload. We obtain a proposal for a set of candidate partial indexes considering also a benefit heuristics. We may consider a self-tuning analysis of an index configuration with both complete and partial indexes. We have implemented techniques and algorithms proposed in this research into DBX, a framework that allows local and global self-tuning regarding relational databases.

[16_MSc_diaz]
Alejandro DIAZ CENTENO. Evaluation of physical-motor status of people with reduced mobility using motion capture with Microsoft Kinect. [Title in Portuguese: Avaliação do estado físico-motor de pessoas com mobilidade reduzida usando captura de movimento com o Microsoft Kinect]. M.Sc. Diss. Eng. Presentation: 30/09/2016. 65 p. Advisor: Alberto Barbosa Raposo.

Abstract: The evaluation of the motor status of stroke patients and elderly people is done by using qualitative scales without standardization or measuring instruments. The scales are most common because they are relatively inexpensive and accessible, but suffer the disadvantage of being subjective, variable, and require prolonged training time. Moreover, assessment instruments, although more accurate and objective, have the problem of being heterogeneous, usually very expensive, and focused on specific goals. The rise in recent times of 3D sensors with high accuracy and low cost, some of them well known as the Microsoft Kinect, allows the use of motion analysis to quantify the deficit or success of a physiotherapeutic or drug treatment in a quantitative and standardized way, enabling the automatic comparison with standards of healthy people, and people with the same stage of disease, or similar characteristics. The aim of this work is to create a system using Microsoft Kinect for capturing and processing motor status of patients with reduced mobility in a non-invasive way, providing clinical feedback that allows the conduction of a quantitative and objective evaluation of patients, enabling monitoring of disease progression and reduced rehabilitation time.

[16_MSc_lucas]
Ana Paula Lima LUCAS. Gestão da manutenção de software: um estudo de caso. [Title in English: Software maintenance management: a case study]. M.Sc. Diss. Port. Presentation: 27/09/2016. 104 p. Advisor: Arndt von Staa.

Abstract: The company participating in this work sought to implement activities related to software maintenance due to problems related to the great occurrence of defects, constant rework, among others. To suppress these problems, a preliminary study of the system in question has been elaborated, evaluating the current state of software maintenance. In view of the diagnosis, the necessity of changes became evident concerning the way that the system maintenance activities were conducted. With this, the search for alternatives of improvements began with the objective of reducing the occurrence of defects and also increase the maintainability of the system. Knowing the problems and what could be done to improve; it was proposed to accede some practices of the software maintenance maturity model - SMmm and to integrate the concepts of these practices into a defined process and adapted to the needs of the system. To support this implementation, the infrastructure used was the Team Foundation Service - TFS platform that collaborated with the implementation of the selected practices according to the requirements of the SMmm model, resulting in a defined process supported by the TFS that partially implements the SMmm model. This paper presents a case study with the objective of evaluating the benefits provided by the use of some practices of the SMmm model. The evaluation carried out compared the data from the preliminary study with data collected after adoption of the practices, the analyzed results pointed out a significant reduction in the number of issues.

[16_MSc_macdowell]
André Victor Gomes de Aboim MAC DOWELL. Uma API para exergames móveis com evento centrados em microlocalização baseada em BLE fingerprinting. [Title in English: An API for mobile exergames micro location events using BLE fingerpriting]. M.Sc. Diss. Port. Presentation: 04/06/2016. 106 p. Advisor: Markus Endler.

Abstract: Smartphones are ever more present in the day to day lives of our society, for both work and entertainment. In this mobile platform, there is a growing number of games that uses sensing capabilities of the Smartphone for its gameplay mechanics, like GPS for location-based games, a category of mobile pervasive games. Although, there are categories of pervasive games that require specific hardware capabilities not normally found in a Smartphone, like precise proximity inference between devices and a more precise, fast and reliable location solution then GPS. Simultaneously, both sensing and beacon technologies for Internet of Things (IoT) are getting cheaper and more available, and there are many micro locating solutions that uses these technologies in different application contexts. In Mobile Exergames, a category of pervasive games where the gameplay is outdoors and fast-paced, with constant interaction between multiple players, a more precise location solution then GPS is necessary. The development of these games includes the execution of game sessions and their components, together with the interoperability of different technologies. That way, in this work, we present a location strategy using Fingerprinting and Bluetooth LE (BLE) beacons and a API for common location requests and events. We analyze the location strategy through tests with different configurations using a Pervasive Game Middleware with Session management, and evaluate the location API through gameplay abstractions for a few pervasive games.

[16_MSc_moreira]
Andrey D' Almeida Rocha RODRIGUES. Visualização de modelos digitais de elevação em multiresolucão utilizando programação em GPU. [Title in English: Muiti-resolution visualization of digital elevaton models using GPU shaders]. M.Sc. Diss. Port. Presentation: 07/04/2016. 37 p. Advisor: Waldemar Celes Filho.

Abstract: 3D CAD Models have played an important role in engineering projects’ management. It is noticeable in many of these files the presence of several objects with implicit representation that end up being represented as triangular meshes. Although suitable for rendering, the triangular mesh representation brings some drawbacks, such as the ambiguity in objects with low discretization rate. The reverse engineering aims to reconstruct this discrete representation to its original continuous representation. In this work, we propose a novel methodology for geometry reconstruction in CAD models using Support Vector Machines and Shape Descriptors.

[16_MSc_castilhoneto]
Arthur Beltrão CASTILHO NETO. Anotador de papeis semânticos para Português. [Title in English: Semantic Role-Labeling for Portuguese]. M.Sc. Diss. Port. Presentation: 16/12/2016. 78 p. Advisor: Ruy Luiz Milidú.

Abstract: Semantic role-labeling (SRL) is an important task of natural language processing (NLP) which allows establishing meaningful relationships between events described in a given sentence and its participants. Therefore, it can potentially improve performance on a large number of NLP systems such as automatic translation, spell correction, information extraction and retrieval and question answering, as it decreases ambiguity in the input text. The vast majority of SRL systems reported so far employed supervised learning techniques to perform the task. For better results, large sized manually reviewed corpora are used. The Brazilian semantic role labeled lexical resource (Propbank.br) is much smaller. Hence, in recent years, attempts have been made to improve performance using semi supervised and unsupervised learning. Even making several direct and indirect contributions to NLP, those studies were not able to outperform exclusively supervised systems. This paper presents an approach to the SRL task in Portuguese language using supervised learning over a set of 114 categorical features. Over those, we apply a combination of two domain regularization methods to cut binary features down to 96%. We test a SVM model (L2-loss dual support vector classification) on PropBank.Br dataset achieving results slightly better than state-of-the-art. We empirically evaluate the system using official CoNLL 2005 Shared Task script pulling 82.17% precision, 82.88% coverage and 82.52% F1. The previous state-of-the-art Portuguese SRL system scores 83.0% precision, 81.7% coverage and 82.3% F1.

[16_MSc_pontes]
Bruno Silva PONTES. Reconhecimento de posturas humanas preservando a privacidade: um estudo de caso usando um sensor térmico de baixa resolução. [Title in English: Human posture recognition preserving privacy: a case study using a low resolution array thermal sensor]. M.Sc. Diss. Port. Presentation: 30/09/2016. 78 p. Advisor: Hugo Fuks.

Abstract: Postures recognition is one of the human sensing challenges, that helps ambient assisted livings in people accompanying. On the other hand, these ambients assist doctors in the diagnosis of their patients’ health, mainly through activities of daily livings real time recognition, which is seen in the medical field as one of the best ways to anticipate critical health situations. In addition, the world’s population aging, lack of hospital resources to meet all people and increased health care costs drive the development of systems to support ambient assisted livings. Preserving privacy in these ambients monitored by sensors is a critical factor for user acceptance, so there is a demand for solutions that does not requires images. This work demonstrates the use of a low resolution thermal array sensor in human sensing, showing that it is feasible to detect the presence and to recognize human postures, using only the data of this sensor.

[16_PhD_mendes]
Carlos Augusto Teixeira MENDES. GeMA, um novo framework para a prototipação, desenvolvimento e integração de simulações multifísicas e multiescalas em grupos multidisciplinares. [Title in English: GeMA, a new framework for prototyping, development and integration of multiphysics and multiscale simulations in multidisciplinary groups]. Ph. D. Thesis. Port. Presentation: 01/04/2016. 168 p. Advisor: Marcelo Gattass.

Abstract: Petroleum exploration and production is a complex task where the use of physical models is imperative to minimize exploration risks and maximize the return on the invested capital during the production phase of new oil fields. Over time, these models have become more and more complex, giving rise to a tendency of integration between several simulators and the need for new multiphysics simulations, where single-physics models are solved together in a coupled way. This work presents the GeMA (Geo Modelling Analysis) framework, a library to support the development of new multiphysics simulators, allowing both the coupling of new models built with the framework as a base and the integration with pre-existing simulators. Its objective is to promote the use of software engineering techniques, such as extensibility, reusability, modularity and portability in the construction of engineering physical models, allowing engineers to focus on the physical problem formulation since the framework takes care of data management and other necessary support functions, speeding up code development. Built to aid during the entire multiphysics simulation workflow, the framework architecture supports multiple simulation and coupling paradigms, with special emphasis given to finite element methods. Being capable of representing the spatial domain by multiple discretizations (meshes) and exchanging values between them, the framework also implements some important concepts of extensibility, through the combined use of plugins and abstract interfaces, configurable orchestration and fast prototyping through the use of the Lua language. This work also presents a set of test cases used to assess the framework correctness and expressiveness, with particular emphasis given to a 2D basin model that couples FEM non-linear temperature calculations based on finite elements, mechanical compaction and hydrocarbon maturation and generation.

[16_MSc_marques]
Daniel dos Santos MARQUES. A decision tree learner for cost-sensitive binary classification. [Title in Portuguese: Uma árvore de decisão para classificação binária sensível ao custo]. M.Sc. Diss. Eng. Presentation: 22/09/2016. 46 p. Advisor: Eduardo Sany Laber.

Abstract: Classifcation problems have been widely studied in the machine learning literature, generating applications in several areas. However, in a number of scenarios, misclassi cation costs can vary substantially, which motivates the study of Cost-Sensitive Learning techniques. In the present work, we discuss the use of decision trees on the more general Example-Dependent Cost-Sensitive Problem (EDCSP), where misclassi cation costs vary with each example. One of the main advantages of decision trees is that they are easy to interpret, which is a highly desirable property in a number of applications. We propose a new attribute selection method for constructing decision trees for the EDCSP and discuss how it can be eciently implemented. Finally, we compare our new method with two other decision tree algorithms recently proposed in the literature, in 3 publicly available datasets.

[16_PhD_baia]
Davy de Medeiros BAÍA. Modelagem de contextos dinâmicos em simulação de gestão de projetos de software baseada em multiagentes. [Title in Portuguese: Dynamic modelling in software project management simulation based on multi-agent]. Ph.D. Thesis. Port. Presentation: 08/03/2016. 201 p. Advisor: Calos José Pereira de Lucena.

Abstract: Software Project Management is not a trivial task, especially with the changes that occur during the course of its execution. Generally, software project has elements such as tasks and human resources. Each of these elements has its features and their relationships with each other. These elements, their features and their relationships defines a context.We define as dynamic context changes that occur in the context during execution of the project. Persons involved in project decision-making need to deal with this dynamic context, thus increasing the complexity of the project management. Simulations often apply to support answer specific questions or phenomena on a domain, through analysis of experiments or executions results. For this, simulations use models that capture details of specifics domains. Multi-agent systems modeling provides robust models to represent real-world environments that are complex and dynamic. One of the advantages of using multi-agent-based simulation is its ability to support realistic aspects of project management, incorporating its elements through agents. However, there is a lack of an approach to simulate the software projects management, to represent the context, execute scenarios to assist in the decision-making process and to support the dynamic context that occurs throughout the project execution. In this context, we propose a conceptual model based on multi-agent system and software project management simulation, the ProMabs. This conceptual model contains five components to model the context and its dynamics, thus, execute simulations of scenarios that assist in the decision-making process. As contributions, this thesis presents ProMabs with three instances based on different technologies: a multi-agent programming platform, a simulation environment, and a simulation tool. Therewith, evaluate the ProMabs from three different perspectives. The application of ProMabs with these technologies, allows to represent elements and their relationships, create scenarios to assist decision-making and support adaptive software project, i.e., with dynamic context. Finally, we present an experiment with qualitative and quantitative analysis, with application of the ProMabs to represent the context, support its dynamic and assist decision-making by means of scenarios. The experimental results are positive indications that the instantiation of the ProMabs as supported by a simulation tool assists the participants in decision making.

[16_MSc_silva]
Djalma Lúcio Soares da SILVA. Uso de estruturas planares extraídas de imagens RGB-D em aplicações de realidade aumentada. [Title in English: Using Planar Structures Extracted from RGB-D Images in Augmented Reality Applications]. M. Sc. Diss. Port. Presentation: 01/08/2016. 68 p. Advisor: Alberto Barbosa Raposo.

Abstract: This dissertation investigates the use of planar geometric structures extracted from RGB-D images in Augmented Reality Applications. The model of a scene is essential for augmented reality applications. RGB-D images can greatly help the construction of these models because they provide geometric and photometric information about the scene. Planar structures are prevalent in many 3D scenes and, for this reason, augmented reality applications use planar surfaces as one of the main components for projection of virtual objects. Therefore, it is extremely important to have robust and efficient methods to acquire and represent the structures that compose these planar surfaces. In this work, we will present a method for identifying, targeting and representing planar structures from RGB-D images. Our planar structures representation is triangulated two-dimensional polygons, simpli- fied and textured, forming a triangle mesh intrinsic to the plane that defines regions in this space corresponding to surface of objects in the 3D scene. We have demonstrated through various experiments and implementation of an augmented reality application, the techniques and methods used to extract the planar structures from the RGB-D images.

[16_PhD_sarmiento]
Edgar SARMIENTO CALISAYA. Analysis of natural language scenarios. [Title in Portuguese: Análise de cenários em linguagem natural]. Ph. D. Thesis. Eng. Presentation: 13/04/2016. 231 p. Advisor: Julio Cesar Sampaio do Prado Leite.

Abstract: Requirements analysis plays a key role in the software development process. Natural language-based scenario representations are often used for writing software requirements specifications (SRS). Scenarios written using natural language may be ambiguous, and, sometimes, inaccurate. This problem is partially due to the fact that relationships among scenarios are rarely represented explicitly. As scenarios are used as input to subsequent activities of the software development process (SD), it is very important to enable their analysis; especially to detect defects due to wrong information or missing information. This work proposes a Petri-Net and Natural Language Processing (NLP) based approach as an effective way to analyze the acquired scenarios, which takes textual description of scenarios (conform to a meta-model defined in this work) as input and generates an analysis report as output. To enable the automated analysis, scenarios are translated into equivalent Place/Transition Petri-Nets. Scenarios and their resulting Petri-Nets can be automatically analyzed to evaluate some properties related to unambiguity, completeness, consistency and correctness. The identified defects can be traced back to the scenarios, allowing their revision. We also discuss how unambiguity, completeness, consistency and correctness of scenariobased SRSs can be decomposed in related properties, and define heuristics for searching defect indicators that hurt these properties. We evaluate our work by applying our analysis approach to four case studies. The evaluation compares the results achieved by our tool-supported approach, with an inspection based approach and with related work.

[16_MSc_reis]
Eduardo de Jesus Coelho REIS. Anotação morfossintática a partir de contexto morfológico. [Title in English: Morphosyntactic annotation based on morphological context]. M.Sc. Diss. Port. Presentation: 27/09/2016. 91 p. Advisors: Ruy Luiz Milidiú.

Abstract: Part-of-speech tagging is one of the primary stages in natural language processing, providing useful features for performing higher complexity tasks. Word level representations have been largely adopted, either through a conventional sparse codification, such as bag-of-words, or through a distributed representation, like the sophisticated word embedded models used to describe syntactic and semantic information. A central issue on these codifications is the lack of morphological aspects. In addition, recent taggers present per-token accuracies around 97%. However, when using a persentence metric, the good taggers show modest accuracies, scoring around 55−57%. In this work, we demonstrate how to use n-grams to automatically derive morphological sparse features for text processing. This representation allows neural networks to perform POS tagging from a character-level input. Additionally, we introduce a regularization strategy capable of selecting specific features for each layer unit. As a result, regarding n-grams selection, using the embedded regularization in our models produces two variants. The first one shares globally selected features among all layer units, whereas the second operates individual selections for each layer unit, so that each unit is sensible only to the n-grams that better stimulate it. Using the proposed approach, we generate a high number of features which represent relevant morphosyntactic affection based on a character-level input. Our POS tagger achieves the accuracy of 96.67 % in the Mac-Morpho corpus for Portuguese.

[16_MSc_cruz]
Felipe João Pontes da CRUZ. Sistemas de recomendação utilizando Máquinas de Boltzmann restritas. [Title in English: Uma análise visual dos dados de GPS dos ônibus no Rio]. M.Sc. Diss. Port. Presentation: 23/02/2016. 53 p. Advisor: Ruy Luiz Milidiú.

Abstract: Recommender systems can be used in many problems in the real world. Many models were proposed to solve the problem of predicting missing entries in a specific dataset. Two of the most common approaches are neighborhood-based collaborative filtering and latent factor models. A more recent alternative was proposed on 2007 by Salakhutdinov, using Restricted Boltzmann Machines. This models belongs to the family of latent factor models, in which, we model latent factors over the data using hidden binary units. RBMs have shown that they can approximate solutions trained with a traditional matrix factorization model. In this work we'll revisit this proposed model and carefully detail how to model and train RBMs for the problem of missing ratings prediction.

[16_MSc_ismerio]
Fernando Cardoso ISMÉRIO. Wearables in core stabilization. [Title in Portuguese: StableBelt: wearables em estabilização segmentar]. M.Sc. Diss. Eng. Presentation: 21/03/2016. 130 p. Advisor: Hugo Fuks.

Abstract: In this dissertation, different types of audio biofeedback (ABF) for core stabilization exercises using motion sensors are investigated. Core stabilization exercises are one of the strategies used in the treatment of low back pain. The Supine Bridge (SB) exercise was chosen as the focus for the investigation. The primary motion sensor used was a tri-axial accelerometer. Flex Sensors, Force Sensitive Resistors and multiple accelerometers were also used in other prototypes. The results of this dissertation, which include data from accelerometer, comments, process, reflections, and implementation of prototypes that generate 3 types of audio biofeedback, were gathered during 5 cycles of action research. In action research, the researcher conducts the research performing successive actions that attempt to reduce a specific problem in a real world environment. In this dissertation, the environment chosen was a place where a patient executes exercises and the problem identified is the difficulty of the patient to perform the exercises correctly. The action was the introduction of a wearable – StableBelt – which generates audio biofeedback based on the patient’s movements during a core stabilization exercise. Different types of audio were investigated: instrumental music, piano and drums. The StableBelt was evaluated through 3 user tests. After a preliminary test with one participant, user tests with 5 and 8 participants were conducted. In the preliminary test, instrumental music was used and piano and drums in later tests. The last cycle of the action research was dedicated to the comfort of the StableBelt. During the investigation, physical therapists which research low back pain and physical therapists which use core stabilization exercises in their clinical practice were interviewed.

[16_PhD_silva]
Greis Francy Mireya SILVA CALPA. Estratégias para suporte à colaboração em sistemas presenciais para pessoas com Transtorno do Espectro Autista. [Title in Portuguese: Strategies to support collaboration in face-to-face systems for people with Autism Spectrum Disorders]. Ph.D. Thesis. Port. Presentation: 22/12/2016. 208 p. Advisor: Alberto Barbosa Raposo.

Abstract: Face-to-Face collaborative systems for people with autism spectrum disorders use strategies to motivate/force the collaboration among users. However, even the collaborative applications developed for this public, still do not consider notions of awareness for these users that present difficulties to understand the most basic concepts of a collaborative activity. Users with autism present difficulties to recognize and to interpret gestures and mental states of others, which restricts their capacity to understand implicit information that are essential to being aware of what is happening around them, and consequently, to perform the collaborative activities. In this work, we investigate some questions about how to offer awareness support, especially for users with low-functioning autism, in order to formulate and evaluate a set of collaborative strategies to support the design of more appropriate collaborative systems. For this purpose, we used the research-action methodology. Following this methodology, we perform four research cycles of action and reflection about proposed solutions, so that we could conceive the set of collaborative strategies proposed. In this cyclic process, we verified that collaborative systems shall offer awareness mechanisms in the interface (based on certain requirements) in different levels of approximation of the collaboration as well as activities to get users to know each dimension of collaboration, and gradually understanding it as a whole. These aspects compose the set of collaborative strategies conceived in this work.

[16_MSc_monteagudo]
Grettel MONTEAGUDO GARCIA. Analyzing, comparing and recommending conferences. [Title in Portuguese: Análise, comparação e recomendação de conferências]. M.Sc. Diss. Port. Presentation: 17/03/2016. 65 p. Advisor: Marco Antonio Casanova.

Abstract: This dissertation discusses techniques to automatically analyze, compare and recommend conferences, using bibliographic data, outlines an implementation of the proposed techniques and describes experiments with data extracted from a triplified version of the DBLP repository. Conference analysis applies statistical and social network analysis measures to the co-authorship network. The techniques for comparing conferences explore familiar similarity measures, such as the Jaccard similarity coefficient, the Pearson correlation similarity and the cosine similarity, and a new measure, the co-authorship network communities similarity index. These similarity measures are used to create a conference recommendation system based on the Collaborative Filtering strategy. Finally, the work introduces two techniques for recommending conferences to a given prospective author based on the strategy of finding the most related authors in the co-authorship network. The first alternative uses the Katz index, which can be quite costly for large graphs, while the second one adopts an approximation of the Katz index, which proved to be much faster to compute. The experiments suggest that the best performing techniques are: the technique for comparing conferences that uses the new similarity measure based on co-authorship communities; and the conference recommendation technique that explores the most related authors in the co-authorship network.

[16_MSc_descragnolle-taunay]
Henrique D'ESCRAGNOLLE-TAUNAY. A spatial partitioning heuristic for automatic adjustment of the 3D navigation speed in multiscale virtual environments. [Title in English: Streamline tracing for oil natural reservoirs based on adaptive numerical methods]. MSc. Diss. Eng. Presentation: 04/03/2016. 47 p. Advisor: Alberto Barbosa Raposo.

Abstract: With technological evolution, 3D virtual environments continuously increase in complexity; such is the case with multiscale environments, i.e., environments that contain groups of objects with extremely diverging levels of scale. Such scale variation makes it difficult to interactively navigate in this kind of environment since it demands repetitive and intuitive adjustments in their velocity or scale, according to the objects that are close to the observer, in order to ensure a comfortable and stable navigation. Recent effort have been developed working with heavy GPU based solutions that are not feasible depending of the complexity of the scene. We present a spatial partitioning heuristics for automatic adjustment of 3D navigation speed in a multiscale virtual environment minimizing the workload and transferring it to the CPU, allowing the GPU to focus on rendering.  Our proposal describes a geometrical strategy during the preprocessing phase that allows us to estimate in real-time phase which is the shortest distance between the observer and the object nearest to him. From this unique information, we are capable to automatically adjusting the speed of navigation according to the characteristic scale of the region where the observer is.   With the scene topological information obtained in a preprocessing phase, we are able to obtain, in real-time, the closest objects and the visible objects, which allow us to propose two different heuristics for automatic navigation velocity. Finally, in order to verify the usability gain in the proposed approaches, user tests were conducted to evaluate the accuracy and efficiency of the navigation, and users' subjective satisfaction.  Results where particularly significant for demonstrating accuracy gain in navigation while using the proposed approaches for both laymen and advanced users.

[16_MSc_bistene]
Joanna Pivatelli BISTENE. A contratação de tecnologia da informação na Administração Pública Federal: o caso do desenvolvimento de software sob demanda. [Title in English: Information technology acquisition in Brazilian Federal Government: the case of on-demand software development]. MSc. Diss. Eng. Presentation: 27/09/2016. 253 p. Advisor: Julio Cesar Sampaio do Prado Leite.

Abstract: Acquisition of Information Technology (IT) by the Brazilian Federal Government is governed by law. In the specific case, the Law 8.666/1993 is intend to establish the rules for such contracts, forcing their planning. The Requirements Engineering literature emphasizes evolves in definition process but this is often disregard. Therefore, exists a clear conflict in requirements definition during the IT acquire in Brazilian Federal Government with current legislation. Define requirements obligation before software procurement is impose by law and can generate problems in contract management. This dichotomy among the mutability requirements and legal rigidity in the procurement process had inspired an exploratory research with public organizations. Our research provide transparency in problems experienced by these agencies in procurement of IT solutions. We prepared a preliminary analysis of these problems and pointed out possible solutions.

[16_MSc_aguiar]
José Luiz do Nascimento AGUIAR. Medidas de similiaridade entre séries temporárias. [Title in English: Time series similarity measuresa abordagem baseada em blueprints para priorização e classificação de anomalias de código críticas]. MSc. Diss. Port. Presentation: 11/03/2016. 75 p. Advisor: Eduardo Sany Laber.

Abstract: Nowadays a very important task in data mining is to understand how to collect the most informative data in a very amount of data. Once every single feld of knowledge have lots of data to summarize in the most representative information, the time series approach is de nitely a very strong way to represent and collect this information from it. On other hand we need to have an appropriate tool to extract the most signifiant data from this time series. To help us we can use some similarity methods to know how similar is one time series from another In this work we will perform a research using some distance-based similarity methods and apply it in some clustering algorithms to do an assessment to see if there is a combination (distance-based similarity methods / clustering algorithm) that present a better performance in relation with all the others used in this work or if there exists one distancebased similarity method that shows a better performance between the others.

[16_MSc_nascimento]
Leonardo Henrique Camello do NASCIMENTO. Um estudo de presença em uma aplicação de realidade virtual para tratamento de pessoas com medo de voar. [Title in English: A presence study in a virtual reality application for the treatment of people with fear of flying]. MSc. Diss. Port. Presentation: 29/04/2016. 63 p. Advisor: Alberto Barbosa Raposo.

Abstract: Fear of flying is a real problem that affects 10% to 25% of the world’s population. Approximately 25% of adults experience a significant increase in their anxiety levels when required to take any type of air transport and 10% of them avoid the situation. The approach that has proven to be the most effective in the treatment of phobias is in vivo exposure. However, the difficulty and the cost, and sometimes even the danger, of using real airplanes and real flights to expose people with fear of flying to these stimuli have daunted many researchers, therapists, and patients despite the prevalence and the impact of the fear of flying. We present in this study a virtual reality application that promotes a systematic exposure to the stimuli that causes significant increase in anxiety levels related to fear of flying through computer generated environments. This application uses the concept of immersion through the Oculus Rift to promote an “almost real” experience to the patients. To evaluate the proposed application, in special the “sense of presence” caused by it, we obtained qualitative data from interviews and questionnaires with its “meta-users”, i.e., the psychiatrists that will apply the treatment to their patients.

[16_PhD_duarte]
Leonardo Seperuelo DUARTE. TopSim: a plugin-based framework for large-scale numerical analysis. [Title in Portuguese: TopSim: um sistema baseado em plugin para análise numérica em larga escala]. Ph.D. Thesis. Eng. Presentation: 09/09/2016. 91 p. Advisor: Waldemar Celes Filho.

Abstract: Computational methods in engineering are used to solve physical problems that do not have analytical solution or their perfect mathematical representation is unfeasible. Numerical techniques, including the largely used finite element method, require the solution of linear systems with hundreds of thousands equations, demanding high computational resources (memory and time). In this thesis, we present a plugin-based framework for large-scale numerical analysis. The framework is used as an original tool to solve topology optimization problems using the finite element method with millions of elements. Our strategy uses an element-by-element technique to implement a highly parallel code for an iterative solver with low memory consumption. Besides, the plugin approach provides a fully flexible and easy to extend environment, where different types of applications, requiring different types of finite elements, materials, linear solvers, and formulations, can be developed and improved. The kernel of the framework is minimum with only a plugin manager module, responsible to load the desired plugins during runtime using an input configuration file. All the features required for a specific application are defined inside plugins, with no need to change the kernel. Plugins may provide or require additional specialized interfaces, where other plugins may be connected to compose a more complex and complete system. We present results for a structural linear elastic static analysis and for a structural topology optimization analysis. The simulations use elements Q4, hexahedron (Brick8), and hexagonal prism (Honeycomb), with direct and iterative solvers using sequential, parallel and distributed computing. We investigate the performance regarding the use of memory and the scalability of the solution for problems with different sizes, from small to very large examples on a single machine and on a cluster. We simulated a linear elastic static example with 500 million elements on 300 machines.

[16_MSc_millan]
Liander MILLÁN FERNÁNDEZ. Concurrent programming in Lua - revisiting the Luaproc library. [Title in Portuguese: TopSim: um sistema baseado em plugin para análise numérica em larga escala]. M. Sc. Diss. Eng. Presentation: 16/12/2016. 68 p. Advisor: Noemi da La Roque Rodriguez.

Abstract: In recent years, the tendency to increase the performance of a microprocessor, as an alternative solution to the increasing demand for computational resources of both applications and systems, has decreased significantly. This has led to an increase of the interest in employing multiprocessing environments. Although many models and libraries have been developed to offer support for concurrent programming, ensuring that several
execution ows access shared resources in a controlled way remains a complex task. The Luaproc library, which provides support for concurrency in Lua, has shown some promise in terms of performance and cases of use. In this thesis, we study the Luaproc library and incorporate to it new functionalities in order to make it more user friendly and extend its use to new scenarios. First, we introduce the motivations to our extensions to Luaproc, discussing alternative ways of dealing with the existing limitations. Then, we present requirements, characteristics of the implementation, and limitations associated to each of the mechanisms implemented as alternative solutions to these limitations. Finally, we employ the incorporated functionalities in implementing some concurrent applications, in order to evaluate the performance and test the proper functioning of such mechanisms.

[16_MSc_talavera]
Luis Eduardo TALAVERA RIOS. An energy-aware IoT gateway, with continuous processing of sensor data. [Title in Portuguese: Um enegy-aware IoT gateway, com processamento contínuo de dados de sensor]. MSc. Diss. Eng. Presentation: 16/03/2016. 73 p. Advisor: Markus Endler.

Abstract: Few studies have investigated and proposed a middleware solution for the Internet of Mobile Things (IoMT), where the smart things (Smart Objects) can be moved, or else can move autonomously, but remain accessible from any other computer over the Internet. In this context, there is a need for energy-ecient gateways to provide connectivity to a great variety of Smart Objects. Proposed solutions have shown that mobile devices (smartphones and tablets) are a good option to become the universal intermediates by providing a connection point to nearby Smart Objects with short-range communication technologies. However, they only focus on the transmission of raw sensor data (obtained from connected Smart Objects) to the cloud where processing (e.g. aggregation) is performed. Internet Communication is a strong battery-draining activity for mobile devices; moreover, bandwidth may not be sucient when large amounts of information is being received from the Smart Objects. Hence, we argue that some of the processing should be pushed as close as possible to the sources. In this regard, Complex Event Processing (CEP) is often used for real-time processing of heterogeneous data and could be a key technology to be included in the gateways. It allows a way to describe the processing as expressive queries that can be dynamically deployed or removed on-the-y. Thus, being suitable for applications that have to deal with dynamic adaptation of local processing. This dissertation describes an extension of a mobile middleware with the inclusion of continuous processing of sensor data, its design and prototype implementation for Android. Experiments have shown that our implementation delivers good reduction in energy and bandwidth consumption.

[16_MSc_netto]
Luiz Felipe NETTO.  Algoritmo de corte com preservação de contexto para visualização de modelos de reservatório. [Title in English: Cutaway algorithm with context preservation for reservoir model visualization]. MSc. Diss. Port. Presentation: 16/09/2016. 71 p. Advisor: Waldemar Celes Filho.

Abstract: Numerical simulation of black oil reservoir is widely used in the oil and gas industry. The reservoir is represented by a model of hexahedral cells with associated properties, and the numerical simulation is used to predict the fluid behavior in the model. Specialists make analysis of such simulations by inspecting, in a graphical interactive environment, the tridimensional model. In this work, we propose a new cutaway algorithm with context preservation to help the inspection of the model. The main goal is to allow the specialist to visualize the wells and their vicinity. The wells represent the object of interest that must be visible while preserving the tridimensional model (the context) in the vicinity as far as possible. In this way, it is possible to visualize the distribution of cell property together with the object of interest. The proposed algorithm makes use of graphics processing units and is valid for arbitrary objects of interest. It is also proposed an extension to the algorithm to allow the cut section to be decoupled from the camera, allowing analysis of the cut model from different points of view. The effectiveness of the proposed algorithm is demonstrated by a set of results based on actual reservoir models.

[16_MSc_silva]
Luiz José Schirmer SILVA. CrimeVis: an interactive visualization system for analyzing criminal data in the state of Rio de Janeiro. [Title in Portuguese: CrimeVis: um sistema interativo de visualização para análise de dados criminais do estado do Rio de Janeiro]. MSc. Diss. Eng. Presentation: 02/06/2016. 53 p. Advisor: Hélio Cortes Vieira Lopes.

Abstract: This work presents the development of an interactive graphic visualization system for analyzing criminal data in the State of Rio de Janeiro, provided by the Public Safety Institute from the State of Rio de Janeiro (ISP-RJ, Instituto de Segurança Pública). The system presents to the user a set of integrated tools that support visualizing and analyzing statistical data on crimes, which make it possible to infer relevant information regarding government policies on public safety and their effects. The tools allow us to visualize multidimensional data, spatiotemporal data, and multivariate data in an integrated manner using brushing and linking techniques. The work also presents a case study to evaluate the set of tools we developed.

[16_MSc_mota]
Marcelo Garnier MOTA. Exploring structured information retrieval for bug localization in C# software projects. [Title in Portuguese: Explorando recuperação de informação estruturada para localização de defeitos em projetos de software C#]. MSc. Diss. Eng. Presentation: 16/09/2016. 91 p. Advisor: Alessandro Fabrício Garcia.

Abstract: Software projects can grow very rapidly, reaching hundreds or thousands of files in a relatively short time span. Therefore, manually finding the source code parts that should be changed in order to fix a bug is a difficult task. Static bug localization techniques provide effective means of finding files related to a bug. Recently, structured information retrieval has been used to improve the effectiveness of static bug localization, being successfully applied by techniques such as BLUiR, BLUiR+, and AmaLgam. However, there are significant shortcomings on how these techniques were evaluated. BLUiR, BLUiR+, and AmaLgam were tested only with four projects, all of them structured with the same language, namely, Java. Moreover, the evaluations of these techniques (i) did not consider appropriate program versions, (ii) included bug reports that already mentioned the bug location, and (iii) did not exclude spurious files, such as test files. These shortcomings suggest the actual effectiveness of these techniques may be lower than reported in recent studies. Furthermore, there is limited knowledge on whether and how the effectiveness of these state-of-the-art techniques can be improved. In this dissertation, we evaluate the three aforementioned techniques on 20 open-source C# software projects, providing a rigorous assessment of the effectiveness of these techniques on a previously untested object-oriented language. Moreover, we address the simplistic assumptions commonly present in bug localization studies, thereby providing evidence on how their findings may be biased. Finally, we study the contribution of different program construct types to bug localization. This is a key aspect of how structured information retrieval is applied in bug localization. Therefore, understanding how each construct type influences bug localization may lead to effectiveness improvements in projects structured with a specific programming language, such as C#.

[16_MSc_martins]
Marcelo Malta Rodrigues MARTINS. Strong lower bounds for the CVRP via column and cut generation. [Title in Portuguese: Limites inferiores fortes para o CVRP via geração de colunas e cortes]. MSc. Diss. Eng. Presentation: 18/01/2016. 67 p. Advisor: Marcus Vinicius Soledade Poggi de Aragão.

Abstract: The Capacitated Vehicle Routing Problem (CVRP) is the seminal version of the vehicle routing problem, a classical problem in Operational Research. Introduced by Dantzig e Ramser, the CVRP generalizes the Traveling Salesman Problem (TSP) and the Bin Packing Problem (BPP). In addition, routing problems arise in several real world applications, often in the context of reducing costs, polluent emissions or energy within transportation activities. In fact, the cost with transportation can be roughly estimated to represent 5% to 20% of the overall cost of a delivered product. This means that any saving in routing can be much relevant. The CVRP is stated as follows: given a set of n + 1 locations – a depot and n customers – the distances between every pair of locations, integer demands associated with each customer, and a vehicle capacity, we are interested in determining the set of routes that start at the depot, visits each customer exactly once and returns to the depot. The total distance traveled by the routes should be minimized and the sum of the demands of customers on each route should not exceed the vehicle capacity. This work considers that the number of available vehicles is given. State of the art algorithms for finding and proving optimal solutions for the CVRP compute their lower bounds through column generation and improving it by adding cutting planes. The columns generated may be elementary routes, where customers are visited only once, or relaxations such as q-routes and the more recent ng-routes, which differ on how they allow repeating customers along the routes. Cuts may be classified as robust, those that are defined over arc variables, and non-robust (or strong), those that are defined over the column generation master problem variables. The term robust used above refers to how adding the cut modifies the efficiency of solving the pricing problem. Besides the description above, the most efficient exact algorithms for the CVRP use too many elements turning its replication a hard long task. The objective of this work is to determine how good can be lower bounds computed by a column generation algorithm on ng-routes using only capacity cuts and a family of strong cuts, the limited memory subset row cuts. We assess the leverage achieved with the consideration of this kind of strong cuts and its combination with others techniques like Decremental Space State Relaxation (DSSR), Completion Bounds, ng-Routes and Capacity Cuts over a Set Partitioning formulation of the problem. Extensive computational experiments are presented along with an analysis of the results obtained.

[16_MSc_silva]
Marcos Vinícius Marques da SILVA. VelvetH-DB:
uma abordagem robusta de banco de dados no processo de montagem de fragmentos de sequências biológicas. [Title in English: VelvetH-DB: a robust database approach for the assembly process of biological sequences]. MSc. Diss. Port. Presentation: 30/03/2016. 66 p. Advisor: Sergio Lifschitz.

Abstract: Recent technological advances, both in assembly algorithms and in sequencing methods, have enabled the reconstruction of whole DNA even without a reference genome available. The assembly of the complete chain involves reading a large volume of genome fragments, called short-reads, which makes the problem a significant computational challenge. A major bottleneck for all existing fragment assembly algorithms is the high consumption of RAM. This dissertation intends to study the implementation of one of these algorithms, called Velvet, which is widely used and recommended. The same possessed a module, VelvetH that performs a pre-processing data with the aim of reducing the consumption of main memory. After a thorough study of code improvements and alternatives, specific changes have been made and proposed a solution with data persistence in secondary memory in order to obtain effectiveness and robustness.

[16_PhD_ferreira]
Marilia Guterres FERREIRA. Anticipating change in software systems supporting organizational Information systems using an organizational based strategic perspective. [Title in Portuguese: Antecipando mudanças em sistemas de software que suportam Sistemas de Informação Organizacionais usando uma perspectiva estratégica baseada em organizações]. Ph.D. Thesis. Eng. Presentation: 18/10/2016. 240 p. Advisor: Julio Cesar Sampaio do Prado Leite.

Abstract: Keeping organizations and their Software Systems supporting Organizational Information Systems (SSsOIS) aligned over time is a complex endeavour. We believe understanding the organizational dynamics of changes, and of the impacts these changes might cause, can support the evolution of SSsOIS. Yet, reasoning about the organizational changes in advance also supports the development of an SSsOIS more likely to be aligned to the dynamics of the organization. Based on it, we ground our work on strategic management theory, which reasons about possible futures of the organization and formulates strategies to achieve new goals in these possible futures. We propose to apply the outcomes of strategic management to prepare SSsOIS for the future, i.e. to prepare SSsOIS for these new requirements raised from the strategic plans. For this, we use scenario planning as a tool to support key people in the organization to think about multiple possible futures and plan strategies. In order to keep the strategic planning of the organization aligned to the SSsOIS, we propose an Organizational Dynamics-based Approach for Requirements Elicitation (ODA4RE) composed by a scenario-based strategic planning (SSP), organizational impact analysis (OIA), and validation of the likely SSsOIS’ requirements (LSRV). OIA also introduces an organizational dynamics metamodel (ODMM) on which to base the reasoning, and an organizational dynamics questions set (ODQS) to explore likely organizational impacts. We evaluate our proposal in four empirical studies with different purposes: first in an academic organization in Rio de Janeiro to analyse specifically the SSP, second in a workshop to evaluate the ODMM’s expressiveness, third in a Post Office branch in London to analyse OIA, and finally the entire approach at a Brazilian research university. Results show contributions towards SSsOIS’ requirements evolution as they align with the organization plans.

[16_MSc_masson]
Matheus Manhães MASSON. Cold Start em recomendação de músicas utilizando deep learning. [Title in English: Cold Start in music recommendation using deep learning]. MSc. Diss. Eng. Presentation: 23/08/2016. 61 p. Advisor: Ruy Luiz Milidiú.

Abstract: Recomender system are used to provide information or Products by learnig the profile of their users automatically using Machine Learning techniques. Tipically these systems are based on previously collected data on their products and users. When there are no previus data these systems do not work; this problem is called Cold Start Problem. This work is focused on the Cold Start Problem that affects the quality of the recommendations and the failure to recomend new songs by methods traditionally used. For this solution we use deep learning with Audio of the Songs and thus extract useful features to recommend. From Latent Factors obtained by Matrix Factorization a convolutional Neural network is trained to learn these factors using the Audio. Thus the network can be used to predict Latent Factors of Songs using only the audio without the need of previous data.  This becomes a viable solution to the Cold Start Problem. The result shows that this is a workable solution to the problem even if they did not reach the best metrics of traditional methods. The Convolutional Network trained learns from the Audio and preditcs factors. Thus the result allows to recommend new songs and may even increase recommendation using hybrid methods.

[16_PhD_viana]
Marx Leles VIANA. Design e implementação de agentes de software adaptativos normativos. [Title in English: Design and implementation of adaptive normative software agents]. Ph. D. Thesis. Port. Presentation: 05/12/2016. 157 p. Advisor: Carlos José Pereira de Lucena.

Abstract: Multi-agent systems have been introduced as a new paradigm for conceptualizing, designing and implementing software systems that are becoming increasingly complex, open, distributed, dynamic, autonomous and highly interactive. However, agent-oriented software engineering has not been widely adopted, mainly due to lack of modeling languages that are expressive and comprehensive enough to represent relevant agent-related abstractions and support the refinement of design models into code. Most modeling languages do not define how these abstractions interact at runtime, but many software applications need to adapt their behavior, react to changes in their environments dynamically, and align with some form of individual or collective normative application behavior (e.g., obligations, prohibitions and permissions). In this paper, we propose a metamodel and an architecture approach to developing adaptive normative agents. We believe the proposed approach will advance the state of the art in agent systems so that software technologies for dynamic, adaptive, norm-based applications can be designed and implemented.

[16_MSc_teixeira]
Otávio Freitas TEIXEIRA. Auto-Sintonia para sistemas de bancos de dados na nuvem. [Title in English: Database self-tuning in the Cloud]. MSc. Diss. Eng. Presentation: 31/03/2016. 63 p. Advisor: Sergio Lifschitz.

Abstract: Cloud computing is changing the way users access and benefit from computer services. A database manager is one of the main features of this new working environment. However, large volumes of data must be properly managed and made available, according to the fluctuations in workloads and function of new and existing parameters. Because of dimensions problems in this new cloud environment, it is very difficult to have a DBA who can manually manage, maintain availability and performance acceptably. In particular, the necessity of a tuning process automatic in the cloud system to meet contractual operation requirements and the necessity of offering to the user resources as if they were unlimited while with excellent performance. This thesis explains and compares the activities of (self)-tuning database systems operating in conventional and cloud environments, emphasizing the differences observed in the cloud service provider's view and users in a context of DBaaS. In particular, it is proposed to extend of tuning ontology in order to automate actions to tuning the Database as a Service.

[16_PhD_ribeiro]
Paula Ceccon RIBEIRO. Uncertainty Analysis of 2D vector fields through the Helmholtz-Hodge Decomposition. [Title in Portuguese: Análise de incertezas em campos vetoriais 2D com o uso da Decomposição de Helmholtz-Hodge]. Ph.D. Thesis Eng. Presentation: 15/12/2016. 109 p. Advisor: Hélio Cortes Vieira Lopes.

Abstract: Vector eld plays an essential role in a large range of scientific applications. They are commonly generated through computer simulations. Such simulations may be a costly process because they usually require high
computational time. When researchers want to quantify the uncertainty in such kind of applications, usually an ensemble of vector fields realizations are generated, making the process much more expensive. The Helmholtz-
Hodge Decomposition is a very useful instrument for vector field interpretation because it traditionally distinguishes conservative (rotational-free) components from mass-preserving (divergence-free) components. In this work, we are going to explore the applicability of such technique on the uncertainty analysis of 2-dimensional vector fields. First, we will present an approach of the use of the Helmholtz-Hodge Decomposition as a basic tool for the analysis of a vector field ensemble. Given a vector field ensemble E, we firstly obtain the corresponding rotational-free, divergence-free and harmonic component ensembles by applying the Natural Helmholtz-Hodge Decomposition to each vector field in E. With these ensembles in hand, our proposal not only quantifies, via a statistical analysis, how much each component ensemble is point-wisely correlated to the original vector field ensemble, but it also allows to investigate the uncertainty of rotational-free, divergence-free and harmonic components separately. Then, we propose two techniques that jointly with the Helmholtz-Hodge Decomposition stochastically generate vector fields from a single realization. Finally, we propose a method to synthesize vector fields from an ensemble, using both the Dimension Reduction and Inverse Projection techniques. We test the proposed methods with synthetic vector fields as well as with simulated vector fields.

[16_MSc_souzafilho] *
Paulo Roberto Pereira de SOUZA FILHO. Auxílio a portabilidade de código em aplicações de alto desempenho. [Title in English: Support for code portability in high performance computing applications]. MSc. Diss. Port. Presentation: 21/03/2016. 117 p. Advisor: Noemi da La Roque Rodriguez.

Abstract: Today’s platforms are becoming increasingly heterogeneous. A given platform may have many different computing elements in it: CPUs, coprocessors and GPUs of various kinds. This work propose a way too keep some portion of code portable without compromising the performance along different heterogeneous platforms. We implemented the HLIB library that handles the preparation code needed by heterogeneous computing, also this library transparently supports the traditional homogeneous platform. To address multiple SIMD architectures we implemented the OpenVec, a tool to help compiler to enable SIMD instructions. This tool provides a set of portable SIMD intrinsics and C++ operators to get a portable explicit vectorization, covering SIMD architectures from the last 17 years like ARM Neon, Intel SSE to AVX-512 and IBM Power8 Altivec+VSX. We demonstrated the combination use of this strategy using both tools with petaflop HPC applications.

[16_MSc_dunker]
Philip Kuster DUNKER. Uma ferramenta de telepresença de baixo custo usando Oculus Rift: desenvolvimento e avaliação num cenário de videoconferência. [Title in English: A low-cost telepresence tool using Oculus Rift: development and evaluation in a videoconference scenarium]. MSc. Diss. Port. Presentation: 06/04/2016. 70 p. Advisor: Alberto Barbosa Raposo.

Abstract: Telepresence refers to a set of technologies that allows a person to feel as if he is in a place other than his true location. Such equipment uses an ordinary camera to film what is happening in an environment and transmits it alive on televisions or monitors to the user in another environment. Sometimes the cameras can be controlled through devices such as keyboards or joysticks. This dissertation presents a tool composed of a head-mounted display (HMD), we used the Oculus Rift DK1, integrated with a device called remote head, able to film with a stereo camera and to transmit to the Oculus Rift the images in 3D. At the same time, the HMD’s gyroscope captures the user's head orientation and sends it to the remote head, which has servo motors able to rotate it in order to allow the user to move the stereo camera without any additional device. The project's goal is to provide the user an experience of immersive telepresence, with a low cost and a simple interface. Some tests with users were performed and indicated the benefit of the tool for videoconferencing.

[16_PhD_nasser]
Rafael Barbosa NASSER. Uma plataforma na nuvem para armazenamento de dados georreferenciados de mobilidade urbana. [Title in English: A cloud computing platform for storing georeferenced mobility data]. MSc. Diss. Port. Presentation: 26/09/2016. 159 p. Advisor: Hélio Cortes Vieira Lopes.

Abstract:  The quality of life in urban centers has been a concern for governments, business and the resident population in general. Public transportation services perform a central role in this discussion, since they determine, especially for that layer of lower-income society, the time wasted daily in their movements. In Brazilian cities, city buses are predominant in public transportion. Users of this service - passengers - do not have updated information of buses and lines. Offer this kind of information contributes to a better everyday experience of this modal and therefore provides greater quality of life for its users. In a broader view, the bus can be considered sensors that enable the understanding of the patterns and identify anomalies in vehicle traffic in urban areas, allowing benefits for the whole population. This work presents a platform in the cloud computing environment that captures, enriches, stores and makes available the data from GPS devices installed on buses, allowing the extraction of knowledge from this valuable and voluminous set of information. Experiments are performed with the buses of the Municipality of Rio de Janeiro, with applications focused on passenger and society. The methodologies, discussions and techniques used throughout the work can be reused for different cities, modal and perspectives.

[16_MSc_portugal]
Roxana Lisette Quintanilha PORTUGAL. Mineração de informação em linguagem natural para apoiar a elicitação de requisitos. [Title in English: Mining information in natural language to support requirements elicitation]. MSc. Diss. Port. Presentation: 19/04/2016. 96 p. Advisor: Julio Cesar Sampaio do Prado Leite.

Abstract:  This work describes the mining of information in natural language from the GitHub repository. It is explained how the content of similar projects, given a search domain, can be useful for the reuse of knowledge and thus help in the Requirement Elicitation tasks. Techniques of text mining, regularities independent from domain, and GitHub metadata are the method used to select relevant projects and the information within them.  One approach to achieve our goal is explained with an exploratory research and the results achieved.

[16_MSc_pinheiro]
Sasha Nicolas da Rocha PINHEIRO. Calibração de câmera usando projeção frontal-paralela e colinearidade dos pontos de controle. [Title in English: Camera calibration using fronto parallel projection and collinearity of control points]. MSc. Diss. Port. Presentation: 01/09/2016. 63 p. Advisor: Alberto Barbosa Raposo.

Abstract:  Crucial for any computer vision or augmented reality application, the camera calibration is the process in which one gets the intrinsics and the extrinsics parameters of a camera, such as focal length, principal point and distortions values. Nowadays, the most used method to deploy the calibration comprises the use of images of a planar pattern in different perspectives, in order to extract control points to set up a system of linear equations whose solution represents the camera parameters, followed by an optimization based on the 2D reprojection error. In this work, the ring calibration pattern was chosen because it offers higher accuracy on the detection of control points. Upon application of techniques such as fronto-parallel transformation, iterative refinement of the control points and adaptative segmentation of ellipses, our approach has reached improvements in the result of the calibration process. Furthermore, we proposed extend the optimization model by modifying the objective function, regarding not only the 2D reprojection error but also the 2D collinearity error.

[16_MSc_fiolgonzalez]
Sonia FIOL GONZÁLEZ. A novel committee-based clustering method. [Title in Portuguese: Um novo método de agrupamento baseado em comitê]. MSc. Diss. Eng. Presentation: 15/09/2016. 70 p. Advisor: Hélio Cortes Vieira Lopes.

Abstract:  In data analysis, in the process of quantitative modeling and in the construction of decision support models, clustering, classification and information retrieval algorithms are very useful. For these algorithms it is crucial to determine the relevant features in the original dataset. To deal with this problem, techniques for feature selection play an important role. Moreover, it is recognized that in unsupervised learning tasks it is also diffcult to define the correct number of clusters. This research proposes a method based on ensemble methods using all features from a dataset and varying the number of clusters to calculate the similarity matrix between any two instances of the dataset. Each element in this matrix stores the probability of the corresponding instances to be in the same cluster in these multiple scenarios. Notice that the similarity matrix might be transformed to a distance matrix to be used in other clustering methods. The experiments were made with a real-world dataset of the crimes in Rio de Janeiro Capital showing the effectiveness of the proposed technique.

[16_MSc_silvaneto]
Vicente Correa SILVA NETO. Uma plataforma de jogos JRPG destinada à educação com entretenimento. [Title in English: A JRPG game platform with the purpose of education and entertainment]. MSc. Diss. Port. Presentation: 22/09/2016. 82 p. Advisor: Waldemar Celes Filho.

Abstract: In this project, inspired by the fields of Pedagogy and Entertainment, we aim to develop a digital games development framework in order to facilitate the creation of educational games of the sub-genre JRPG (Japanese Role-Playing Games), more interesting than the majority of educational games available for now. The RPG genre is, by definition, based in storytelling and role-playing principles, identified by the literature as important tools that stimulates the students imagination, engage them emotionally and arouse their interests for the traditional educational program. The subgenre JRPG, in turn, represents a special category of eletronic RPGs that inherit those same educational principles, but have well defined delimitations in respect of game mechanics and artistic identity. These delimitations are positive in a sense that they work as guidelines for the development process of this kind of games.

[16_PhD_segura]
Vinicius Costa Villas Bôas SEGURA. BONNIE: Building Online Narratives from Noteworthy Interaction Events. [Title in Portuguese: BONNIE: Construindo narrativas online a partir de eventos de interacão relevantes]. MSc. Diss. Eng. Presentation: 26/09/2016. 154 p. Advisor: Simone Diniz Junqueira Barbosa.

Abstract: Nowadays, we have access to data of unprecedentedly large size, high dimensionality, and complexity. To extract unknown and unexpected information from such complex and dynamic data, we need effective and efficient strategies. One such strategy is to combine data analysis and visualization techniques, which is the essence of visual analytics applications. After the knowledge discovery process, a major challenge is to filter the essential information that led to a discovery and to communicate the findings to other people. We propose to take advantage of the trace left by the exploratory data analysis, in the form of user interaction history, to aid in this process. With the trace, the user can choose the desired interaction steps and create a narrative, sharing the acquired knowledge with readers. To achieve our goal, we have developed the BONNIE (Building Online Narratives from Noteworthy Interaction Events) framework. The framework comprises a log model to register the interaction events, auxiliary code to help the developer instrument his or her own code, and an environment to view the user's own interaction history and build narratives. This thesis presents our proposal for communicating discoveries in visual analytics applications, the BONNIE framework, and a few empirical studies we conducted to evaluate our solution.

[16_MSc_costa]
Vinícius de Lima COSTA. Uma ferramenta de RV para tratamento de fobia de voar controlada pelo terapeuta. [Title in English: A VR tool for fear of flying treatment controlled by the therapist]. MSc. Diss. Port. Presentation: 27/10/2016. 47 p. Advisor: Alberto Barbosa Raposo.

Abstract: The problem known as fear of flying is common nowadays. Also known by other names such as aviophobia or aerophobia, this kind of fear can be defined as “ a specific phobia noted by a persistent excessive fear of travelling or possibility of travel through the air”. Many people suffer from this kind of phobia, creating a high demand for treatments in this area. The most effective way to treat someone is by in vivo exposition. However, this kind of treatment is usually expensive, since there is a need to go to an airport and to get aboard a plane. At the end, the patient may not even try to go through with the flight because of his/her excessive fear. The present work focuses on creating a 3D virtual reality flight simulator, from the passenger point of view. In addition to this simulator, there is also a mobile application that controls the current state of the main application and the stimulus that can be passed to the patient without interrupting the immersion on the main application. The effectiveness of the virtual reality application in transmitting the sense of fear and the effectiveness of the mobile application were evaluated with the help of psychiatrists from IPUB/UFRJ and a pilot test, plus a presentation to PUC-Rio psychiatrists.

[16_MSc_barroso]
Yanely Milanés BARROSO. Structured learning with incremental feature induction and selection for Portuguese dependency parsing. [Title in Portuguese: Aprendizado estruturado com indução e seleção incrementais de atributos para análise de dependência em Português]. MSc. Diss. Eng. Presentation: 09/03/2016. 92 p. Advisor: Ruy Luiz Milidiú.

Abstract: Natural language processing requires solving several tasks of increasing complexity, which involve learning to associate structures like graphs and sequences to a given text. For instance, dependency parsing involves learning of a tree that describes the dependency-based syntactic structure of a given sentence. A widely used method to improve domain knowledge representation in this task is to consider combinations of features, called templates, which are used to encode useful information with nonlinear pattern. The total number of all possible feature combinations for a given template grows exponentialy in the number of features and can result in computational intractability. Also, from an statistical point of view, it can lead to overfitting. In this scenario, it is required a technique that avoids over fitting and that reduces the feature set. A very common approach to solve this task is based on scoring a parse tree, using a linear function of a defined set of features. It is well known that sparse linear models simultaneously address the feature selection problem and the estimation of a linear model, by combining a small subset of available features. In this case, sparseness helps control over fitting and performs the selection of the most informative features, which reduces the feature set. Due to its exibility, robustness and simplicity, the perceptron algorithm is one of the most popular linear discriminant methods used to learn such complex representations. This algorithm can be modified to produce sparse models and to handle nonlinear features. We propose the incremental learning of the combination of a sparse linear model with an induction procedure of non-linear variables in a structured prediction scenario. The sparse linear model is obtained through a modifications of the perceptron algorithm. The induction method is the Entropy-Guided Feature Generation. The empirical evaluation is performed using the Portuguese Dependency Parsing data set from the CoNLL 2006 Shared Task. The resulting parser attains 92.98% of accuracy, which is a competitive performance when compared against the state-of-art systems. On its regularized version, it accomplishes an accuracy of 92.83%, shows a striking reduction of 96.17% in the number of binary features and reduces the learning time in almost 90%, when compared to its non regularized version.