Doutorado em Ciência da Computação

URI Permanente para esta coleção

http://repositorio.ufes.br/handle/10/17626

Nível: Doutorado
Ano de início:
Conceito atual na CAPES:
Ato normativo:
Periodicidade de seleção:
Área(s) de concentração:
Url do curso:

Navegar

Agora exibindo 1 - 20 de 52

A commitment-based reference ontology for service: harmonizing service perspectives
(Universidade Federal do Espírito Santo, 2014-12-10) Nardi, Julio Cesar; Almeida, João Paulo Andrade; Falbo, Ricardo de Almeida; Pires, Luiz Ferreira; Amorim, Fernanda Araújo Baião; Guizzardi, Renata Silva Souza; Barcellos, Monalessa Perini
Nowadays, the notion of service has been widely adopted in the practice of economic sectors (e.g., Service, Manufacturing, and Extractive sectors), as well as, in the research focus of various disciplines (e.g., Marketing, Business, and Computer Science). Due to that, a number of research initiatives (e.g., service ontologies, conceptual models, and theories) have tried to understand and characterize the complex notion of service. However, due to particular views of these disciplines and economic sectors, a number of different characterizations of service (e.g., “service as interaction”, “service as value co-creation”, and “service as capability / manifestation of competence”, among others) have been proposed. The existence of these various non-harmonized characterizations, and the focus on a terminological debate about the “service” concept, instead of about the service phenomena from a broad perspective, make the establishment of a unified body of knowledge for service difficult. This limitation impacts, e.g., the establishment of unified conceptualization for supporting the smooth alignment between Business and IT views in service-oriented enterprise architecture (SoEA), and the design and usage of service modeling languages. In this thesis we define a theoretical foundation for service based on the notion of service commitment and claims as basic elements in the characterization of service relations along service life cycle phases (service offer, service negotiation, and service delivery). As discussed in this work, this theoretical foundation is capable of harmonizing a number of service perspectives found in the literature. Such theoretical foundation is specified in a well-founded core reference ontology, named UFO-S, which was designed by adopting a sound ontological engineering apparatus (mainly, a wellfounded ontology representation language, OntoUML, and approaches of model verification and model validation). As a kind of “theory”, UFO-S was applied in the analysis of SoEA structuring principles in order to define a “commitment-based SoEA view”, which remarks social aspects inherent in service relations usually underexplored in widely adopted service-oriented approaches (such as SOA-RM by OASIS, ITIL, and ArchiMate). Based on this, UFO-S was also applied in an ontological analysis of service modeling at ArchiMate’s Business layer. Such ontological analysis showed some limitations concerned to semantic ambiguity and lack of expressiveness for representing service offerings (and type thereof) and service agreements in SoEA. In order to address these limitations, three service modeling patterns (service offering type pattern, service offering pattern, and service agreement pattern) were proposed taking as basis UFO-S. The usefulness of these patterns for addressing these limitations was evidentiated by means of an empirical evaluation. Finally, we can say that, beyond offering a broad and well-founded theoretical foundation for service able to harmonize service perspectives, UFO-S presented benefits as a reference model in the analysis of SoEA structuring principles, and in the (re)design of service modeling languages.
A conceptual architecture and a framework for dealing with variability in mulsemedia systems
(Universidade Federal do Espírito Santo, 2019-12-03) Saleme, Estevao Bissoli; Santos, Celso Alberto Saibel; https://orcid.org/0000000232875843; http://lattes.cnpq.br/7614206164174151; https://orcid.org/0000000318563824; http://lattes.cnpq.br/8757661847246456; Pereira Filho, Jose Goncalves ; https://orcid.org/0000-0002-8056-3836; http://lattes.cnpq.br/8265854560095874; Martinello, Magnos; https://orcid.org/0000-0002-8111-1719; http://lattes.cnpq.br/7471111924336519; Saade, Debora Christina Muchaluat; https://orcid.org/0000-0002-1233-9736; http://lattes.cnpq.br/2448703093928632
The increasing interest in digital immersive experiences has drawn the attention of researchers into understanding human perception whilst adding sensory effects to multimedia systems such as VR (Virtual Reality) and AR (Augmented Reality) applications, multimedia players, and games. These so-called mulsemedia—multiple sensorial media—systems are capable of delivering wind, smell, vibration, among others, along with audiovisual content with the aim of enhancing users’ Quality of Experience (QoE) in areas such as entertainment, healthcare, education, culture, and marketing. To support the researchers’ investigation, there have been developed many standalone software solutions and incipient architectural proposals to bind these applications to sensory effects devices, such as wind fans, scent emitters, vibration chairs, etc. These devices, in turn, are constantly evolving, making it difficult to update applications to be compatible with them. There is little or no interoperability between software and hardware in this realm, hindering reuse in other contexts. Every time a mulsemedia application is needed, new software is built mostly from scratch. This model has proven to be demanding, time-consuming, and costly mainly because it requires researchers and developers alike to gain knowledge about new devices, connectivity, communication protocols, and other particulars. The fact is that building such systems imposes a number of challenges and requirements (which are discussed in this thesis) due mainly to their ever-evolving and heterogeneous traits. As a result, few mulsemedia systems have remained reusable to be applied to different research purposes as opposed to the use of open mulsemedia datasets. Therefore, the main contribution of this thesis is a decoupled conceptual architecture to deal with variability of scenarios in mulsemedia delivery systems, which includes recommendations to cope with the variation of end-user applications and sensory effect devices through the support and reuse of even unforeseen communication and connectivity protocols, and sensory effects metadata (SEM). To evaluate it, an open-source and robust mulsemedia framework was developed. Then, a performance assessment was carried out on communication protocols for the integration between event-based applications, whereby temporal restrictions play a role, and the framework. Results indicated statistically significant differences in response time providing directions for optimized integrations. Finally, a user QoE subjective evaluation comparing a monolithic mulsemedia system with this framework was undertaken with results suggesting no evinced statistically significant differences in user perceived QoE between the systems under different aspects. Therefore, it is hoped that this work fosters the area of mulsemedia and HCI (Human-Computer Interaction) in the sense that researchers can leverage either the conceptual architecture to design mulsemedia delivery systems or the framework to carry out their experiments
A formal analysis of Identity and Sortality in the Unified Foundational Ontology (UFO)
(Universidade Federal do Espírito Santo, 2021-04-20) Nicola, João Rafael Moraes; Guizzardi, Giancarlo; https://orcid.org/0000-0002-3452-553X; http://lattes.cnpq.br/5297252436860003; https://orcid.org/0000-0001-8731-291X; http://lattes.cnpq.br/7572689551845981; Almeida, João Paulo Andrade; https://orcid.org/0000-0002-9819-3781; http://lattes.cnpq.br/4332944687727598
The Unified Foundational Ontology (UFO) is a conceptual framework grounded on principles derived from Ontology discipline of Philosophy, with applications in the field of Software Engineering, specially in Conceptual Modeling, as the semantic reference for the OntoUML modeling language. Among the concepts described in the UFO fragment of endurants (UFO-A), the concept of sortality plays a central role in the classification of UFO substantial universals. However, this concept, and the related concepts of identity and individuality currently lack a systematic formal characterization, hindering their application in the analysis of substantial universals. This research enriches the Unified Foundational Ontology (UFO) literature with a formal specification for a fragment of UFO-A that allows the characterization of these concepts. This specification is presented in Isabelle/HOL, a logical formalism that allows a machine-assisted verification. We construct a formal framework, based on this specification and on categoric-theoretic concepts through which we propose formal definitions for the concepts of individuality and identity, and, from these definitions, we propose a formal characterization of the concept of sortality. Illustrations and the application of the proposed definitions on the domain of conceptual modeling are also presented.
A Place-as-you-go In-network Framework by Modular Decomposition for Flexible Embedding to Software/Hardware Co-design
(Universidade Federal do Espírito Santo, 2022-10-21) Mafioletti, Diego Rossi; Martinello, Magnos; https://orcid.org/0000-0002-8111-1719; http://lattes.cnpq.br/7471111924336519; https://orcid.org/0000-0002-1513-9414; http://lattes.cnpq.br/2470233635439757; Mota, Vinicius Fernandes Soares; https://orcid.org/0000-0002-8341-8183; http://lattes.cnpq.br/9305955394665920; Both, Cristiano Bonato; https://orcid.org/0000-0002-9776-4888; http://lattes.cnpq.br/2658002010026792; Verdi, Fabio Luciano; https://orcid.org/0000-0002-5455-8910; http://lattes.cnpq.br/9143186843657940; Giaccone, Paolo
Cloud computing has become very popular as a computation platform, being part of people’s daily life. In order to deliver the cloud-required super-computing power, the network must play a crucial role by connecting hundreds of thousands of machines within data centres. However, with the rise of a new massive set of cloud-native and intensive applications (e.g. 5G, cloud robotics, deep learning, etc) combined with network function virtualisation (NFV), a significant strain has been placed on the processing capacity of server CPUs, demanding even higher performance and turning the network into a bottleneck. Typically, the network interface card (NIC) has been used to connect servers to the network. Although, smart network interface cards (SmartNICs) are becoming an increasingly popular method of offloading intensive packet processing tasks from servers, thus freeing CPU cycles to drive application performance. The critical challenge towards using a SmartNIC is how to make efficient use of these in-network heterogeneous computing resources, as there is a significant gap between application software and the computing capabilities of programmable devices. First, this device lacks generic programming models or abstractions, being usually programmed using low-level primitives or proprietary APIs. Second, the network developer needs to deal with the internal complexity of hardware resources, as well as manage the balance on offloading workloads, trying to find out the tradeoff between additional overheads and offloading benefits. It is needed to figure out how to co-design application logic between programmable network hardware and end-host servers within the edge computing paradigm. This thesis presents a novel framework for prototyping and deploying in-network applications. The framework is structured into a set of components that rely on i) application functional decomposition; ii) identification of logic blocks; iii) aggregation of overlapped functions merged and mapped on network functionalities in P4 language; iv) intercepting, interacting and forwarding with data structures for balancing the offloading of network flows. In order to demonstrate the principle of co-designing on diverse applications, Virtual Network Functions (VNFs) are created, and some of their functionally decomposed elements are deployed as small embedded Network Functions (eNFs) on in-network processors in sorted use cases, reviewing the framework components raised previously: decomposing network functions (i, ii, iii) and unifying components to fit into a set of VNFs/eNFs, examining latency, throughput and vCPU usage against their software implementation counterpart; intercepting and interacting with cloud robotics (iv), raising security concerns when using programmable data planes; evolving to security network functions running at in-network computing (i, ii, iii, iv), checking the overhead added by them; and modifying a PON upstream scheduler mechanism (iv), providing low latency requirements for applications.
A planning pipeline for software systems of autonomous vehicles
(Universidade Federal do Espírito Santo, 2023-09-04) Azevedo, Pedro Henrique Vieira de Oliveira; Gonçalves, Claudine Santos Baduê; https://orcid.org/0000-0003-1810-8581; http://lattes.cnpq.br/1359531672303446; http://lattes.cnpq.br/5958251435708396; Komati, Karin Satie; https://orcid.org/0000-0001-5677-4724; http://lattes.cnpq.br/9860697624155451; Santos, Thiago Oliveira dos; https://orcid.org/0000-0001-7607-635X; http://lattes.cnpq.br/5117339495064254; Boeres, Maria Claudia Silva; https://orcid.org/0000-0001-9801-2410; http://lattes.cnpq.br/0528154281423964; Wolf, Denis Fernando
The use of autonomous vehicles on public roads and in industrial environments has been growing in recent years and to achieve the higher level of autonomy, where no human intervention is needed, the vehicle must know the roads of the environment where it will operate, which allows it to travel from an initial location to a desired destination dealing with the uncertainties of the environments, such as static obstacles. In this work, we propose a planning pipeline for autonomous vehicle software systems. The pipeline is composed of an offline process and an online process. The offline process consists of two modules: (i) the Waypoints Editor, used to edit or even create entire new paths using an open-source vector graphics editor. (ii) The Multi-Level Road Network Generator constructs a representation of the environment in two levels of representation where the first one is used to guide the vehicle in the autonomous mission and the second one represents the paths and is used is route computation. The online process consists of three modules: (i) the Route Planner module to compute routes, (ii) the Off-Road Planner module to compute short paths in order to bring the car to the start of the route or to take the car from the end of the route to the final goal, and (iii) the Frenét Frames Path Planner to generate alternative paths to the right and left of the route and, in the presence of static obstacles, overtake them. We evaluated the performance of the proposed planning pipeline through simulations in three different real-world datasets using our Autonomous Vehicle Simulator module. Simulated experimental results showed that the proposed planning pipeline allowed the autonomous vehicle to know the environment as much as the user of the system wants and successfully execute missions from an initial position to a desired destination computing faster routes using the second level of the Multi-Level Road Network, dealing with static obstacles in the environment obeying the overtake safe distance restrictions imposed by the autonomous vehicle system.
ALGORITMOS HEURÍSTICOS INTELIGENTES PARA PROBLEMAS DE LAYOUT EM LINHA ÚNICA E LINHA DUPLA
(Universidade Federal do Espírito Santo, 2021-08-03) Cravo, Gildásio Lecchi; Amaral, André Renato Sales; https://orcid.org/0000-0001-7344-3994; http://lattes.cnpq.br/4695002674556067; https://orcid.org/0000-0003-2497-2835; http://lattes.cnpq.br/4301368869549226; Mauri, Geraldo Regis; https://orcid.org/0000-0002-8393-7741; http://lattes.cnpq.br/7870111209439581; Mestria, Mário; https://orcid.org/0000-0001-8283-0806; http://lattes.cnpq.br/5866381928751063; Boeres, Maria Claudia Silva; https://orcid.org/0000-0001-9801-2410; http://lattes.cnpq.br/0528154281423964; Lorenzoni, Luciano Lessa; https://orcid.org/0000-0003-4859-7750; http://lattes.cnpq.br/7959495705859101
The study of layout of facilities aims to determine the best use of available space, resulting in more effective manufacturing processes in the context of industry. This thesis addresses four facility layout problems, categorized as row layout, in which facilities must be arranged in one or two straight lines, respecting some allocation constraints. Initially, the problem of single-row facility layout (SRFLP) is addressed, which consists of arranging facilities along a straight line, in order to minimize the weighted sum of the distances between all pairs of facilities. The other three problems addressed in this study are SRFLP extensions, in which the facilities are arranged in two lines, namely: the double-row layout problem (DRLP), the parallel row ordering problem (PROP) and the bi-objective corridor allocation problem (bCAP). An algorithm called GRASP-F is proposed for SRFLP. The computational experiments show the efficiency of the method by improving the known-values for 29 out of 93 instances in the literature with up to 1000 facilities. To date, this is the second work to consider problems of this magnitude. For DRLP, a purely heuristic approach, called PSO-DRLP, is proposed based on the Particle Swarm Optimization meta-heuristic. The PSO-DRLP presented values equal to or better than the known-values for 35 of 51 instances in the literature, and, for the remaining 16, the values found are very close to the best-known values. The solution algorithm for PROP, called AILS, is based on the ILS meta-heuristic, but unlike the standard, two phases with different intensification and diversification characteristics were used, in addition to using techniques to accelerate the calculation of the gain in the objective function in the neighborhood move used in the local search. The results found improved 49 out of 100 instances with previous known results and for the remaining 51 instances the best-known results were achieved. In addition to these tests, experiments were carried out using 6 instances with up to 300 facilities, unprecedented in the context of PROP. For bCAP, an algorithm similar to AILS-PROP was proposed, also in two phases and with techniques to accelerate the calculation of the gain in the objective function in the neighborhood movements used in the local search, having obtained excellent results for 76 tested instances. In general, the proposed solutions for the four problems can be considered excellent alternatives to solve layout problems in single and double lines, being possible to obtain high quality results for large problems in low computational times.
An alternative approach of parallel preconditioning for 2D finite element problems
(Universidade Federal do Espírito Santo, 2018-06-29) Lima, Leonardo Muniz de; Catabriga, Lucia; Almeida, Regina Célia Cerqueira de; Santos, Isaac Pinheiro dos; Souza, Alberto Ferreira de; Elias, Renato Nascimento
We propose an alternative approach of parallel preconditioning for 2D finite element problems. This technique consists in a proper domain decomposition with reordering that produces narrowband linear systems from finite element discretization, allowing to apply, without significant efforts, traditional preconditioners as Incomplete LU Factorization (ILU) or even sophisticated parallel preconditioners as SPIKE. Another feature of that approach is the facility to recalculate finite element matrices whether for nonlinear corrections or for time integration schemes. That means parallel finite element application is performed indeed in parallel, not just to solve the linear system. We also employ preconditioners based on element-by-element storage with minimal adjustments. Robustness and scalability of these parallel preconditioning strategies are demonstrated for a set of benchmark experiments. We consider a group of two-dimensional fluid flow problems modeled by transport, and Euler equations to evaluate ILU, SPIKE, and some element-by-element preconditioners. Moreover, our approach provides load balancing and improvement to MPI communications. We study the load balancing and MPI communications through analyzer tools as TAU (Tuning Analysis Utilities).
An Ontology Network to support Knowledge Representation and Semantic Interoperability in the HCI Domain
(Universidade Federal do Espírito Santo, 2022-07-08) Costa, Simone Dornelas; Barcellos, Monalessa Perini; https://orcid.org/; http://lattes.cnpq.br/8826584877205264; https://orcid.org/; http://lattes.cnpq.br/; Almeida, Joao Paulo Andrade; https://orcid.org/; http://lattes.cnpq.br/4332944687727598; Zaina, Luciana Aparecida Martinez; https://orcid.org/; http://lattes.cnpq.br/; Pereira, Roberto; https://orcid.org/; http://lattes.cnpq.br/; Souza, Vitor Estevao Silva; https://orcid.org/0000000318695704; http://lattes.cnpq.br/2762374760685577
abstract
An ontology-based process for domain-specific visual language design
(Universidade Federal do Espírito Santo, 2017-08-17) Teixeira, Maria das Graças da Silva; Falbo, Ricardo de Almeida; Gailly, Frederik; Guizzardi, Giancarlo; Almeida, João Paulo Andrade; Campos, Maria Luiza Machado; Poels, Geert; Looy, Amy Van
In het domein van de conceptuele modellering wordt er steeds meer aandacht besteed aan visuele domeinspecifieke modelleertalen en hoe deze talen ondersteuning kunnen bieden bij het representeren van een bepaald domein voor verschillenden belanghebbenden. Bijgevolg is er een absolute noodzaak aan richtlijnen die men kan volgen bij het ontwikkelen van deze domeinspecifieke modelleertalen. Bestaand onderzoek voorziet een aantal richtlijnen maar deze focussen meestal op de abstracte syntax van deze talen en niet op de visuele aspecten (concrete syntax) van deze talen. Er is nochtans een absolute noodzaak aan richtlijnen specifiek voor de ontwikkeleng van de concrete syntax want deze heeft een significante impact op de efficiëntie van de communicatie en probleemoplossende eigenschappen van de met deze talen ontwikkelde conceptuele modellen. De meest gebruikte theorie voor de evaluatie van de concrete syntax van een visuele modelleertaal is de Physics of Notations(PoN). PoN definieert een verzameling principes die men kan gebruiken voor de analyse en ontwerp van een cognitief effectieve visuele notatie voor een modelleertaal. PoN heeft echt ook een aantal tekortkomingen: i) het bevat geen methode die aangeeft hoe de principes moeten gebruikt worden en ii) het helpt niet bij het ontwikkelen van symbolen die overeenstemmen met het domein. In dit PhD project wordt de Physics of Notations Systematized (PoN-S) ontwikkeld en voorgesteld als een oplossing voor de eerste tekortkoming van PoN. PoN-S voorziet een sequentiële set van activiteiten en geeft voor elke activiteit aan welk principe moet worden gebruikt. Bovendien voorziet het ook een groepering voor de verschillende principes die de gebruiker moet helpen bij het gebruik. De tweede tekortkoming wordt in dit PhD project opgelost door gebruik te maken van foundational ontologies. Foundational ontologies worden gebruikt voor het verbeteren van de kwaliteit van zowel de abstracte syntax van een modelleertaal als ook voor het rechtstreeks verbeteren van het conceptueel model. In dit doctoraat wordt het onderzoek van Guizzardi (2013) en meer specifiek het onderzoek rond UFO gebaseerde ontologische richtlijnen gecombineerd met de eerder ontwikkelde verbetering van PoN. Dit resulteert in de Physics of Notations Ontologized and Systematized (PoNTO-S), een systematisch ontwikkelingsproces voor de concrete syntax van visuele modelleertalen waarbij ook rekening wordt gehouden met de ontologische betekenis van de abstracte syntax. Het onderzoek dat uitgevoerd werd in het kader van dit PhD project stemt overeen met een Design Science project met verschillende iteraties die resulteren in verschillende Design Science artefacten die ook werden geëvalueerd. Na de ontwikkeling van PoN-S en PoNTO-S werd er één labo experiment uitgevoerd en werden de artefacten ook deels geëvalueerd door gebruik te maken van twee case studies. Deze studies tonen aan dat PoN-D en PonTO-S nuttig zijn tijdens de ontwikkeling van visuele domeinspecifeke modelleertalen.
An Ontology-based Reference Model for the Software Systems Domain with a focus on Requirements Traceability
(Universidade Federal do Espírito Santo, 2022-04-29) Duarte, Bruno Borlini; Souza, Vitor Estevao Silva; https://orcid.org/0000000318695704; http://lattes.cnpq.br/2762374760685577; https://orcid.org/; http://lattes.cnpq.br/; Leite, Julio Cesar Sampaio do Prado; https://orcid.org/0000-0002-0355-0265; http://lattes.cnpq.br/6871006250321522; Almeida, Joao Paulo Andrade; https://orcid.org/0000-0002-9819-3781; http://lattes.cnpq.br/4332944687727598; Guerra, Eduardo Martins; https://orcid.org/; http://lattes.cnpq.br/3413978291577451; Barcellos, Monalessa Perini; http://lattes.cnpq.br/8826584877205264
Software plays an essential role in modern society, as it has become indispensable in many aspects of our lives, such as social, business and even personal. Because of this importance, many researchers are dedicated to study the nature of software, how it is related to us and how it is able to change aspects in our society. It is accepted by the scientific community that software is a complex social artifact. This notion comes from the fact that a modern software system can be understood as the combination of interacting elements that exist inside a computer, such as programs and data, and in our world, such as sensors, other systems or even people, all of which are specifically organized to provide a set of functionalities or services and so, fulfill its purposes. A major concern in the development of modern complex software-based systems, is ensuring that the design of the system is capable of satisfying the current set of requirements. In this context, it is widely accepted in the scientific literature and in international standards that the requirements have an important role in the software process. Because of that, requirements need to be developed, refined, managed and traced to their origins, in a controlled engineering process, to control their changing nature and mitigate risks. In order to support these activities, we argue, based on the conceptual modeling scientific literature, that we can use ontologies to provide a better understanding of the software systems domain, reducing the inherent complexity and improving the requirements engineering process. In this work, we propose an ontology-based requirements traceability theory centered in different types of software systems requirements. Based on that, we developed the Reference Ontology of Software Systems (ROSS) and the Ontology of Software Defects Errors and Failures (OSDEF). ROSS and OSDEF are domain ontologies about the software systems that are intended to be used together and combined with other existing ontologies, as reference models for requirements traceability. Besides, we developed machine- readable operational ontologies, based on the reference versions of ROSS and OSDEF. The operational ontologies are created to support an ontology-based requirements traceability process that is based on the relationships that exist between the concepts in the ontologies.
Analysis of the impacts of label dependence in multi-label learning
(Universidade Federal do Espírito Santo, 2021-10-19) Mello, Lucas Henrique Sousa; Varejão, Flavio Miguel; https://orcid.org/0000-0002-5444-1974; http://lattes.cnpq.br/6501574961643171; https://orcid.org/0000-0003-3601-8782; http://lattes.cnpq.br/1436724861273417; Haeusler, Edward Hermann; https://orcid.org/0000-0002-4999-7476; http://lattes.cnpq.br/6075905438020841; Boldt, Francisco de Assis; https://orcid.org/0000-0001-6919-5377; http://lattes.cnpq.br/0385991152092556; Santos, Thiago Oliveira dos; https://orcid.org/0000-0001-7607-635X; http://lattes.cnpq.br/5117339495064254; Rodrigues, Alexandre Loureiros; https://orcid.org/0000-0002-7619-2681; Rauber, Thomas Walter; https://orcid.org/0000000263806584; http://lattes.cnpq.br/0462549482032704
Conclusions in the field of multi-label learning are often drawn from experiments using real benchmark datasets, which is a good practice for comparing results. However, it hardly proves or clearly shows how dependencies among class labels impact on the performance and behaviour of multi-label algorithms. A reasonable approach to tackle this issue consists of adopting a mathematical or statistical formulation of the problem and using it to elaborate theoretical proofs. Another approach consists of elaborating experiments in a well-controlled environment where the dependence among labels can be easier controlled and analyzed, which is the case for many works based on artificial datasets. Both approaches are adopted in this thesis to understand the role of label dependence in multi-label learning. The work done in this thesis is composed of several contributions regarding the analysis of multi-label algorithms from a statistical perspective. One contribution is that calibrated label ranking is an algorithm that can perform extremely poor in particular scenarios where label dependence is present, due to the way that pairwise comparison of labels is done by the algorithm. Another contribution is that the label dependence present in multi-label learning makes the optimization of the expected coverage a NP-hard problem, even at restricted conditions. Finally, a proposal is presented on how to build an experimental environment where the label dependence can conveniently be controlled for comparing performance among multi-label learning algorithms.
Aplicando crowdsourcing na sincronização de vídeos gerados por usuários
(Universidade Federal do Espírito Santo, 2017-10-30) Costa Segundo, Ricardo Mendes; Santos, Celso Alberto Saibel; Pereira Filho, José Gonçalves; Guimarães, Rodrigo Laiola; Souza Filho, Guido Lemos de; Willrich, Roberto
Crowdsourcing is a solve-problem strategy based on collecting contributions with partial results from individuals and aggregating them into a major problem solution. Based on this strategy, this thesis shows how the crowd can synchronize a set of videos generated by users, correlated to an event. Each user captures the event with its personal viewpoint and according to its limitations. In this scenario, it is not possible to ensure that all generated contents have homogeneous characteristics (starting time and duration, resolution, quality, etc.), hindering the use of a purely automatic synchronization process. Additionally, user generated videos are distributed available between several independent content servers. The assumption of this thesis is that the ability of human intelligence to adapt can be used to render a group of videos produced in an uncoordinated and distributed manner generating its synchronization. To prove this hypothesis, the following steps were executed: (i) the development of a synchronization method for multiple videos from independent sources; (ii) The execution of a systematic mapping about the use of crowdsourcing for processing videos; (iii) The development of techniques for the use of the crowd in synchronizing videos; (iv) The development of a functional model for developing synchronization applications using crowdsourcing, which can be extended for general video applications; and (v) The execution of experiments to show the ability of the crowd to perform the synchronization. The results found show that the crowd can participate in the synchronization process and that several factors can influence the accuracy of the results obtained.
Appearance-based global localization with a hybrid weightless-weighted neural network approach
(Universidade Federal do Espírito Santo, 2018-02-02) Silva, Avelino Forechi; Santos, Thiago Oliveira dos; Souza, Alberto Ferreira de; Oliveira, Elias Silva de; Gonçalves, Claudine Santos Badue; Aguiar, Edilson de; Ciarelli, Patrick Marques
Currently, self-driving cars rely greatly on the Global Positioning System (GPS) infrastructure, albeit there is an increasing demand for global localization alternative methods in GPS-denied environments. One of them is known as appearance-based global localization, which associates images of places with their corresponding position. This is very appealing regarding the great number of geotagged photos publicly available and the ubiquitous devices fitted with ultra-high-resolution cameras, motion sensors and multicore processors nowadays. The appearance-based global localization can be devised in topological or metric solution regarding whether it is modelled as a classification or regression problem, respectively. The topological common approaches to solve the global localization problem often involve solutions in the spatial dimension and less frequent in the temporal dimension, but not both simultaneously. It was proposed an integrated spatio-temporal solution based on an ensemble of kNN classifiers, where each classifier uses the Dynamic Time Warping (DTW) and the Hamming distance to compare binary features extracted from sequences of images. Each base learner is fed with its own binary set of features extracted from images. The solution was designed to solve the global localization problem in two phases: mapping and localization. During mapping, it is trained with a sequence of images and associated locations that represents episodes experienced by a robot. During localization, it receives subsequences of images of the same environment and compares them to its previous experienced episodes, trying to recollect the most similar “experience” in time and space at once. Then, the system outputs the positions where it “believes” these images were captured. Although the method is fast to train, it scales linearly with the number of training samples in order to compute the Hamming distance and compare it against the test samples. Often, while building a map, one collects high correlated and redundant data around the environment of interest. Some reasons are due to the use of high frequency sensors or to the case of repeating trajectories. This extra data would carry an undesired burden on memory and runtime performance during test if not treated appropriately during the mapping phase. To tackle this problem, it is employed a clustering algorithm to compress the network’s memory after mapping. For large scale environments, it is combined the clustering algorithms with a multi hashing data structure seeking the best compromise between classification accuracy, runtime performance and memory usage. So far, this encompasses solely the topological solution part for the global localization problem, which is not precise enough for autonomous cars operation. Instead of just recognizing places and outputting an associated pose, it is desired that a global localization system regresses a pose given a current image of a place. But, inferring poses for city-scale scenes is unfeasible at least for decimetric precision. The proposed approach to tackle this problem is as follows: first take a live image from the camera and use the localization system aforementioned to return the image-pose pair most similar to a topological database built as before in the mapping phase. And then, given the live and mapped images, a visual localization system outputs the relative pose between those images. To solve the relative camera pose estimation problem, it is trained a Convolutional Neural Network (CNN) to take as input two separated images in time and space in order to output a 6 Degree of Freedom (DoF) pose vector, representing the relative position and orientation between the input images. In conjunction, both systems solve the global localization problem using topological and metric information to approximate the actual robot pose. The proposed hybrid weightless-weighted neural network approach is naturally combined in a way that the output of one system is the input to the other producing competitive results for the Global Localization task. The full approach is compared against a Real Time Kinematic GPS system and a Visual Simultaneous Localization and Mapping (SLAM) system. Experimental results show that the proposed combined approach is able to correctly global localize an autonomous vehicle 90% of the time with a mean error of 1.20m compared to 1.12m of the Visual SLAM system and 0.37m of the GPS, 89% of the time.
Avaliação da aprendizagem em jogos digitais baseada em learning analytics sobre dados multimodais
(Universidade Federal do Espírito Santo, 2018-09-21) Pereira Junior, Heraclito Amancio; Menezes, Crediné Silva de; Souza, Alberto Ferreira de; Castro Junior, Alberto Nogueira de; Queiroz, Sávio Silveira de; Tavares, Orivaldo de Lira; Cury, Davidson
The use of digital games as a pedagogical tool has been successfully applied in the development of the skills, abilities and attitudes required of 21st century professionals, both in primary and secondary education, as well as in vocational training. Despite this, one issue has worried educators who think of using digital games: "How to assess the learning of digital games?". Assessment is an important part of the teaching-learning process. This importance, especially with regard to learning based on computational resources, including digital games, led to the emergence of a research area called Learning Analytics, which "applies techniques and methods of Computer Science, Pedagogy, Sociology, Psychology, Neuroscience and Statistics for the analysis of data collected during educational processes". In order to better understand these assessments, the collection has also considered multimodal data, those from different manifestations of the student, captured by sensors, during the learning process (touches, gestures, voices and facial expressions). Although the publications indicate that some methods, techniques and tools have been researched to support learning assessments in learning computing environments, and these studies have already obtained some results, they have not yet been sufficient to provide clear, comprehensive. In particular, with regard to digital games, there is still limited availability of consolidated resources for assessing student learning during play, which has been one of the major factors hindering a broadening of its use for educational purposes. This work brings a contribution to the solution of this problem through: a computational platform, in the form of a framework, designed based on the techniques and methods of Learning Analytics; a specialization of the ECD (Evidency Center Design) approach, for project evaluations of learning based on digital games, and a Process that organizes the stages and activities of this type of evaluation. Experiments, reported here, using a framework instance, have demonstrated both their own merit as an assessment tool and the specialization of ECD and the said process.
Classifier ensemble feature selection for automatic fault diagnosis
(Universidade Federal do Espírito Santo, 2017-07-14) Boldt, Francisco de Assis; Varejão, Flávio Miguel; Rauber, Thomas Walter; Salles, Evandro Ottoni Teatini; Carvalho, André Carlos Ponce de Leon Ferreira de; Santos, Thiago Oliveira dos; Conci, Aura
An efficient ensemble feature selection scheme applied for fault diagnosis is proposed, based on three hypothesis: a. A fault diagnosis system does not need to be restricted to a single feature extraction model, on the contrary, it should use as many feature models as possible, since the extracted features are potentially discriminative and the feature pooling is subsequently reduced with feature selection; b. The feature selection process can be accelerated, without loss of classification performance, combining feature selection methods, in a way that faster and weaker methods reduce the number of potentially non-discriminative features, sending to slower and stronger methods a filtered smaller feature set; c. The optimal feature set for a multi-class problem might be different for each pair of classes. Therefore, the feature selection should be done using an one versus one scheme, even when multi-class classifiers are used. However, since the number of classifiers grows exponentially to the number of the classes, expensive techniques like Error-Correcting Output Codes (ECOC) might have a prohibitive computational cost for large datasets. Thus, a fast one versus one approach must be used to alleviate such a computational demand. These three hypothesis are corroborated by experiments. The main hypothesis of this work is that using these three approaches together is possible to improve significantly the classification performance of a classifier to identify conditions in industrial processes. Experiments have shown such an improvement for the 1-NN classifier in industrial processes used as case study.
Combining heterogeneous data and deep learning models for skin cancer detection
(Universidade Federal do Espírito Santo, 2020-11-12) Pacheco, Andre Georghton Cardoso; Krohling, Renato Antonio; https://orcid.org/; http://lattes.cnpq.br/5300435085221378; https://orcid.org/; http://lattes.cnpq.br/; Mota, Vinicius Fernandes Soares; https://orcid.org/; http://lattes.cnpq.br/9305955394665920; Cavalieri, Daniel Cruz; https://orcid.org/; http://lattes.cnpq.br/; Papa, Joao Paulo; https://orcid.org/; http://lattes.cnpq.br/; Santos, Celso Alberto Saibel; https://orcid.org/0000000232875843; http://lattes.cnpq.br/7614206164174151
Currently, Deep Neural Networks (DNN) are the most successful and common methodologies to tackle medical image analysis. Despite the success, applying Deep Learning for these types of problems involves several challenges such as the lack of large training
Copycat CNN: convolutional neural network extraction attack with unlabeled natural images
(Universidade Federal do Espírito Santo, 2023-04-25) Silva, Jacson Rodrigues Correia da; Santos, Thiago Oliveira dos; https://orcid.org/0000-0001-7607-635X; http://lattes.cnpq.br/5117339495064254; https://orcid.org/0000-0002-4314-1693; http://lattes.cnpq.br/0637308986252382; Goncalves, Claudine Santos Badue; https://orcid.org/0000-0003-1810-8581; http://lattes.cnpq.br/1359531672303446; Luz, Eduardo Jose da Silva; https://orcid.org/0000-0001-5249-1559; http://lattes.cnpq.br/5385878413487984; Almeida Junior, Jurandy Gomes de; https://orcid.org/0000-0002-4998-6996; http://lattes.cnpq.br/4495269939725770; Rauber, Thomas Walter; https://orcid.org/0000000263806584; http://lattes.cnpq.br/0462549482032704
Convolutional Neural Networks (CNNs) have been achieving state-of-the-art performance on a variety of problems in recent years, leading to many companies developing neuralbased products that require expensive data acquisition, annotation, and model generation. To protect their models from being copied or attacked, companies often deliver them as black-boxes only accessible through APIs, that must be secure, robust, and reliable across different problem domains. However, recent studies have shown that state-of-the-art CNNs have vulnerabilities, where simple perturbations in input images can change the model’s response, and even images unrecognizable to humans can achieve a higher level of confidence in the model’s output. These methods need to access the model parameters, but there are studies showing how to generate a copy (imitation) of a model using its probabilities (soft-labels) and problem domain data. By using the surrogate model, an adversary can perform attacks on the target model with a higher possibility of success. We further explored these vulnerabilities. Our hypothesis is that by using publicly available images (accessible to everyone) and responses that any model should provide (even blackboxes), it is possible to copy a model achieving high performance. Therefore, we proposed a method called Copycat to explore CNN classification models. Our main goal is to copy the model in two stages: first, by querying it with random natural images, such as those from ImageNet, and annotating its maximum probabilities (hard-labels). Then, using these labeled images to train a Copycat model that should achieve similar performance to the target model. We evaluated this hypothesis on seven real-world problems and against a cloud-based API. All Copycat models achieved performance (F1-Score) above 96.4% when compared to target models. After achieving these results, we performed several experiments to consolidate and evaluate our method. Furthermore, concerned about such vulnerability, we also analyzed various existing defenses against the Copycat method. Among the experiments, defenses that detect attack queries do not work against our method, but defenses that use watermarking can identify the target model’s Intellectual Property. Thus, the method proved to be effective in model extraction, having immunity to the literature defenses, but being identified only by watermark defenses.
CRF+LG: uma abordagem híbrida para o reconhecimento de entidades nomeadas em português
(Universidade Federal do Espírito Santo, 2019-02-07) Pirovani, Juliana Pinheiro Campos; Oliveira, Elias Silva de; Laporte, Éric; Lima, Priscila Machado Vieira; Ciarelli, Patrick Marques; Gonçalves, Claudine Santos Badue
Named Entity Recognition involves automatically identifying and classifying entities such as persons, places, and organizations, and it is a very important task in Information Extraction. Named Entity Recognition systems can be developed using the following approaches: linguistics, machine learning or hybrid. This work proposes the use of a hybrid approach, called CRF+LG, for Named Entity Recognition in Portuguese texts in order to explore the advantages of both linguistics and machine learning approaches. The proposed approach uses Conditional Random Fields (CRF) considering the term classification obtained by a Local Grammar (LG) as an additional informed feature. Conditional Random Fields is a probabilistic method for structured prediction. Local grammars are handmade rules to identify expressions within the text. The aim was to study this way of including the human expertise (Local Grammar) in the machine learning Conditional Random Fields approach and to analyze how it can contribute to the performance of this approach. To achieve this aim, a Local Grammar was built to recognize the 10 named entities categories of HAREM, a joint assessment for the Named Entity Recognition in Portuguese. Initially, the Golden Collection of the First and Second HAREM, considered as a reference for Named Entity Recognition systems in Portuguese, were used as training and test sets, respectively, for evaluation of the CRF+LG. After that, the proposed approach was evaluated in two other datasets. The results obtained outperform the results of systems reported in the literature that were evaluated under equivalent conditions. This gain was approximately 8 percentage points in F-measure in comparison to a system that also used CRF and 2 points in comparison to a system that used Neural Networks. Some systems that used Neural Networks presented superior results, but using massive corpora for unsupervised learning of features, which was not the case of this work. The Local Grammar built can be used individually when there is no training set available and in conjunction with other machine learning techniques to improve its performance. We also analyzed the boundaries (lower bound and upper bound) of the proposed approach. The lower bound indicates the minimum performance and the upper bound indicates the maximum gain that we can achieve for the task in question when using this approach.
CROSS-LAYER NETWORK PROGRAMMABILITY FOR EXPRESSIVE AND AGILE ORCHESTRATION ACROSS HETEROGENEOUS RESOURCES
(Universidade Federal do Espírito Santo, 2021-05-28) Guimaraes, Rafael Silva; Martinello, Magnos; https://orcid.org/; http://lattes.cnpq.br/7471111924336519; https://orcid.org/; http://lattes.cnpq.br/; Mota, Vinicius Fernandes Soares; https://orcid.org/; http://lattes.cnpq.br/9305955394665920; Ribeiro, Moises Renato Nunes; https://orcid.org/0000000291492391; http://lattes.cnpq.br/1005553714687743; Filho, Jugurta Rosa Montalvao; https://orcid.org/; http://lattes.cnpq.br/
Orchestration can be viewed as an inter-working technology-agnostic glue thatdecouples, understands, supports, and provides end-to-end communication based ona unified optical-wireless-packet-cloud view. Software-defined network (SDN) andnetwork function v
Crossing domains for accuracy: in-network stacking of machine learning classifiers
(Universidade Federal do Espírito Santo, 2024-06-21) Xavier, Bruno Missi; Ruffini, Marco; https://orcid.org/0000-0001-6220-0065; Martinello, Magnos; https://orcid.org/0000-0002-8111-1719; Pacheco, André; Kirian, Mariam; Pasquini, Rafael; Aparicio, Albert Cabellos
Traffic management plays a crucial role for this expansive global connectivity. In this context, traffic classification strategically differentiates a range of applications and its requirements. This transformation enhances network agility and facilitates the direct integration of Machine Learning (ML) into the network infrastructure, fundamentally changing traffic management by promoting proactive data processing and analysis within the network. The synergy of Network Softwarization (NS) with ML not only leads to reduced latency and improved load management, but also increases the capacity to effectively handle larger volumes of data. Consequently, networks are better equipped to meet the complex demands of modern digital ecosystems, ensuring robust and efficient connectivity. This thesis delved into the domain of programmable networks, with a specific focus on implementing ML for traffic classification within the network architecture components, including the Radio Access Network (RAN) and programmable data planes. This approach represents a significant departure from the traditional traffic classification techniques, which are typically deployed on end-hosts. It also paves the way for integrating Cross Domain Artificial Intelligence (AI) capabilities within the network, facilitated by multi-view learning. More specifically, we advanced the state-of-the-art with four main contributions: (i) We design a framework named Early Attack Guarding and Learning Engine (EAGLE) as the first defense line against a set of cyber threats. Our framework explores Open Radio Access Network (O-RAN) to collect measurements from the air interface (that is, Physical and Medium Access Control) to identify and early mitigate malicious flows; (ii) We introduce MAP4 that demonstrates the feasibility of deploying ML models (that is, decision trees and Random Forest) in the data plane. To achieve this, we rely on the P4 language to deploy a pre-trained model to accurately classify flows at line rate; (iii) We proposed an In-Network Concept Drift to deal with the dynamic nature of the network traffic. This approach detects changes in traffic distribution by implementing Exponentially Weighted Moving Average (EWMA) overcoming the P4 limitations; (iv) Our Cross-Domain AI integrates multiple layers (RAN and programmable data planes) to form an in-network stacking of ML classifiers under the multi-view learning perspective. This innovative approach overcomes the challenges of a single layer in order to improve the overall accuracy of the classification system.

Navegar

Navegando Doutorado em Ciência da Computação por Título

Resultados por página

Opções de Ordenação