TECHNOLOGY OF COMPUTERS

Author: LOPEZ PELAYO FERNANDO.
Year: 2003.
University: CASTILLA-LA MANCHA [More theses of this university] [www.uclm.es].
Place of defense: ESCUELA POLITÉCNICA SUPERIOR DE ALBACETE.
Place of preparation: ESCUELA POLITÉCNICA SUPERIOR DE ALBACETE.

Summary: This thesis presents a theoretical model to validate qualitative and quantitative characteristics in concurrent systems, and real-time multi-Algebra processes Markovianos ROSA. It presents an operational semantics and other denotacional to capture the functional properties of a system, in particular non-deterministic systems can be analyzed, with probabilities (generativas) and distributions Exponenciales and Poisson. Relations are provided on these semantic equivalence permitting processes compare functionally. It also presents an algorithm that allows temporary assessment compare the efficiency of processes. It applies to the analysis of ABP ROSA (a classic case study) and MPEG-2 (an application realcuyas characteristics are crucial temporary) getting interesting results. This demonstrates the usefulness of formal models in the performance assessment in real systems. Finally, unable to describe accurately the actual turnout by ORSA leads to the use of other less intuitive formalism but with better skills for the analysis of actual attendance: Petri Nets with Arcos Temporizados. Through its implementation very interesting results in improving the implementation of the algorithm ciompresión of vido MPEG2, such as an analysis of the number of processors required when permits the maximum turnout, as well as the improvement obtained in the theoretical benefits of this algorithm.

Author: PARCERISA BUNDO JOAN-MANUEL.
Year: 2003.
University: POLITÉCNICA DE CATALUÑA [More theses of this university] [www.upc.edu].
Place of defense: ARQUITECTURA DE COMPUTADORS.
Place of preparation: EDIFICI D6 DESPATX 206 Campus NORD.

Summary: The objective of this thesis is to propose new techniques to design efficient clustered superscalar microarchitectures. Clustered microarchitectures partition the layout of several critical hardware components as a means to keep most of the parallelism while improving the scalability. A clustered processor core, made up of several low complex blocks or clusters, can efficiently execute chains of dependent instructions without paying the overheads of a long issue, register read or bypass latencies. Of course, when two dependent instructions execute in different clusters, an inter-cluster communication penalty is incurred. Moreover, distributed structures usually imply lower dynamic power requirements, and simplify power management via techniques such a selective clock/power gating and voltage scaling. The first target of this research is the assignment of instructions to clusters, since it plays a major role on performance, with the goals of keeping the workload of clusters balanced and reducing the penalty of critical communications. Two different approaches are proposed: first, a family of new schemes that dynamically identify groups of data-dependent instructions called slices, and make cluster assignments on a per-slice basis. The proposed schemes differ from previous approaches either because they are dynamic and/or because they include new mechanisms to deal explicitly with workload balance information gathered at runtime. Second, it proposes a family of new dynamic schemes that assign instructions to clusters in a per-instruction basis, based on prior assignment of the source register producers, on the cluster location of the source physical registers, and on the workload of clusters. The second contribution proposes value prediction as a means to mitigate the penalties of wire delays and, in particular, to hide inter-cluster communications while also improving workload balance. First, it is proven that the benefit of breaking dependences with value prediction grows with the number of clusters and the communication latency, thus it is higher than for a centralized architecture. Second, it is proposed a cluster assignment scheme that exploits the less dense data dependence graph that results from predicting values to achieve a better workload balance. The third aspect considered is the cluster interconnect, which mainly determines communication latency, seeking for the best trade-off between cost and performance. First, several cost-effective point-to-point interconnects are proposed, both synchronous and partially asynchronous, that approach the IPC of an ideal model with unlimited bandwidth while keeping the complexity low. The proposed interconnects have much lower impact than other approaches on the complexity of bypasses, issue queues and register files. Second, possible router implementations are proposed, which illustrate their feasibility with very simple and low-latency hardware solutions. Third, a new topology-aware improvement to the cluster assignment scheme is proposed to reduce the distance (and latency) of inter-cluster communications. The last contribution proposes techniques for distributing the main components of the processor front-end with the goals of reducing their complexity and avoiding replication. In particular, effective techniques are proposed to cluster the branch predictor and the steering logic, that minimize the wire delay penalties caused by broadcasting recursive dependences in two critical hardware loops: the fetch address generation, and the cluster assignment logic, respectively. In the former case, the proposed technique converts the cross-structure wire delays of a centralized predictor into cross-cluster communication delays, which are smoothly pipelined. In the latter case, the partitioning of the instruction steering logic involves the parallelization of an inherently sequential task such as the dependence based cluster assignment of instructions.

Author: CANAL CORRETGER RAMON.
Year: 2003.
University: POLITÉCNICA DE CATALUÑA [More theses of this university] [www.upc.edu].
Place of defense: D6-S103.
Place of preparation: EDIFICI D6 DESPATX 206 Campus NORD.

Summary: Energy consumption and power dissipation have become a key constraint in the design of processors. In the embedded segment, battery life is an issue, so the processor energy consumption has to be minimal. In the high-performance segment, the power dissipation is a limiting factor since the cooling mechanisms are becoming more expensive or even reaching its limits. Although significant research is targeted to produce longer-life batteries and better cooling systems, contributions from other areas are becoming critical. This thesis is proposes several techniques that focus either on a special part of the processor or on it as a whole. The main contributions of this thesis are explained below: Contributions to the Issue Logic Design Novel issue logic designs have been proposed in this thesis. These proposals can be divided into two families: Dependence-tracking schemes. These schemes keep track of the producer-consumer relationships between instructions. Since the direct dependences are stored, the scheme eliminates most of the need for the associative wake-up logic. Prescheduling schemes. Since most of the functional units have a fixed execution latency (all but the memory); at dispatch time, the mechanism schedules the instructions for execution according to the estimated availability cycle of its source operands and functional unit. As a consequence, these schemes eliminate most of the logic needed by the issue logic since the instructions will be only considered for issue just once (at the time they have been scheduled). Contributions to the Ultra-Low Power Processors Value compression is shown to be an effective way to reduce energy requirements. Both for ultralow power processors and high-performance ones, the values that flow through the pipeline determine the energy consumption of the structures in the datapath (mainly the register file, the data cache and the functional units). Since the energy consumption of these components depends directly on the values manipulated, we analyzed and proposed several value compression schemes that reduce the energy consumption of the processor by reducing the number of bits needed to be manipulated for each data value. Several datapaths are redesigned by considering value compression. From this setup, different microarchitectures that exhibit different energyperformance tradeoffs are proposed and analyzed. A new microarchitecture targeted to embedded processors is proposed. In this architecture the datapath is redesigned in order to compute just the bytes that are significant. Several alternatives for a different set of energy/performance trade-offs are proposed. While the alternatives are still targeted to embedded processors, these show a compromise between the energy budget and the performance reduction. Contributions to the high-performance processors Value compression is shown to reduce significantly the energy consumed by conventional highperformance processors. We show how hardware only schemes are able to reduce significantly the energy budget and, at the same time, we show how compile-time techniques can be an important support for reducing the energy requirements apart or on top of hardware-only value compression.

Author: ATIENZA ALONSO DAVID.
Year: 2004.
University: COMPLUTENSE DE MADRID [More theses of this university] [www.ucm.es].
Place of defense: FACULTAD DE INFORMÁTICA.
Place of preparation: FACULTAD DE INFORMÁTICA.

Summary: The objective of the research proposed here is the creation of a comprehensive methodology for design and implemetnación which serves to efficiently manage dynamic memory in the new systems for high performance embedded multimedia. To that end, this methodology provides two fundamental aspects. On the one hand, from the point of view of hardware is considered all factors leading design of a system embedded, ie performance requirements on time (eg, in 3D rendering is required to reach 25-30 frames / second non - a higher number and reduced memory consumption and energy. Moreover, from the point of view of software, methodology defines a flow of structured design (ie, with detailed steps of analysis, design and implementation) at two levels major abstraction existing environments dedicated embedded with a capacity of dynamic memory management: dynamic data structures and support at the operating system of allocating and freeing real-time memory blocks.

Author: MOLINA PREGO M. CARMEN.
Year: 2004.
University: COMPLUTENSE DE MADRID [More theses of this university] [www.ucm.es].
Place of defense: FACULTAD DE INFORMÁTICA.
Place of preparation: FACULTAD DE INFORMÁTICA.

Summary: High-Level Synthesis (SAN) part of a behavioral description of a circuit to produce an implementation of the same level of transfer between registers. The algorithms that represent the state of art in the SAN try to minimize the area of the circuits synthesized refusal to maximizing the resources present in hardware implementations for what they are trying to balance the number of operations carried out by different cycle. However, the total balance is virtually impossible to achieve in most designs, which appear in the hardware resources that some of the cycles are not used to carry, store or transmit data or operations of the specification. This implies a certain loss of the hardware components of the route data is even more dramatic when synthesize specifications heterogeneous, composed of transactions with different data representations and widths. In this case, besides the loss of hardware due to the non-use of hardware in specific cycles, joins wasted due to the partial delivery, storage or transmission of a transaction or data using an appeal for greater width. The objective of the research collection in memory of this thesis has been to identify the causes of the loss of hardware and proposed a series of design methodologies applied to the planning of operations and the selection and assignment of hardware contribute to the reduction area synthesized circuits. These methodologies are based on the fragmentation selective operations in the specification sets new operations simpler that can be treated independently, which leads to implementations in which some of the operations of the original specification are implemented in several cycles, not necessarily consecutive, and several functional units. Similar techniques are applied to data storage and transmission in order to reduce the area for the storage resources and routing. This report also presents two algorithms operations planning and selection and allocation of hardware resources efficiently implementing some of the technical proposals, demonstrating its viability and effectiveness.

Author: LOPEZ GARCIA FERNANDO.
Year: 2004.
University: POLITÉCNICA DE VALENCIA [More theses of this university] [www.upv.es].
Place of defense: Dep. Informatica de Sistemas y Computadores.
Place of preparation: Universidad Politécnica de Valencia.

Summary: This thesis presents a case of study of the development and performance analysis of a surface grading application with real-time compliance. We address the issue of spatial and temporal uniformity in the acquisition system. In a surface grading application it is crucial to ensure the uniform response of the system through time and space. All the results presented for surface grading were obtained using real data from the ceramic tile industry. The VxC TSG database is public and can be accessed at www.disca.upv.es/vision/vxctsg. We present a method based on soft colour-texture descriptors computed in perceptually uniform colour spaces. The method is parameterized and the involved factors are studied using two statistical procedures; experimental design and logistic regression. Although it is not a new theoretical contribution, we have found and demonstrate that a simple set of global colour and texture statistics, together with well-known classifiers, are powerful enough to meet stringent factory requirements for real-time and performance. Also the method is compared with two other approaches from the scientific literature; Colour Histograms and Centile-LBP. Finally, we explore the method's capacity for on-line inspection in a study of real-time compliance and parallelization.

Author: SÁNCHEZ ÉLEZ MARTÍN MARCOS.
Year: 2004.
University: COMPLUTENSE DE MADRID [More theses of this university] [www.ucm.es].
Place of defense: CC. FÍSICAS.
Place of preparation: CC. FISICAS..

Summary: The research presented falls within the planning of multimedia applications in reconfigurable systems. These systems are specially designed to process these applications because they enable the parallel processing of a large amount of data in the shortest possible time. The processing of this large amount of data involves the transfer from the external memory to the internal memories of the reconfigurable platform, resulting in consumption of a large amount of energy. This can be reduced significantly if there is a planning transfers of such data and proper management of its storage in various internal reports. The work has been designed from the beginning to be applied to the system MorphoSys introducing the common characteristics of most of coarse grain reconfigurable systems. This thesis part of a compilation environment previously developed for such systems and makes the study of one of its main phases, planning data, and their impact on pre - and posterioresdel environment. In the development of this research work has studied the influence of different alternatives storage hardware data on the performance of the final system. There has been an attempt not only that the planner Data will be able to adapt to the memory hierarchy imposed by the platform, but also that the study results planner guide tuning of the elements of data storage platform. To allow the use of the platforms reconfigúdbles for some of the most demanding applications .. requires that the environment may try compiling applications that could describe dynamic or interactive. The planner Data presented in this thesis, although it was initially developed for static applications, after minor modifications can adapt to a dynamic planning of the tasks of the application.

Author: HUEDO CUESTA EDUARDO.
Year: 2004.
University: COMPLUTENSE DE MADRID [More theses of this university] [www.ucm.es].
Place of defense: FACULTAD DE INFORMÁTICA.
Place of preparation: FACULTAD DE INFORMÁTICA.

Summary: The need to tap the resources available in the computer systems connected to the Internet and simplify its use has led to a new form of information technology known as Grid Computing. This new technology is análogoa to networks of power supplies: the idea is to provide a single point of access to a set of geographically distributed resources. (Superordenadores, clusters, storage, information, tools, personnel, etc.). Thus, distributed systems can be used as a single virtual system in data-intensive applications, with high computational demand or require collaboration among multiple resources. Despite our efforts over the past few years, the development and implementation of applications in the Grid continues to require a high level of expertise because of their complexity. The developer or user is responsible for scheduling or perform manually every stage of planning work to get some functionality. Furthermore, one should look at all the circumstances that may occur in an environment highly diverse and dynamic as the Grid. The main objective of this thesis is to study different techniques for planning and implementing work on Grids that provide greater efficiency, reliability and ease of use. Of all existing technologies, perhaps the most interesting is the implementation of adaptive work. The key aspect of the implementation is adaptive recognition of the changing conditions, both of the resources of Grid and the demands of the application. To achieve this functionality, we propose a model of application aware Grid and a shipping agent, which provides delivery mechanisms necessary to adapt the application execution. The techniques studied are included in a new experimental tool based on Globus, which allows delivery of simpler and more efficient work in dynamic environments Grid. Adapting to changing conditions is achieved by implementing a replanificación automatic applications when they occur degradation in performance, discoveries of "best" resources, changes in the demands of the application, cancellation or suspension of work, failures, etc.. The replanificación a job can lead to their migration to other resources more appropriately. The viability of the techniques developed is demonstrated with the implementation of several types of applications: high-productivity applications in Computational Fluid Dynamics (analysis of the dynamics of a fluid on a thin plate) and Bioinformatics (predición the structure and thermodynamic properties of protein sequences) and the NAS Benchmarks Grid, which include applications with complex workflows. The results obtained demonstrate the decisive impact that it has the ability to adjust the performance of work on performance and robustness that can be achieved.

Author: SAEZ ACHAERANDIO YAGO.
Year: 2004.
University: POLITÉCNICA DE MADRID [More theses of this university] [www.upm.es].
Place of defense: FACULTAD DE INFORMATICA.
Place of preparation: FACULTAD DE INFORMATICA.

Summary: The Evolutionary Computation Interactive brings together a set of techniques based on a metaphor of biological evolution. The success of many lies in the maintenance of genetic diversity, it is recommended to work with sufficiently large populations of individuals. However, it is not always possible to work with such people, such as when the values of adequacy must be estimated by a human being (Interactive Evolutionary Computation, IEC), or when the evaluation function requires a very high computational cost. This paper presents a new computational method that is capable of working with very small populations (micropoblaciones), which accelerates the convergence of the algorithm end making it more effective.

Author: SAQUETE BORO ESTELA.
Year: 2004.
University: ALICANTE [More theses of this university] [www.ua.es].
Place of defense: ESCUELA POLITÉCNICA SUPERIOR.
Place of preparation: UNIVERSIDAD DE ALICANTE.

Summary: This thesis has developed a system for the recognition and resolution of temporary expressions in written texts, making such expressions on dates or periods of time obsolutas according to a formal model temporary. This transformation allows for the management of events in the texts and their application to the search for answers. The system called TERSEO developed in Spanish in a first phase, but was subsequently extended to work at a level multilingual.

Author: MEDINA PASAJE JULIO LUIS.
Year: 2004.
University: LEÓN [More theses of this university] [www.unileon.es].
Place of defense: FACULTAD DE CIENCIAS.
Place of preparation: FACULTAD DE CIENCIAS.

Summary: The objective of this work is to define the methodology UML-MAST for representation and analysis of the behavior of real-time systems that have been designed using object-oriented paradigm. The methodology proposed reconciles differences in the design of real-time systems from the point of view of the designer of object-oriented systems. In object-oriented systems are grouped attributes and operations using criteria domain, whereas the real-time systems are reactive and studied by the joint concurrent transactions under consideration by the whole of concurrent transactions that are executed in response to events coming from the environment or the timing. To this end, it has established a level of abstraction suitable for the elements of modeling the behavior of real-time models that can be made with a parallel structure to the logical architecture of the system, and also forge partnerships among elements of the standard time royal and the logical model. The semantic modeling has been used to follow the recommendations of UML profile for planificabilidad, performance and time (SPT) standardized by the GMO, and that UML-MAST can be considered an implementation. The proposed methodology has been developed using UML in order to integrate it with a conceptual representations functional deployment and fulfillment for object-oriented systems that are employed, in order to be supported by tools in common use. The UML language has been used to define the elements of a conceptual modeling through a metamoldeo, that the model of any system is an instance, as well as the language with which makes these models. The methodology proposed is part of the tools of analysis and design of real-time systems MAST (Modeling and Analysis Suite for Real-Time Applications) as a new tool for generating models. These models are compiled and analyzed MAST, and the results are referenced to the initial model for its interpretation by the designer. It has also defined criteria for the extension of this methodology to other levels of abstraction such as systems based on components and systems implemented using Ada 95. Part of the results of this work have been incorporated by the GMO their profile PTS.

Author: FERNANDEZ GOMEZ MANUEL.
Year: 2004.
University: POLITÉCNICA DE CATALUÑA [More theses of this university] [www.upc.edu].
Place of defense: C6-E101.
Place of preparation: D-6 201 Nord.

Summary: Two of the most important limitations performance in processors today are derived from the operations of the memory and control units. To solve these problems, cache memories and forecasters jump are two alternatives well known hardware that operate, among other factors, reuse temporary memory and the correlation of jumps. In other words, these structures are trying to exploit redundancy dynamics existing programs. This redundancy comes partly from the way programmers write code, but also limitations on the traditional model of compilation, which introduces instructions from memory and jump unnecessary. We think that the compilers should be very aggressive optimizing programs, and thus be able to eliminate a significant portion of this redundancy. On the other hand, optimizations implemented in time or link directly to the final executable program have received increasing attention in recent years, due to limitations on the traditional model of compilation. Even applying sophisticated analysis and transformations interprocedurales, a traditional compiler is not able to optimize a program as a complete entity. A similar problem appears techniques compile led profiling: major projects were forced to recompile all and each of its modules to exploit such information. Rather, it would build more convenient application complete, instrumentarla to obtain information about profiling and then optimizing the binary end without recompiling not a single source file. In this thesis we present new techniques compile profiling led to eliminate redundancy found in a standard binary executable programs (ie redundant binary), even though these programs have been compiled aggressively with a brand new commercial compiler. Our removal techniques are designed to eliminate redundant operations and jump redundant memory, which are the most important to alleviate the performance problems that we have mentioned. These proposals are based on techniques for eliminating redundancy partial sensitive to the path of execution. The results show that by applying our optimizations, we are able to achieve a reduction of 14% in the execution time of our whole program. This paper also reviewed the problems of the analysis of aliases in executable programs, identifying why the memory disambiguation is one of the weaknesses in modifying object code. We propose several analysis to be applied in the context of optimizers binaries. First, an analysis of aliases strict memory units to discover sensitive to the path of implementation, which is used in our optimizations for the elimination of redundancies memory. Then, two speculative analysis of potential aliases for detection of independence memory. These tests are based on speculative enter information in time for analysis, which increases the accuracy substantial portions of code to maintain efficient analysis. The results show that our proposals are highly useful for enhancing memory disambiguation binary code, which translates into opportunities for optimizations. All our algorithms, analysis and optimization, have been implemented in a binary optimization, emphasizing the most important problems in the applications of our algorithms into executable code, without the help of much of the information present in high-level compilers traditional .

Author: FILALI BOUAMI MOUNCEF.
Year: 2004.
University: GRANADA [More theses of this university] [www.ugr.es].
Place of defense: ESCUELA TÉCNICA SUPERIOR DE INGENIERÍA INFORMÁTICA.
Place of preparation: ESCUELA TÉCNICA SUPERIOR DE INGENIERÍA INFORMÁTICA.

Summary: The dissertation develops and optimization of new models of networks based on the basic functions provided by radio. D. Mouncef Filali Bouami, graduated in Physics from the University of Fez (Morocco) and directed by Professors Doctors D. Ignacio Rojas Ruiz and D. Carlos Garcia Puntonet Department of Computer Architecture and Technology at the University of Grenada, describes a new method and new structures for the development of systems based radio neurons. In this paper presents a structure novel esturcturas for the development of systems based radio neurons. In this paper presents a novel structure called S-RBF (Sum Radial Basis Function) with new entrants T-norma and T-conorma. These operators, combined with weights involved in consexiones "entrada-capa hidden" allow for a "hierarchy" of input vectors you in the activation of each neuron, advocating that the neurons in the hidden layer have greater flexibility, since its activation can be located within a subset of the input space. S-RBF uses an algorithm to find generic, automatically network functions based radical solve a given problem, in an integrated architecture, and taking into account the principal goal of minimizing the computational error and the cost to make it task. In the roles of baser radio parameters connecting calls d elas hidden neurons with the output neuron, normalmenta are simple parameters (constant). But there is the possibility of generalizing the expression of these parameters, thereby increasing their ability expressive, using what are called regression weights. In this thesis are considered expressions, polynomial of order 2, for the weights of neural output of the system, along with the development of algorithms to optimize the structure learning focused proposal. The robustness of the method S-RBF has been tested by applying it to a battery of roles of different sizes, and the results are really satisfying. Taking into account the contents of the thesis, fields considered in the same mainly focuses on: A. Our structures of neural networks based radio. B. Algorithms genérticos. C. Neural Architecture with regression in the variable output. D. System bio-inspirados.

Author: ANAYA RIVERA ELY KARINA.
Year: 2004.
University: GRANADA [More theses of this university] [www.ugr.es].
Place of defense: ETS. DE INGENIERÍA INFORMÁTICA.
Place of preparation: ETS. DE INGENIERÍA INFORMÁTICA.

Summary: Proposal to implement a model of learning in a virtual Mexican technological university. To achieve our objective develop three phases: * Analysis of the educational model of the Mexican technological universities with the aim of identifying, its current problems and thus present a series of viable alternatives to the Virtual Learning. We present a methodology for change, starting from the study. * Design model didáctico-tecnológico learning for application in the virtuality based on the methodology of change in the Technological University. IMPLEMENTATION * Implementation of a real experience for training through the launch of a virtual classroom. The course is based on the virtual model didáctico-tecnológico proposed. * Implementation of a scheme assessment to evaluate the functionality of the model through the satisfaction of the students who participated in the virtual classroom. With this implementation process: analysis, design and implementation is intended to address a number of needs for flexible training that the university demanding on this University.

Author: CATALDI LEIVA ZULMA.
Year: 2005.
University: SEVILLA [More theses of this university] [www.us.es].
Place of defense: AULA DE GRADOS - UNIVERSIDAD DE SEVILLA.
Place of preparation: FACULTAD DE CIENCIAS DE LA EDUCACIÓN - UNIVERSIDAD DE SEVILLA.

Summary: This thesis presents the investigations carried out during between 2001 and 2004 to inquire about the contributions of major information technology based learning problems at the university level. In the experiments described arises in the group work cooperatively and collaborative mediated through computer technology and networks aimed at promoting learning more durable involving the development of skills that students need in their workplace. It seeks to use the resources that are available, such as electronic mail in the form of discussion forum and chat in the most productive way possible to enhance the level of interaction between participants. Particularly, explores what's happening in a discussion forum considering what incentives can be generic skills in participants through mediations. It describes the experiments and comparing levels of promoting students with previous data. In these contexts rich interactions highlights the profile of the student as the main actor in the process of teaching and as coordinator. One of the issues explored, focuses on the study of what happens in the group that negotiates and builds meanings in its evolution. The group will pass through different stages, which are characterized by acquiring certain configurations where each process group gives it some uniqueness. They pass through various types of conflicts and their resolution allows the continuation of the process. Finally, based on the comments, a proposal of theoretical model of evolution of knowledge in the learning groups.

Author: COLOMO PALACIOS RICARDO.
Year: 2005.
University: POLITÉCNICA DE MADRID [More theses of this university] [www.upm.es].
Place of defense: FACULTAD DE INFORMATICA UNIVERSIDAD POLITECNICA DE MADRID.
Place of preparation: FACULTAD DE INFORMATICA UNIVERSIDAD POLITECNICA DE MADRID.

Summary: This thesis develops a framework for evaluating the general skills and techniques of software engineers within the organizations in which they carry out their work. The Framework aims to serve as a basis for determining levels of powers objective cash for diferntes professional profiles that are filed with the Software Engineering. The framework consists of two elements: a proposal for professional structure for the Software Engineering and a method of competency assessment for such engineers. The first professional structure, and integrates career professionals, skills and competence levels for each professional figure. The second, the Framework consists of a module based competency assessment and Feedback 360 that essoportado by a software tool. The Framework has been implemented and validated empirically three Spanish organizations.

Author: ROSALES RAMIREZ MISAEL DARIO.
Year: 2005.
University: AUTÓNOMA DE BARCELONA [More theses of this university] [www.uab.es].
Place of defense: CENTRE DE VISIO PER COMPUTADOR.
Place of preparation: CENTRE DE VISIO PER COMPUTADOR.

Summary: Using images UltraSonidoIntraVascular (IVUS) for the extraction of quantitative information is helpful in the diagnosis and treatment of vessels arterioscleróticos. At present, this kind of analysis is carried out by doctors in the form semi-empírica without taking into account the principles of image generation in the same way that neither takes into account the temporal evolution of the wall glass. This impoverishes the use of this technique, because the analysis and interpretation of the images is based only on the values of gray levels and the behavior morfo-geométrico static wall of glass. Taking into account this lack technical and conceptual, we have divided this dissertation research in four ways: First, referring to this experimental work, which aims to justify our main hypotheses. Secondly, we have developed and implemented a basic model simulation for the purpose of obtaining data IVUS in synthetic form. Thirdly, we have developed a new methodology that allows us to estimate overall terms catheter kinematic behavior of the vessel wall and finally we have proposed a generic model that allows for the reconstruction of geometric glass in 2.5D. The experimental work was done in Vitro with hearts of pig and calf, focusing primarily to study the micro-arquitectura the wall of glass and geometric amendment because external parameters. We know rado and analyzed from the standpoint of images IVUS, the evolution morpho-geométrica the wall of glass in the longitudinal cuts due to the pressure pulsed radio and an external rotation induced heart experimentation. The dynamic analysis was carried out in Vitro extrapolated to a sample of 30 clinical cases conducted by the Laboratory of Hemodinámica "University Hospital Germans Tries i Pu- jol" Badalona-España in cooperation with Boston Scientic-USA. In regard to the simulation model to obtain data IVUS 2D, is based on the basic principles of transmission and reception of sound waves of high frequency to penetrate these radially structure simulated blood. We have made an extensive validation by comparing simulated and real images, contrasting with the view of experts who have outlined manually the actual configuration of layers of glass. For this purpose, we define procedures for extracting quantitative parameters to measure global and local similarities in the images obtained. The model has different utilities. First, medical experts and researchers in the field of vision, can generate simulated images with the aim of observing various arterial structures of interest and clinical comparison with the distribution of gray levels in actual images. Second, researchers and clinicians can use our model to learn and compare the influence of different physical parameters in image formation IVUS. Thirdly, this model can be used to generate large databases synthetic under different conditions and parameters of acquisition, to be used in the validation of the robustness of image processing techniques. The third area of research that we have addressed in this thesis has been the development of a geometrical and kinematic model, which aims to study the evolution of the wall of the coronary arteries from the point of view of the catheter. The model is based on experimental results which show that changes in the arterial wall in a sequence of IVUS, can be modeled by assuming two major contributions, which come from different natural reasons, the first contribution is caused by the inherent geometric properties artery and the second contribution comes from the dynamics of oscillating ventricle. These contributions govern greater appearance of the profiles of the arterial wall in the longitudinal view. Using these statements, we have generated strategies methodological 8 with 5fb purpose of estimating and remove distortions dynamic sequences IVUS. The new conceptual formulation allows us to separate the contributions dependent geometric architecture inherent in the glass wall of the contributions from the dynamic movement of the heart. Finally, given that this separation is possible, we have created a basic foundation for the reconstruction 2.5D vessel from the point of view of the catheter, only using data IVUS. This approach could also be used as an alternative technique for estimating the dynamics of heart since the model contains explicit in the parameters needed to quantify. Here, in this thesis provides a new option that allows enhanced technical IVUS as a tool for exploration and medical diagnosis.

Author: MARTIN MONTES ANTONIO.
Year: 2005.
University: SEVILLA [More theses of this university] [www.us.es].
Place of defense: E.T.SUPERIOR DE INGENIERIA INFORMATICA.
Place of preparation: E.T.SUPERIOR DE INGENIERÍA INFORMÁTICA.

Summary: The management expert networks and services has been an area where traditionally have been imposed solutions and mecanisos owners from different manufacturers. The current management models we are endowed with the necessary characteristics for efficient administration of knowledge management provided by Expert Systems. There is a need to develop models of control and supervision offering greater possibilities. Here it is proposed standardization of the database of knowledge management necessary to manage resources in the various telecommunications networks. This is a key aspect in the management model proposed, the aim is to achieve a definition sintacticamente uniform of all knowledge management contributing experts, regardless of the manufacturer managed resource. This poses a great effort, it is necessary to conduct a uniform definition of all knowledge management applicable to all communication resources available. We need cocebir the definition of knowledge management as a task in the design of new communications resources. From this arise denominated Integrated Management System Expert, a new paradigm that will contain aspects clearly direrenciadores to the management techniques used so far, making a separate application of knowledge provided by the expert systems and information management resources managed. In our proposal knowledge management and information management are part of the management model, in our case the OSI Model. Knowledge of the Expert Systems will be treated as a property belonging to more managed object.

Author: Mouriño Gallego Jose Carlos.
Year: 2005.
University: A CORUÑA [More theses of this university] [www.udc.es].
Place of defense: Facultad de Informática.
Place of preparation: Facultad de Informática.

Summary: It has addressed the shortening of application execution STEM-II for the simulation of the dispersion of pollutants produced by a thermal power station. We have managed to get response times that allow real application with a high precision and has set the agenda for a graphical user interface for the user. The codes have been adapted to parallel multiprocessor systems and distributed memory systems GRID. For systems GRID has been added fault tolerance techniques checkpointing.

Author: Molina Fernández Carlos.
Year: 2005.
University: GRANADA [More theses of this university] [www.ugr.es].
Place of defense: E.T.S. Ingneniería en Informática.
Place of preparation: E.T.S. Ingeniería en Informática.

Summary: The use of OLAP technology in new areas of knowledge and use of unstructured data sources have made the emergence of new requirements on the models used in defining the datacubos. It requires the use of flexible structures that enable modeling and work on the vagueness of the resulting integration. One source of the information would be interesting data compiled by experts to enrich the analysis. Normally this information will be given so vague. In these systems also tend to integrate data mining techniques to enrich the possibilities of obtaining information for the user. These techniques should also be able to work on the inaccuracy of this data. What we propose in this dissertation is on the one hand a multidimensional model which through the use of fuzzy logic allows for the representation and management of information inaccurate. This allows both facts fuzzy definition of the use of hierarchies diffuse to relate the concepts in question. This would represent concepts closest to the user. But it is possible that the vagueness arises not only in defining the concepts but also the relationships between them. To assist in these cases, the model allows to define the hierarchical relationships using labels lingüsticas, allowing the expert delivering using expressions lingüsticas. This model has set up normal operations on datacubos. To isolate the user from the complexity added, the proposed structure has been wrapped using operators summary diffuse, so that the user is accessing the information in a more intuitive. To demonstrate its application had been built three datacubos on actual data taken from surgical interventions, accounting data and the third on census data. To complete the possibility of analysis, has developed a technique for extracting association rules on the multidimensional structure. The algorithm is called COGARE and its main objective is to reduce the complexity of the outcome. This is a collection of rules that works with multi-level relationships between elements of different dimensions. To control the complexity have identified two factors: the number of rules and the level of detail of the elements being considered. Both have been combined into a single measure to control the extraction process. The technique has been tested on datacubos commented earlier, getting an average reduction of the complexity around 39% (to be around 43% for the number of rules). Both results are implemented in a prototype system OLAP: F-Cube Factory. The report presents the main features of the same.