Composição de dicionários visuais utilizando agrupamento de dados por Florestas de Caminhos Ótimos

Pós-graduação em Ciência da Computação - IBILCE

Semi-supervised learning with concept drift using particle dynamics applied to network intrusion detection data

Concept drift, which refers to non stationary learning problems over time, has increasing importance in machine learning and data mining. Many concept drift applications require fast response, which means an algorithm must always be (re)trained with the latest available data. But the process of data labeling is usually expensive and/or time consuming when compared to acquisition of unlabeled data, thus usually only a small fraction of the incoming data may be effectively labeled. Semi-supervised learning methods may help in this scenario, as they use both labeled and unlabeled data in the training process. However, most of them are based on assumptions that the data is static. Therefore, semi-supervised learning with concept drifts is still an open challenging task in machine learning. Recently, a particle competition and cooperation approach has been developed to realize graph-based semi-supervised learning from static data. We have extend that approach to handle data streams and concept drift. The result is a passive algorithm which uses a single classifier approach, naturally adapted to concept changes without any explicit drift detection mechanism. It has built-in mechanisms that provide a natural way of learning from new data, gradually "forgetting" older knowledge as older data items are no longer useful for the classification of newer data items. The proposed algorithm is applied to the KDD Cup 1999 Data of network intrusion, showing its effectiveness.

Designing Software Product Lines for Testability

Software product line (SPL) engineering offers several advantages in the development of families of software products such as reduced costs, high quality and a short time to market. A software product line is a set of software intensive systems, each of which shares a common core set of functionalities, but also differs from the other products through customization tailored to fit the needs of individual groups of customers. The differences between products within the family are well-understood and organized into a feature model that represents the variability of the SPL. Products can then be built by generating and composing features described in the feature model. Testing of software product lines has become a bottleneck in the SPL development lifecycle, since many of the techniques used in their testing have been borrowed from traditional software testing and do not directly take advantage of the similarities between products. This limits the overall gains that can be achieved in SPL engineering. Recent work proposed by both industry and the research community for improving SPL testing has begun to consider this problem, but there is still a need for better testing techniques that are tailored to SPL development. In this thesis, I make two primary contributions to software product line testing. First I propose a new definition for testability of SPLs that is based on the ability to re-use test cases between products without a loss of fault detection effectiveness. I build on this idea to identify elements of the feature model that contribute positively and/or negatively towards SPL testability. Second, I provide a graph based testing approach called the FIG Basis Path method that selects products and features for testing based on a feature dependency graph. This method should increase our ability to re-use results of test cases across successive products in the family and reduce testing effort. I report the results of a case study involving several non-trivial SPLs and show that for these objects, the FIG Basis Path method is as effective as testing all products, but requires us to test no more than 24% of the products in the SPL.

A regional multirnodulus algorithm for blind equalization of QAM signals: Introduction and steady-state analysis

It is well known that constant-modulus-based algorithms present a large mean-square error for high-order quadrature amplitude modulation (QAM) signals, which may damage the switching to decision-directed-based algorithms. In this paper, we introduce a regional multimodulus algorithm for blind equalization of QAM signals that performs similar to the supervised normalized least-mean-squares (NLMS) algorithm, independently of the QAM order. We find a theoretical relation between the coefficient vector of the proposed algorithm and the Wiener solution and also provide theoretical models for the steady-state excess mean-square error in a nonstationary environment. The proposed algorithm in conjunction with strategies to speed up its convergence and to avoid divergence can bypass the switching mechanism between the blind mode and the decision-directed mode. (c) 2012 Elsevier B.V. All rights reserved.

Über das Vermessen tubulärer Strukturen in der Computertomographie

Die chronisch obstruktive Lungenerkrankung (engl. chronic obstructive pulmonary disease, COPD) ist ein Überbegriff für Erkrankungen, die zu Husten, Auswurf und Dyspnoe (Atemnot) in Ruhe oder Belastung führen - zu diesen werden die chronische Bronchitis und das Lungenemphysem gezählt. Das Fortschreiten der COPD ist eng verknüpft mit der Zunahme des Volumens der Wände kleiner Luftwege (Bronchien). Die hochauflösende Computertomographie (CT) gilt bei der Untersuchung der Morphologie der Lunge als Goldstandard (beste und zuverlässigste Methode in der Diagnostik). Möchte man Bronchien, eine in Annäherung tubuläre Struktur, in CT-Bildern vermessen, so stellt die geringe Größe der Bronchien im Vergleich zum Auflösungsvermögen eines klinischen Computertomographen ein großes Problem dar. In dieser Arbeit wird gezeigt wie aus konventionellen Röntgenaufnahmen CT-Bilder berechnet werden, wo die mathematischen und physikalischen Fehlerquellen im Bildentstehungsprozess liegen und wie man ein CT-System mittels Interpretation als lineares verschiebungsinvariantes System (engl. linear shift invariant systems, LSI System) mathematisch greifbar macht. Basierend auf der linearen Systemtheorie werden Möglichkeiten zur Beschreibung des Auflösungsvermögens bildgebender Verfahren hergeleitet. Es wird gezeigt wie man den Tracheobronchialbaum aus einem CT-Datensatz stabil segmentiert und mittels eines topologieerhaltenden 3-dimensionalen Skelettierungsalgorithmus in eine Skelettdarstellung und anschließend in einen kreisfreien Graphen überführt. Basierend auf der linearen System Theorie wird eine neue, vielversprechende, integral-basierte Methodik (IBM) zum Vermessen kleiner Strukturen in CT-Bildern vorgestellt. Zum Validieren der IBM-Resultate wurden verschiedene Messungen an einem Phantom, bestehend aus 10 unterschiedlichen Silikon Schläuchen, durchgeführt. Mit Hilfe der Skelett- und Graphendarstellung ist ein Vermessen des kompletten segmentierten Tracheobronchialbaums im 3-dimensionalen Raum möglich. Für 8 zweifach gescannte Schweine konnte eine gute Reproduzierbarkeit der IBM-Resultate nachgewiesen werden. In einer weiteren, mit IBM durchgeführten Studie konnte gezeigt werden, dass die durchschnittliche prozentuale Bronchialwandstärke in CT-Datensätzen von 16 Rauchern signifikant höher ist, als in Datensätzen von 15 Nichtrauchern. IBM läßt sich möglicherweise auch für Wanddickenbestimmungen bei Problemstellungen aus anderen Arbeitsgebieten benutzen - kann zumindest als Ideengeber dienen. Ein Artikel mit der Beschreibung der entwickelten Methodik und der damit erzielten Studienergebnisse wurde zur Publikation im Journal IEEE Transactions on Medical Imaging angenommen.

Hypergraph Mining for Social Networks

Nowadays, more and more data is collected in large amounts, such that the need of studying it both efficiently and profitably is arising; we want to acheive new and significant informations that weren't known before the analysis. At this time many graph mining algorithms have been developed, but an algebra that could systematically define how to generalize such operations is missing. In order to propel the development of a such automatic analysis of an algebra, We propose for the first time (to the best of my knowledge) some primitive operators that may be the prelude to the systematical definition of a hypergraph algebra in this regard.

A framework for cyber-physical system simulation

In these last years, systems engineering has became one of the major research domains. The complexity of systems has increased constantly and nowadays Cyber-Physical Systems (CPS) are a category of particular interest: these, are systems composed by a cyber part (computer-based algorithms) that monitor and control some physical processes. Their development and simulation are both complex due to the importance of the interaction between the cyber and the physical entities: there are a lot of models written in different languages that need to exchange information among each other. Normally people use an orchestrator that takes care of the simulation of the models and the exchange of informations. This orchestrator is developed manually and this is a tedious and long work. Our proposition is to achieve to generate the orchestrator automatically through the use of Co-Modeling, i.e. by modeling the coordination. Before achieving this ultimate goal, it is important to understand the mechanisms and de facto standards that could be used in a co-modeling framework. So, I studied the use of a technology employed for co-simulation in the industry: FMI. In order to better understand the FMI standard, I realized an automatic export, in the FMI format, of the models realized in an existing software for discrete modeling: TimeSquare. I also developed a simple physical model in the existing open source openmodelica tool. Later, I started to understand how works an orchestrator, developing a simple one: this will be useful in future to generate an orchestrator automatically.

Using citation data to improve retrieval from MEDLINE.

OBJECTIVE: To determine whether algorithms developed for the World Wide Web can be applied to the biomedical literature in order to identify articles that are important as well as relevant. DESIGN AND MEASUREMENTS A direct comparison of eight algorithms: simple PubMed queries, clinical queries (sensitive and specific versions), vector cosine comparison, citation count, journal impact factor, PageRank, and machine learning based on polynomial support vector machines. The objective was to prioritize important articles, defined as being included in a pre-existing bibliography of important literature in surgical oncology. RESULTS Citation-based algorithms were more effective than noncitation-based algorithms at identifying important articles. The most effective strategies were simple citation count and PageRank, which on average identified over six important articles in the first 100 results compared to 0.85 for the best noncitation-based algorithm (p < 0.001). The authors saw similar differences between citation-based and noncitation-based algorithms at 10, 20, 50, 200, 500, and 1,000 results (p < 0.001). Citation lag affects performance of PageRank more than simple citation count. However, in spite of citation lag, citation-based algorithms remain more effective than noncitation-based algorithms. CONCLUSION Algorithms that have proved successful on the World Wide Web can be applied to biomedical information retrieval. Citation-based algorithms can help identify important articles within large sets of relevant results. Further studies are needed to determine whether citation-based algorithms can effectively meet actual user information needs.

Simulation in vascular access surgery

In the last years, simulation training has become widespread in different areas of medicine due to social expectations, political accountability and professional regulation. Different types of simulators allow to improve knowledge, skills, communication and team behavior. Simulation sessions have been proven to shorten the learning curve and allow education in a safe environment. Patients on dialysis are an expanding group. They often suffer from several comorbidities and need complex surgical procedures with regard to their dialysis access. Therefore, education in evidence-based algorithms is as important as teaching of practical skills. In this chapter, we are presenting an overview of available dialysis access training modalities. We are convinced that simulation will become more important in the near future and has a substantial impact on strategies to improve aspects of patient safety. © 2015 S. Karger AG, Basel.

Fully Automatic Segmentation of Hip CT Images

Automatic segmentation of the hip joint with pelvis and proximal femur surfaces from CT images is essential for orthopedic diagnosis and surgery. It remains challenging due to the narrowness of hip joint space, where the adjacent surfaces of acetabulum and femoral head are hardly distinguished from each other. This chapter presents a fully automatic method to segment pelvic and proximal femoral surfaces from hip CT images. A coarse-to-fine strategy was proposed to combine multi-atlas segmentation with graph-based surface detection. The multi-atlas segmentation step seeks to coarsely extract the entire hip joint region. It uses automatically detected anatomical landmarks to initialize and select the atlas and accelerate the segmentation. The graph based surface detection is to refine the coarsely segmented hip joint region. It aims at completely and efficiently separate the adjacent surfaces of the acetabulum and the femoral head while preserving the hip joint structure. The proposed strategy was evaluated on 30 hip CT images and provided an average accuracy of 0.55, 0.54, and 0.50 mm for segmenting the pelvis, the left and right proximal femurs, respectively.

Decision Support Models for Composing and Navigating through e-Learning Objects

Libraries of learning objects may serve as basis for deriving course offerings that are customized to the needs of different learning communities or even individuals. Several ways of organizing this course composition process are discussed. Course composition needs a clear understanding of the dependencies between the learning objects. Therefore we discuss the metadata for object relationships proposed in different standardization projects and especially those suggested in the Dublin Core Metadata Initiative. Based on these metadata we construct adjacency matrices and graphs. We show how Gozinto-type computations can be used to determine direct and indirect prerequisites for certain learning objects. The metadata may also be used to define integer programming models which can be applied to support the instructor in formulating his specifications for selecting objects or which allow a computer agent to automatically select learning objects. Such decision models could also be helpful for a learner navigating through a library of learning objects. We also sketch a graph-based procedure for manual or automatic sequencing of the learning objects.

Retinal slit lamp video mosaicking

Purpose To this day, the slit lamp remains the first tool used by an ophthalmologist to examine patient eyes. Imaging of the retina poses, however, a variety of problems, namely a shallow depth of focus, reflections from the optical system, a small field of view and non-uniform illumination. For ophthalmologists, the use of slit lamp images for documentation and analysis purposes, however, remains extremely challenging due to large image artifacts. For this reason, we propose an automatic retinal slit lamp video mosaicking, which enlarges the field of view and reduces amount of noise and reflections, thus enhancing image quality. Methods Our method is composed of three parts: (i) viable content segmentation, (ii) global registration and (iii) image blending. Frame content is segmented using gradient boosting with custom pixel-wise features. Speeded-up robust features are used for finding pair-wise translations between frames with robust random sample consensus estimation and graph-based simultaneous localization and mapping for global bundle adjustment. Foreground-aware blending based on feathering merges video frames into comprehensive mosaics. Results Foreground is segmented successfully with an area under the curve of the receiver operating characteristic curve of 0.9557. Mosaicking results and state-of-the-art methods were compared and rated by ophthalmologists showing a strong preference for a large field of view provided by our method. Conclusions The proposed method for global registration of retinal slit lamp images of the retina into comprehensive mosaics improves over state-of-the-art methods and is preferred qualitatively.

A Comparison of Hydrographically and Optically Derived Mixed Layer Depths

Efforts to understand and model the dynamics of the upper ocean would be significantly advanced given the ability to rapidly determine mixed layer depths (MLDs) over large regions. Remote sensing technologies are an ideal choice for achieving this goal. This study addresses the feasibility of estimating MLDs from optical properties. These properties are strongly influenced by suspended particle concentrations, which generally reach a maximum at pycnoclines. The premise therefore is to use a gradient in beam attenuation at 660 nm (c660) as a proxy for the depth of a particle-scattering layer. Using a global data set collected during World Ocean Circulation Experiment cruises from 1988-1997, six algorithms were employed to compute MLDs from either density or temperature profiles. Given the absence of published optically based MLD algorithms, two new methods were developed that use c660 profiles to estimate the MLD. Intercomparison of the six hydrographically based algorithms revealed some significant disparities among the resulting MLD values. Comparisons between the hydrographical and optical approaches indicated a first-order agreement between the MLDs based on the depths of gradient maxima for density and c660. When comparing various hydrographically based algorithms, other investigators reported that inherent fluctuations of the mixed layer depth limit the accuracy of its determination to 20 m. Using this benchmark, we found a similar to 70% agreement between the best hydrographical-optical algorithm pairings.

Análisis biomecánico del segmento de la carrera a pie de la competición en triatlón

El principal objetivo de la presente investigación fue el conocer el perfil de rendimiento técnico de los triatletas, desde un punto de vista biomecánica, en el segmento carrera a pie durante la competición en triatlón. Asimismo, como el genero y el nivel deportivo del triatleta podrían influir en su respuesta motriz durante la competicion. Para ello, se necesitaba desarrollar y validar una técnica experimental que fuera lo suficientemente precisa (validez interna), con una alta fiabilidad y con una gran validez externa (ecologica) debido al entorno de la competicion. La muestra la formaron un total de 64 deportistas: 32 triatletas participantes en la Copa del Mundo de Triatlon de Madrid-2008 (16 hombres y 16 mujeres) y 32 triatletas participantes en el Clasificatorio del Campeonato de Espana Elite (16 hombres y 16 mujeres). El análisis de la técnica de carrera de los deportistas se realizo mediante un sistema fotogramétrico en 2d que permitió calcular las coordenadas (x,y) de los centros articulares con un error de 1.66% en el eje x y de un 2.10% en el eje y. Las imágenes fueron obtenidas por una cámara que filmaba el movimiento en un plano antero-posterior del triatleta. Algoritmos basados en la DLT (Abdel-Aziz & Karara, 1971) permitieron conocer las coordenadas reales a partir de las coordenadas digitalizadas en el plano y posteriormente las distintas variables analizadas. El análisis biomecánica de la carrera se realizo en 4 ocasiones diferentes durante la competición, correspondiendo con cada una de las vueltas de 2,5 km, que el triatleta tenía que realizar. La velocidad de carrera resulto estar íntimamente ligada al nivel deportivo del triatleta. Del mismo modo, 3 de los 4 grupos analizados presentaron valores inferiores a 3 minutos 30 segundos por kilometro recorrido, poniendo de manifiesto el altísimo nivel de los sujetos analizados. Del mismo modo parece que las chicos consiguen una mayor velocidad gracias a una mayor longitud de ciclo en relación a las chicas, ya que estas muestran valores mayores en cuanto a frecuencia de zancada. La frecuencia de zancada presento los valores más altos en la primera vuelta en todos los deportistas analizados. Asimismo, los triatletas de nivel internacional y las chicas fueron los que mostraron los mayores valores. La longitud de zancada presento distintas tendencias en función del nivel y el género del deportista. Así pues, en los deportistas internacionales y en los chicos los mayores valores se encontraron en la primera vuelta mientras que la tendencia fue al descenso, siendo probablemente la fatiga acumulada la causante de dicha tendencia. En cambio, aquellos deportistas de nivel nacional y las chicas mostraron valores mayores en la segunda vuelta que en la primera, evidenciando que además de la fatiga, el ciclismo previo tiene una incidencia directa sobre su rendimiento. Los tiempos de vuelo permanecieron constantes durante toda la carrera, encontrando cierta evolución en los tiempos de apoyo, la cual provoca una modificación en los porcentajes relativos en los tiempos de vuelo. Los tiempos de apoyo más bajos se encontraron en la primera vuelta. Del mismo modo, los deportistas de nivel internacional y los chicos mostraron valores inferiores. También, estos grupos fueron más constantes en sus valores a lo largo de las vueltas. Por el contrario, se encontraron tendencias al aumento en los triatletas de nivel nacional y en las chicas, los cuales no fueron capaces de mantener el mismo rendimiento debido seguramente a su menor nivel deportivo. La oscilación vertical de la cadera se mostro constante en los triatletas de mayor nivel, encontrándose tendencias al aumento en los de menor nivel. Del mismo modo, los valores más altos correspondieron a las chicas y a los deportistas de nivel nacional. La distancia de la cadera al apoyo permaneció constante a lo largo de las vueltas en todos los grupos, obteniéndose valores mayores en los triatletas de nivel internacional y en los chicos. El ángulo de la rodilla apoyada en el momento del despegue no mostro una tendencia clara. Los deportistas de nivel internacional y los chicos presentaron los valores más bajos. El ángulo de la rodilla libre en el momento del despegue mostro una correlación muy alta con la velocidad de carrera. Del mismo modo, los ángulos más pequeños se encontraron en los triatletas internacionales y en los chicos, debido seguramente a los mayores valores de velocidad registrados por ambos grupos. Los ángulos de los tobillos no mostraron ninguna tendencia clara durante la competición analizada. Los cuatro grupos de población presentaron valores similares, por lo que parece que no representan una variable que pueda incidir sobre el rendimiento biomecánica del triatleta. Los resultados obtenidos en el presente estudio de investigación avalan la utilización de la fotogrametría-video en 2d para el análisis de la técnica de carrera durante la competición en triatlón. Su aplicación en una competición de máximo nivel internacional ha posibilitado conocer el perfil técnico que presentan los triatletas a lo largo del segmento de carrera a pie. Del mismo modo, se ha podido demostrar como los estudios realizados en laboratorio no reflejan la realidad competitiva de un triatlón de máximo nivel. The aim of this research was to determine the running technique profile during a triathlon competition from a biomechanical perspective. Also, to analyze the triathlete gender’s and level of performance’s influence on this profile in competition. An accurate (internal validity) and reliable methodology with a high external validity (ecological) had to be developed to get those aims in competition. Sixty-four triathletes were analyzed. 32 (16 males, 16 females) took part in the Madrid 2008 Triathlon World Cup and 32 (16 males and 16 females) took part in the Spanish Triathlon National Championships. The biomechanical analyses were carried out by a photogrammetric system that allow to calculate the landmarks coordinates (x,y) with a 1.66% error in x axis, and a 2.10% error in y axis. The frames were obtained with a camera situated perpendicular to the triathletes’ trajectory, filming the saggittal plane. DLT based algorithms (Abdel-Aziz & Karara, 1971) were used to calculate the real coordinates from the digitalized ones and the final variables afterwards. The biomechanical analisys itself was performed in four different moments during the competition, according to each 2.5 km lap the triathletes had to do. Running speed was highly related to performance level. Also, 3 of the 4 analyzed groups showed speed values under the 3 minutes and 30 seconds per kilometer. It demonstrated the very high performance level of the analized triathletes. Furthermore, it seems that men get higher speeds because their longer stride length, while women shows higher stride frequency values. The highest stride frequency values were found in the first lap. Women and the international level triathletes showed the highest values. Stride length showed different tendencies according to the gender and level of performance. Men and international level triathletes showed the highest level in the first lap and a decreasing tendency after that. The accumulated fatigue was probably the reason of this tendency. On the other hand, higher values than in first lap were found in the second one in women and national level triathletes. It demonstrated the previous cycling can affect to those groups in terms of biomechanics. Flight times remained constant during the running part, while the contact times showed an increasing tendency that caused a variation in flight times percents. The lowest contact times were found in the first lap and in men and international triathletes’ values. Also, these two groups were more consistent during the whole running. On the other hand, increasing tendencies were found in women and national level triathletes, who were not able to maintain the same values probably due to their lower level of performance. Higher level triathletes showed more consistent hip vertical oscillation values than lower level triathletes, who presented increasing tendencies. The highest values were found in women and national level triathletes. The horizontal distance hip-toe cap remained constant among the laps in all the groups. Men and international level triathletes showed the highest values. The support knee angle at toe-off did not show a clear tendency. The lowest values were found in men and international level triathletes. A high correlation was found between the non-support knee angle and the running speed. Furthermore, men and international level triathletes showed the smallest values, due to the higher velocities reached by these two groups. Ankles angles did not show any tendency during the running part. Similar values were found in the four analyzed groups, so this variable does not seem to represent an important one within the triathlete’s performance. The results obtained in the present research support the use of the bidimensional photogrammetric video-system to analyze the running technique during a triathlon competition. Its application in international triathlon meetings has allowed determining the triathletes’ technique profile during the running part. Also, it has been demonstrated the laboratory-based studies does not reproduce a top-level competition.

A hierarchical strategy for real-time tracking on-board UAVs

In this paper, we present a real-time tracking strategy based on direct methods for tracking tasks on-board UAVs, that is able to overcome problems posed by the challenging conditions of the task: e.g. constant vibrations, fast 3D changes, and limited capacity on-board. The vast majority of approaches make use of feature-based methods to track objects. Nonetheless, in this paper we show that although some of these feature-based solutions are faster, direct methods can be more robust under fast 3D motions (fast changes in position), some changes in appearance, constant vibrations (without requiring any specific hardware or software for video stabilization), and situations where part of the object to track is out the field of view of the camera. The performance of the proposed strategy is evaluated with images from real-flight tests using different evaluation mechanisms (e.g. accurate position estimation using a Vicon sytem). Results show that our tracking strategy performs better than well known feature-based algorithms and well known configurations of direct methods, and that the recovered data is robust enough for vision-in-the-loop tasks.