Application of Bayesian approach in geotechnical engineering

En esta Tesis Doctoral se emplean y desarrollan M��todos Bayesianos para su aplicaci��n en an��lisis geot��cnicos habituales, con un ��nfasis particular en (i) la valoraci��n y selecci��n de modelos geot��cnicos basados en correlaciones emp��ricas; en (ii) el desarrollo de predicciones acerca de los resultados esperados en modelos geot��cnicos complejos. Se llevan a cabo diferentes aplicaciones a problemas geot��cnicos, como es el caso de: (1) En el caso de rocas intactas, se presenta un m��todo Bayesiano para la evaluaci��n de modelos que permiten estimar el m��dulo de Young a partir de la resistencia a compresi��n simple (UCS). La metodolog��a desarrollada suministra estimaciones de las incertidumbres de los par��metros y predicciones y es capaz de diferenciar entre las diferentes fuentes de error. Se desarrollan modelos "espec��ficos de roca" para los tipos de roca m��s comunes y se muestra c��mo se pueden "actualizar" esos modelos "iniciales" para incorporar, cuando se encuentra disponible, la nueva informaci��n espec��fica del proyecto, reduciendo las incertidumbres del modelo y mejorando sus capacidades predictivas. (2) Para macizos rocosos, se presenta una metodolog��a, fundamentada en un criterio de selecci��n de modelos, que permite determinar el modelo m��s apropiado, entre un conjunto de candidatos, para estimar el m��dulo de deformaci��n de un macizo rocoso a partir de un conjunto de datos observados. Una vez que se ha seleccionado el modelo m��s apropiado, se emplea un m��todo Bayesiano para obtener distribuciones predictivas de los m��dulos de deformaci��n de macizos rocosos y para actualizarlos con la nueva informaci��n espec��fica del proyecto. Este m��todo Bayesiano de actualizaci��n puede reducir significativamente la incertidumbre asociada a la predicci��n, y por lo tanto, afectar las estimaciones que se hagan de la probabilidad de fallo, lo cual es de un inter��s significativo para los dise��os de mec��nica de rocas basados en fiabilidad. (3) En las primeras etapas de los dise��os de mec��nica de rocas, la informaci��n acerca de los par��metros geomec��nicos y geom��tricos, las tensiones in-situ o los par��metros de sostenimiento, es, a menudo, escasa o incompleta. Esto plantea dificultades para aplicar las correlaciones emp��ricas tradicionales que no pueden trabajar con informaci��n incompleta para realizar predicciones. Por lo tanto, se propone la utilizaci��n de una Red Bayesiana para trabajar con informaci��n incompleta y, en particular, se desarrolla un clasificador Na��ve Bayes para predecir la probabilidad de ocurrencia de grandes deformaciones (squeezing) en un t��nel a partir de cinco par��metros de entrada habitualmente disponibles, al menos parcialmente, en la etapa de dise��o. This dissertation employs and develops Bayesian methods to be used in typical geotechnical analyses, with a particular emphasis on (i) the assessment and selection of geotechnical models based on empirical correlations; on (ii) the development of probabilistic predictions of outcomes expected for complex geotechnical models. Examples of application to geotechnical problems are developed, as follows: (1) For intact rocks, we present a Bayesian framework for model assessment to estimate the Young���s moduli based on their UCS. Our approach provides uncertainty estimates of parameters and predictions, and can differentiate among the sources of error. We develop ���rock-specific��� models for common rock types, and illustrate that such ���initial��� models can be ���updated��� to incorporate new project-specific information as it becomes available, reducing model uncertainties and improving their predictive capabilities. (2) For rock masses, we present an approach, based on model selection criteria to select the most appropriate model, among a set of candidate models, to estimate the deformation modulus of a rock mass, given a set of observed data. Once the most appropriate model is selected, a Bayesian framework is employed to develop predictive distributions of the deformation moduli of rock masses, and to update them with new project-specific data. Such Bayesian updating approach can significantly reduce the associated predictive uncertainty, and therefore, affect our computed estimates of probability of failure, which is of significant interest to reliability-based rock engineering design. (3) In the preliminary design stage of rock engineering, the information about geomechanical and geometrical parameters, in situ stress or support parameters is often scarce or incomplete. This poses difficulties in applying traditional empirical correlations that cannot deal with incomplete data to make predictions. Therefore, we propose the use of Bayesian Networks to deal with incomplete data and, in particular, a Na��ve Bayes classifier is developed to predict the probability of occurrence of tunnel squeezing based on five input parameters that are commonly available, at least partially, at design stages.

Bayesian network modeling of the consensus between experts: an application to neuron classification

Neuronal morphology is hugely variable across brain regions and species, and their classification strategies are a matter of intense debate in neuroscience. GABAergic cortical interneurons have been a challenge because it is difficult to find a set of morphological properties which clearly define neuronal types. A group of 48 neuroscience experts around the world were asked to classify a set of 320 cortical GABAergic interneurons according to the main features of their three-dimensional morphological reconstructions. A methodology for building a model which captures the opinions of all the experts was proposed. First, one Bayesian network was learned for each expert, and we proposed an algorithm for clustering Bayesian networks corresponding to experts with similar behaviors. Then, a Bayesian network which represents the opinions of each group of experts was induced. Finally, a consensus Bayesian multinet which models the opinions of the whole group of experts was built. A thorough analysis of the consensus model identified different behaviors between the experts when classifying the interneurons in the experiment. A set of characterizing morphological traits for the neuronal types was defined by performing inference in the Bayesian multinet. These findings were used to validate the model and to gain some insights into neuron morphology.

Evaluation of a User-Adapted Spoken Language Dialogue System: Measuring the Relevance of the Contextual Information Sources

We present an evaluation of a spoken language dialogue system with a module for the management of userrelated information, stored as user preferences and privileges. The ���exibility of our dialogue management approach, based on Bayesian Networks (BN), together with a contextual information module, which performs different strategies for handling such information, allows us to include user information as a new level into the Context Manager hierarchy. We propose a set of objective and subjective metrics to measure the relevance of the different contextual information sources. The analysis of our evaluation scenarios shows that the relevance of the short-term information (i.e. the system status) remains pretty stable throughout the dialogue, whereas the dialogue history and the user pro���le (i.e. the middle-term and the long-term information, respectively) play a complementary role, evolving their usefulness as the dialogue evolves.

What is behind a summary-evaluation decision?

Research in psychology has reported that, among the variety of possibilities for assessment methodologies, summary evaluation offers a particularly adequate context for inferring text comprehension and topic understanding. However, grades obtained in this methodology are hard to quantify objectively. Therefore, we carried out an empirical study to analyze the decisions underlying human summary-grading behavior. The task consisted of expert evaluation of summaries produced in critically relevant contexts of summarization development, and the resulting data were modeled by means of Bayesian networks using an application called Elvira, which allows for graphically observing the predictive power (if any) of the resultant variables. Thus, in this article, we analyzed summary-evaluation decision making in a computational framework

A Multi-agent System for Emergency Decision Support

This paper describes the multi-agent organization of a computer system that was designed to assist operators in decision making in the presence of emergencies. The application was developed for the case of emergencies caused by river floods. It operates on real-time receiving data recorded by sensors (rainfall, water levels, flows, etc.) and applies multi-agent techniques to interpret the data, predict the future behavior and recommend control actions. The system includes an advanced knowledge based architecture with multiple symbolic representation with uncertainty models (bayesian networks). This system has been applied and validated at two particular sites in Spain (the Jucar basin and the South basin).

Regularization for sparsity in statistical analysis and machine learning

Pragmatism is the leading motivation of regularization. We can understand regularization as a modification of the maximum-likelihood estimator so that a reasonable answer could be given in an unstable or ill-posed situation. To mention some typical examples, this happens when fitting parametric or non-parametric models with more parameters than data or when estimating large covariance matrices. Regularization is usually used, in addition, to improve the bias-variance tradeoff of an estimation. Then, the definition of regularization is quite general, and, although the introduction of a penalty is probably the most popular type, it is just one out of multiple forms of regularization. In this dissertation, we focus on the applications of regularization for obtaining sparse or parsimonious representations, where only a subset of the inputs is used. A particular form of regularization, L1-regularization, plays a key role for reaching sparsity. Most of the contributions presented here revolve around L1-regularization, although other forms of regularization are explored (also pursuing sparsity in some sense). In addition to present a compact review of L1-regularization and its applications in statistical and machine learning, we devise methodology for regression, supervised classification and structure induction of graphical models. Within the regression paradigm, we focus on kernel smoothing learning, proposing techniques for kernel design that are suitable for high dimensional settings and sparse regression functions. We also present an application of regularized regression techniques for modeling the response of biological neurons. Supervised classification advances deal, on the one hand, with the application of regularization for obtaining a na����ve Bayes classifier and, on the other hand, with a novel algorithm for brain-computer interface design that uses group regularization in an efficient manner. Finally, we present a heuristic for inducing structures of Gaussian Bayesian networks using L1-regularization as a filter. El pragmatismo es la principal motivaci��n de la regularizaci��n. Podemos entender la regularizaci��n como una modificaci��n del estimador de m��xima verosimilitud, de tal manera que se pueda dar una respuesta cuando la configuraci��n del problema es inestable. A modo de ejemplo, podemos mencionar el ajuste de modelos param��tricos o no param��tricos cuando hay m��s par��metros que casos en el conjunto de datos, o la estimaci��n de grandes matrices de covarianzas. Se suele recurrir a la regularizaci��n, adem��s, para mejorar el compromiso sesgo-varianza en una estimaci��n. Por tanto, la definici��n de regularizaci��n es muy general y, aunque la introducci��n de una funci��n de penalizaci��n es probablemente el m��todo m��s popular, ��ste es s��lo uno de entre varias posibilidades. En esta tesis se ha trabajado en aplicaciones de regularizaci��n para obtener representaciones dispersas, donde s��lo se usa un subconjunto de las entradas. En particular, la regularizaci��n L1 juega un papel clave en la b��squeda de dicha dispersi��n. La mayor parte de las contribuciones presentadas en la tesis giran alrededor de la regularizaci��n L1, aunque tambi��n se exploran otras formas de regularizaci��n (que igualmente persiguen un modelo disperso). Adem��s de presentar una revisi��n de la regularizaci��n L1 y sus aplicaciones en estad��stica y aprendizaje de m��quina, se ha desarrollado metodolog��a para regresi��n, clasificaci��n supervisada y aprendizaje de estructura en modelos gr��ficos. Dentro de la regresi��n, se ha trabajado principalmente en m��todos de regresi��n local, proponiendo t��cnicas de dise��o del kernel que sean adecuadas a configuraciones de alta dimensionalidad y funciones de regresi��n dispersas. Tambi��n se presenta una aplicaci��n de las t��cnicas de regresi��n regularizada para modelar la respuesta de neuronas reales. Los avances en clasificaci��n supervisada tratan, por una parte, con el uso de regularizaci��n para obtener un clasificador naive Bayes y, por otra parte, con el desarrollo de un algoritmo que usa regularizaci��n por grupos de una manera eficiente y que se ha aplicado al dise��o de interfaces cerebrom��quina. Finalmente, se presenta una heur��stica para inducir la estructura de redes Bayesianas Gaussianas usando regularizaci��n L1 a modo de filtro.

Regularized model learning in EDAs for continuous and multi-objective optimization

Probabilistic modeling is the de���ning characteristic of estimation of distribution algorithms (EDAs) which determines their behavior and performance in optimization. Regularization is a well-known statistical technique used for obtaining an improved model by reducing the generalization error of estimation, especially in high-dimensional problems. `1-regularization is a type of this technique with the appealing variable selection property which results in sparse model estimations. In this thesis, we study the use of regularization techniques for model learning in EDAs. Several methods for regularized model estimation in continuous domains based on a Gaussian distribution assumption are presented, and analyzed from di���erent aspects when used for optimization in a high-dimensional setting, where the population size of EDA has a logarithmic scale with respect to the number of variables. The optimization results obtained for a number of continuous problems with an increasing number of variables show that the proposed EDA based on regularized model estimation performs a more robust optimization, and is able to achieve signi���cantly better results for larger dimensions than other Gaussian-based EDAs. We also propose a method for learning a marginally factorized Gaussian Markov random ���eld model using regularization techniques and a clustering algorithm. The experimental results show notable optimization performance on continuous additively decomposable problems when using this model estimation method. Our study also covers multi-objective optimization and we propose joint probabilistic modeling of variables and objectives in EDAs based on Bayesian networks, speci���cally models inspired from multi-dimensional Bayesian network classi���ers. It is shown that with this approach to modeling, two new types of relationships are encoded in the estimated models in addition to the variable relationships captured in other EDAs: objectivevariable and objective-objective relationships. An extensive experimental study shows the e���ectiveness of this approach for multi- and many-objective optimization. With the proposed joint variable-objective modeling, in addition to the Pareto set approximation, the algorithm is also able to obtain an estimation of the multi-objective problem structure. Finally, the study of multi-objective optimization based on joint probabilistic modeling is extended to noisy domains, where the noise in objective values is represented by intervals. A new version of the Pareto dominance relation for ordering the solutions in these problems, namely ���-degree Pareto dominance, is introduced and its properties are analyzed. We show that the ranking methods based on this dominance relation can result in competitive performance of EDAs with respect to the quality of the approximated Pareto sets. This dominance relation is then used together with a method for joint probabilistic modeling based on `1-regularization for multi-objective feature subset selection in classi���cation, where six di���erent measures of accuracy are considered as objectives with interval values. The individual assessment of the proposed joint probabilistic modeling and solution ranking methods on datasets with small-medium dimensionality, when using two di���erent Bayesian classi���ers, shows that comparable or better Pareto sets of feature subsets are approximated in comparison to standard methods.

Desarrollo de un entorno inteligente en el contexto del hogar digital

A lo largo de las ��ltimas d��cadas el desarrollo de la tecnolog��a en muy distintas ��reas ha sido vertiginoso. Su propagaci��n a todos los aspectos de nuestro d��a a d��a parece casi inevitable y la electr��nica de consumo ha invadido nuestros hogares. No obstante, parece que la dom��tica no ha alcanzado el grado de integraci��n que cab��a esperar hace apenas una d��cada. Es cierto que los dispositivos aut��nomos y con un cierto grado de inteligencia est��n abri��ndose paso de manera independiente, pero el hogar digital, como sistema capaz de abarcar y automatizar grandes conjuntos de elementos de una vivienda (gesti��n energ��tica, seguridad, bienestar, etc.) no ha conseguido extenderse al hogar medio. Esta falta de integraci��n no se debe a la ausencia de tecnolog��a, ni mucho menos, y numerosos son los estudios y proyectos surgidos en esta direcci��n. Sin embargo, no ha sido hasta hace unos pocos a��os que las instituciones y grandes compa����as han comenzado a prestar verdadero inter��s en este campo. Parece que estamos a punto de experimentar un nuevo cambio en nuestra forma de vida, concretamente en la manera en la que interactuamos con nuestro hogar y las comodidades e informaci��n que este nos puede proporcionar. En esa corriente se desarrolla este Proyecto Fin de Grado, con el objetivo de aportar un nuevo enfoque a la manera de integrar los diferentes dispositivos del hogar digital con la inteligencia artificial y, lo que es m��s importante, al modo en el que el usuario interact��a con su vivienda. M��s concretamente, se pretende desarrollar un sistema capaz de tomar decisiones acordes al contexto y a las preferencias del usuario. A trav��s de la utilizaci��n de diferentes tecnolog��as se dotar�� al hogar digital de cierta autonom��a a la hora de decidir qu�� acciones debe llevar a cabo sobre los dispositivos que contiene, todo ello mediante la interpretaci��n de ��rdenes procedentes del usuario (expresadas de manera coloquial) y el estudio del contexto que envuelve al instante de ejecuci��n. Para la interacci��n entre el usuario y el hogar digital se desarrollar�� una aplicaci��n m��vil mediante la cual podr�� expresar (de manera conversacional) las ��rdenes que quiera dar al sistema, el cual intervendr�� en la conversaci��n y llevar�� a cabo las acciones oportunas. Para todo ello, el sistema har�� principalmente uso de ontolog��as, an��lisis sem��ntico, redes bayesianas, UPnP y Android. Se combinar�� informaci��n procedente del usuario, de los sensores y de fuentes externas para determinar, a trav��s de las citadas tecnolog��as, cu��l es la operaci��n que debe realizarse para satisfacer las necesidades del usuario. En definitiva, el objetivo final de este proyecto es dise��ar e implementar un sistema innovador que se salga de la corriente actual de interacci��n mediante botones, men��s y formularios a los que estamos tan acostumbrados, y que permita al usuario, en cierto modo, hablar con su vivienda y expresarle sus necesidades, haciendo a la tecnolog��a un poco m��s transparente y cercana y aproxim��ndonos un poco m��s a ese concepto de hogar inteligente que imagin��bamos a finales del siglo XX. ABSTRACT. Over the last decades the development of technology in very different areas has happened incredibly fast. Its propagation to all aspects of our daily activities seems to be inevitable and the electronic devices have invaded our homes. Nevertheless, home automation has not reached the integration point that it was supposed to just a few decades ago. It is true that some autonomic and relatively intelligent devices are emerging, but the digital home as a system able to control a large set of elements from a house (energy management, security, welfare, etc.) is not present yet in the average home. That lack of integration is not due to the absence of technology and, in fact, there are a lot of investigations and projects focused on this field. However, the institutions and big companies have not shown enough interest in home automation until just a few years ago. It seems that, finally, we are about to experiment another change in our lifestyle and how we interact with our home and the information and facilities it can provide. This Final Degree Project is developed as part of this trend, with the goal of providing a new approach to the way the system could integrate the home devices with the artificial intelligence and, mainly, to the way the user interacts with his house. More specifically, this project aims to develop a system able to make decisions, taking into account the context and the user preferences. Through the use of several technologies and approaches, the system will be able to decide which actions it should perform based on the order interpretation (expressed colloquially) and the context analysis. A mobile application will be developed to enable the user-home interaction. The user will be able to express his orders colloquially though out a conversational mode, and the system will also participate in the conversation, performing the required actions. For providing all this features, the system will mainly use ontologies, semantic analysis, Bayesian networks, UPnP and Android. Information from the user, the sensors and external sources will be combined to determine, through the use of these technologies, which is the operation that the system should perform to meet the needs of the user. In short, the final goal of this project is to design and implement an innovative system, away from the current trend of buttons, menus and forms. In a way, the user will be able to talk to his home and express his needs, experiencing a technology closer to the people and getting a little closer to that concept of digital home that we imagined in the late twentieth century.

Conditional density approximations with mixtures of polynomials

Mixtures of polynomials (MoPs) are a non-parametric density estimation technique especially designed for hybrid Bayesian networks with continuous and discrete variables. Algorithms to learn one- and multi-dimensional (marginal) MoPs from data have recently been proposed. In this paper we introduce two methods for learning MoP approximations of conditional densities from data. Both approaches are based on learning MoP approximations of the joint density and the marginal density of the conditioning variables, but they differ as to how the MoP approximation of the quotient of the two densities is found. We illustrate and study the methods using data sampled from known parametric distributions, and we demonstrate their applicability by learning models based on real neuroscience data. Finally, we compare the performance of the proposed methods with an approach for learning mixtures of truncated basis functions (MoTBFs). The empirical results show that the proposed methods generally yield models that are comparable to or significantly better than those found using the MoTBF-based method.

Inteligencia artificial para la predicci��n y control del acabado superficial en procesos de fresado a alta velocidad

El objetivo principal de esta tesis doctoral es profundizar en el an��lisis y dise��o de un sistema inteligente para la predicci��n y control del acabado superficial en un proceso de fresado a alta velocidad, basado fundamentalmente en clasificadores Bayesianos, con el prop��osito de desarrollar una metodolog����a que facilite el dise��o de este tipo de sistemas. El sistema, cuyo prop��sito es posibilitar la predicci��n y control de la rugosidad superficial, se compone de un modelo aprendido a partir de datos experimentales con redes Bayesianas, que ayudar��a a comprender los procesos din��micos involucrados en el mecanizado y las interacciones entre las variables relevantes. Dado que las redes neuronales artificiales son modelos ampliamente utilizados en procesos de corte de materiales, tambi��n se incluye un modelo para fresado us��ndolas, donde se introdujo la geometr��a y la dureza del material como variables novedosas hasta ahora no estudiadas en este contexto. Por lo tanto, una importante contribuci��n en esta tesis son estos dos modelos para la predicci��n de la rugosidad superficial, que se comparan con respecto a diferentes aspectos: la influencia de las nuevas variables, los indicadores de evaluaci��n del desempe��o, interpretabilidad. Uno de los principales problemas en la modelizaci��n con clasificadores Bayesianos es la comprensi��n de las enormes tablas de probabilidad a posteriori producidas. Introducimos un m��etodo de explicaci��n que genera un conjunto de reglas obtenidas de ��rboles de decisi��n. Estos ��rboles son inducidos a partir de un conjunto de datos simulados generados de las probabilidades a posteriori de la variable clase, calculadas con la red Bayesiana aprendida a partir de un conjunto de datos de entrenamiento. Por ��ltimo, contribuimos en el campo multiobjetivo en el caso de que algunos de los objetivos no se puedan cuantificar en n��meros reales, sino como funciones en intervalo de valores. Esto ocurre a menudo en aplicaciones de aprendizaje autom��tico, especialmente las basadas en clasificaci��n supervisada. En concreto, se extienden las ideas de dominancia y frontera de Pareto a esta situaci��n. Su aplicaci��n a los estudios de predicci��n de la rugosidad superficial en el caso de maximizar al mismo tiempo la sensibilidad y la especificidad del clasificador inducido de la red Bayesiana, y no solo maximizar la tasa de clasificaci��n correcta. Los intervalos de estos dos objetivos provienen de un m��etodo de estimaci��n honesta de ambos objetivos, como e.g. validaci��n cruzada en k rodajas o bootstrap.---ABSTRACT---The main objective of this PhD Thesis is to go more deeply into the analysis and design of an intelligent system for surface roughness prediction and control in the end-milling machining process, based fundamentally on Bayesian network classifiers, with the aim of developing a methodology that makes easier the design of this type of systems. The system, whose purpose is to make possible the surface roughness prediction and control, consists of a model learnt from experimental data with the aid of Bayesian networks, that will help to understand the dynamic processes involved in the machining and the interactions among the relevant variables. Since artificial neural networks are models widely used in material cutting proceses, we include also an end-milling model using them, where the geometry and hardness of the piecework are introduced as novel variables not studied so far within this context. Thus, an important contribution in this thesis is these two models for surface roughness prediction, that are then compared with respecto to different aspects: influence of the new variables, performance evaluation metrics, interpretability. One of the main problems with Bayesian classifier-based modelling is the understanding of the enormous posterior probabilitiy tables produced. We introduce an explanation method that generates a set of rules obtained from decision trees. Such trees are induced from a simulated data set generated from the posterior probabilities of the class variable, calculated with the Bayesian network learned from a training data set. Finally, we contribute in the multi-objective field in the case that some of the objectives cannot be quantified as real numbers but as interval-valued functions. This often occurs in machine learning applications, especially those based on supervised classification. Specifically, the dominance and Pareto front ideas are extended to this setting. Its application to the surface roughness prediction studies the case of maximizing simultaneously the sensitivity and specificity of the induced Bayesian network classifier, rather than only maximizing the correct classification rate. Intervals in these two objectives come from a honest estimation method of both objectives, like e.g. k-fold cross-validation or bootstrap.

Caracterizaci��n y simulaci��n de arborizaciones dentr��ticas con redes bayesianas incluyendo variables angulares

El funcionamiento interno del cerebro es todav��a hoy en d��a un misterio, siendo su comprensi��n uno de los principales desaf��os a los que se enfrenta la ciencia moderna. El c��rtex cerebral es el ��rea del cerebro donde tienen lugar los procesos cerebrales de m��s alto nivel, c��mo la imaginaci��n, el juicio o el pensamiento abstracto. Las neuronas piramidales, un tipo espec��fico de neurona, suponen cerca del 80% de los cerca de los 10.000 millones de que componen el c��rtex cerebral, haciendo de ellas un objetivo principal en el estudio del funcionamiento del cerebro. La morfolog��a neuronal, y m��s espec��ficamente la morfolog��a dendr��tica, determina c��mo estas procesan la informaci��n y los patrones de conexi��n entre neuronas, siendo los modelos computacionales herramientas imprescindibles para el estudio de su rol en el funcionamiento del cerebro. En este trabajo hemos creado un modelo computacional, con m��s de 50 variables relativas a la morfolog��a dendr��tica, capaz de simular el crecimiento de arborizaciones dendr��ticas basales completas a partir de reconstrucciones de neuronas piramidales reales, abarcando desde el n��mero de dendritas hasta el crecimiento los los ��rboles dendr��ticos. A diferencia de los trabajos anteriores, nuestro modelo basado en redes Bayesianas contempla la arborizaci��n dendr��tica en su conjunto, teniendo en cuenta las interacciones entre dendritas y detectando de forma autom��tica las relaciones entre las variables morfol��gicas que caracterizan la arborizaci��n. Adem��s, el an��lisis de las redes Bayesianas puede ayudar a identificar relaciones hasta ahora desconocidas entre variables morfol��gicas. Motivado por el estudio de la orientaci��n de las dendritas basales, en este trabajo se introduce una regularizaci��n L1 generalizada, aplicada al aprendizaje de la distribuci��n von Mises multivariante, una de las principales distribuciones de probabilidad direccional multivariante. Tambi��n se propone una distancia circular multivariante que puede utilizarse para estimar la divergencia de Kullback-Leibler entre dos muestras de datos circulares. Comparamos los modelos con y sin regularizaci ��n en el estudio de la orientaci��n de la dendritas basales en neuronas humanas, comprobando que, en general, el modelo regularizado obtiene mejores resultados. El muestreo, ajuste y representaci��n de la distribuci��n von Mises multivariante se implementa en un nuevo paquete de R denominado mvCircular.---ABSTRACT---The inner workings of the brain are, as of today, a mystery. To understand the brain is one of the main challenges faced by current science. The cerebral cortex is the region of the brain where all superior brain processes, like imagination, judge and abstract reasoning take place. Pyramidal neurons, a specific type of neurons, constitute approximately the 80% of the more than 10.000 million neurons that compound the cerebral cortex. It makes the study of the pyramidal neurons crucial in order to understand how the brain works. Neuron morphology, and specifically the dendritic morphology, determines how the information is processed in the neurons, as well as the connection patterns among neurons. Computational models are one of the main tools for studying dendritic morphology and its role in the brain function. We have built a computational model that contains more than 50 morphological variables of the dendritic arborizations. This model is able to simulate the growth of complete dendritic arborizations from real neuron reconstructions, starting with the number of basal dendrites, and ending modeling the growth of dendritic trees. One of the main diferences between our approach, mainly based on the use of Bayesian networks, and other models in the state of the art is that we model the whole dendritic arborization instead of focusing on individual trees, which makes us able to take into account the interactions between dendrites and to automatically detect relationships between the morphologic variables that characterize the arborization. Moreover, the posterior analysis of the relationships in the model can help to identify new relations between morphological variables. Motivated by the study of the basal dendrites orientation, a generalized L1 regularization applied to the multivariate von Mises distribution, one of the most used distributions in multivariate directional statistics, is also introduced in this work. We also propose a circular multivariate distance that can be used to estimate the Kullback-Leibler divergence between two circular data samples. We compare the regularized and unregularized models on basal dendrites orientation of human neurons and prove that regularized model achieves better results than non regularized von Mises model. Sampling, fitting and plotting functions for the multivariate von Mises are implemented in a new R packaged called mvCircular.

Predicci��n de tr��fico de contenedores a corto plazo mediante t��cnicas de miner��a de datos : redes neuronales artificiales y redes bayesianas

El correcto pron��stico en el ��mbito de la log��stica de transportes es de vital importancia para una adecuada planificaci��n de medios y recursos, as�� como de su optimizaci��n. Hasta la fecha los estudios sobre planificaci��n portuaria se basan principalmente en modelos emp��ricos; que se han utilizado para planificar nuevas terminales y desarrollar planes directores cuando no se dispone de datos iniciales, anal��ticos; m��s relacionados con la teor��a de colas y tiempos de espera con formulaciones matem��ticas complejas y necesitando simplificaciones de las mismas para hacer manejable y pr��ctico el modelo o de simulaci��n; que requieren de una inversi��n significativa como para poder obtener resultados aceptables invirtiendo en programas y desarrollos complejos. La Miner��a de Datos (MD) es un ��rea moderna interdisciplinaria que engloba a aquellas t��cnicas que operan de forma autom��tica (requieren de la m��nima intervenci��n humana) y, adem��s, son eficientes para trabajar con las grandes cantidades de informaci��n disponible en las bases de datos de numerosos problemas pr��cticos. La aplicaci��n pr��ctica de estas disciplinas se extiende a numerosos ��mbitos comerciales y de investigaci��n en problemas de predicci��n, clasificaci��n o diagnosis. Entre las diferentes t��cnicas disponibles en miner��a de datos las redes neuronales artificiales (RNA) y las redes probabil��sticas o redes bayesianas (RB) permiten modelizar de forma conjunta toda la informaci��n relevante para un problema dado. En el presente trabajo se han analizado dos aplicaciones de estos casos al ��mbito portuario y en concreto a contenedores. En la Tesis Doctoral se desarrollan las RNA como herramienta para obtener previsiones de tr��fico y de recursos a futuro de diferentes puertos, a partir de variables de explotaci��n, obteni��ndose valores continuos. Para el caso de las redes bayesianas (RB), se realiza un trabajo similar que para el caso de las RNA, obteni��ndose valores discretos (un intervalo). El principal resultado que se obtiene es la posibilidad de utilizar tanto las RNA como las RB para la estimaci��n a futuro de par��metros f��sicos, as�� como la relaci��n entre los mismos en una terminal para una correcta asignaci��n de los medios a utilizar y por tanto aumentar la eficiencia productiva de la terminal. Como paso final se realiza un estudio de complementariedad de ambos modelos a corto plazo, donde se puede comprobar la buena aceptaci��n de los resultados obtenidos. Por tanto, se puede concluir que estos m��todos de predicci��n pueden ser de gran ayuda a la planificaci��n portuaria. The correct assets��� forecast in the field of transportation logistics is a matter of vital importance for a suitable planning and optimization of the necessary means and resources. Up to this date, ports planning studies were basically using empirical models to deal with new terminals planning or master plans development when no initial data are available; analytical models, more connected to the queuing theory and the waiting times, and very complicated mathematical formulations requiring significant simplifications to acquire a practical and easy to handle model; or simulation models, that require a significant investment in computer codes and complex developments to produce acceptable results. The Data Mining (DM) is a modern interdisciplinary field that include those techniques that operate automatically (almost no human intervention is required) and are highly efficient when dealing with practical problems characterized by huge data bases containing significant amount of information. These disciplines��� practical application extends to many commercial or research fields, dealing with forecast, classification or diagnosis problems. Among the different techniques of the Data Mining, the Artificial Neuronal Networks (ANN) and the probabilistic ��� or Bayesian ��� networks (BN) allow the joint modeling of all the relevant information for a given problem. This PhD work analyses their application to two practical cases in the ports field, concretely to container terminals. This PhD work details how the ANN have been developed as a tool to produce traffic and resources forecasts for several ports, based on exploitation variables to obtain continuous values. For the Bayesian networks case (BN), a similar development has been carried out, obtaining discreet values (an interval). The main finding is the possibility to use ANN and BN to estimate future needs of the port���s or terminal���s physical parameters, as well as the relationship between them within a specific terminal, that allow a correct assignment of the necessary means and, thus, to increase the terminal���s productive efficiency. The final step is a short term complementarily study of both models, carried out in order to verify the obtained results. It can thus be stated that these prediction methods can be a very useful tool in ports��� planning.

Fusi��n de redes Bayesianas Gaussianas

Las redes Bayesianas constituyen un modelo ampliamente utilizado para la representaci��n de relaciones de dependencia condicional en datos multivariantes. Su aprendizaje a partir de un conjunto de datos o expertos ha sido estudiado profundamente desde su concepci��n. Sin embargo, en determinados escenarios se demanda la obtenci��n de un modelo com��n asociado a particiones de datos o conjuntos de expertos. En este caso, se trata el problema de fusi��n o agregaci��n de modelos. Los trabajos y resultados en agregaci��n de redes Bayesianas son de naturaleza variada, aunque escasos en comparaci��n con aquellos de aprendizaje. En este documento, se proponen dos m��todos para la agregaci��n de redes Gaussianas, definidas como aquellas redes Bayesianas que modelan una distribuci��n Gaussiana multivariante. Los m��todos presentados son efectivos, precisos y producen redes con menor cantidad de par��metros en comparaci��n con los modelos obtenidos individualmente. Adem��s, constituyen un enfoque novedoso al incorporar nociones exploradas tradicionalmente por separado en el estado del arte. Futuras aplicaciones en entornos escalables hacen dichos m��todos especialmente atractivos, dada su simplicidad y la ganancia en compacidad de la representaci��n obtenida.---ABSTRACT---Bayesian networks are a widely used model for the representation of conditional dependence relationships among variables in multivariate data. The task of learning them from a data set or experts has been deeply studied since their conception. However, situations emerge where there is a need of obtaining a consensuated model from several data partitions or a set of experts. This situation is referred to as model fusion or aggregation. Results about Bayesian network aggregation, although rich in variety, have been scarce when compared to the learning task. In this context, two methods are proposed for the aggregation of Gaussian Bayesian networks, that is, Bayesian networks whose underlying modelled distribution is a multivariate Gaussian. Both methods are effective, precise and produce networks with fewer parameters in comparison with the models obtained by individual learning. They constitute a novel approach given that they incorporate notions traditionally explored separately in the state of the art. Future applications in scalable computer environments make such models specially attractive, given their simplicity and the gaining in sparsity of the produced model.

Metodolog��a para la evaluaci��n de la calidad de la localizaci��n de puertos secos

La estructura econ��mica mundial, con centros de producci��n y consumo descentralizados y el consiguiente aumento en el tr��fico de mercanc��as en todo el mundo, crea considerables problemas y desaf��os para el sector del transporte de mercanc��as. Esta situaci��n ha llevado al transporte mar��timo a convertirse en el modo m��s econ��mico y m��s adecuado para el transporte de mercanc��as a nivel global. De este modo, los puertos mar��timos se configuran como nodos de importancia capital en la cadena de suministro al servir como enlace entre dos sistemas de transporte, el mar��timo y el terrestre. El aumento de la actividad en los puertos mar��timos produce tres efectos indeseables: el aumento de la congesti��n vial, la falta de espacio abierto en las instalaciones portuarias y un impacto ambiental significativo en los puertos mar��timos. Los puertos secos nacen para favorecer la utilizaci��n de cada modo de transporte en los segmentos en que resultan m��s competitivos y para mitigar estos problemas moviendo parte de la actividad en el interior. Adem��s, gracias a la implantaci��n de puertos secos es posible discretizar cada uno de los eslabones de la cadena de transporte, permitiendo que los modos m��s contaminantes y con menor capacidad de transporte tengan itinerarios lo m��s cortos posible, o bien, sean utilizados ��nicamente para el transporte de mercanc��as de alto valor a��adido. As��, los puertos secos se presentan como una oportunidad para fortalecer las soluciones intermodales como parte de una cadena integrada de transporte sostenible, potenciando el transporte de mercanc��as por ferrocarril. Sin embargo, su potencial no es aprovechado al no existir una metodolog��a de planificaci��n de la ubicaci��n de uso sencillo y resultados claros para la toma de decisiones a partir de los criterios ingenieriles definidos por los t��cnicos. La decisi��n de d��nde ubicar un puerto seco exige un an��lisis exhaustivo de toda la cadena log��stica, con el objetivo de transferir el mayor volumen de tr��fico posible a los modos m��s eficientes desde el punto de vista energ��tico, que son menos perjudiciales para el medio ambiente. Sin embargo, esta decisi��n tambi��n debe garantizar la sostenibilidad de la propia localizaci��n. Esta Tesis Doctoral, pretende sentar las bases te��ricas para el desarrollo de una herramienta de Herramienta de Ayuda a la Toma de Decisiones que permita establecer la localizaci��n m��s adecuada para la construcci��n de puertos secos. Este primer paso es el desarrollo de una metodolog��a de evaluaci��n de la sostenibilidad y la calidad de las localizaciones de los puertos secos actuales mediante el uso de las siguientes t��cnicas: Metodolog��a DELPHI, Redes Bayesianas, An��lisis Multicriterio y Sistemas de Informaci��n Geogr��fica. Reconociendo que la determinaci��n de la ubicaci��n m��s adecuada para situar diversos tipos de instalaciones es un importante problema geogr��fico, con significativas repercusiones medioambientales, sociales, econ��micos, locacionales y de accesibilidad territorial, se considera un conjunto de 40 variables (agrupadas en 17 factores y estos, a su vez, en 4 criterios) que permiten evaluar la sostenibilidad de las localizaciones. El An��lisis Multicriterio se utiliza como forma de establecer una puntuaci��n a trav��s de un algoritmo de scoring. Este algoritmo se alimenta a trav��s de: 1) unas calificaciones para cada variable extra��das de informaci��n geogr��fica analizada con ArcGIS (Criteria Assessment Score); 2) los pesos de los factores obtenidos a trav��s de un cuestionario DELPHI, una t��cnica caracterizada por su capacidad para alcanzar consensos en un grupo de expertos de muy diferentes especialidades: log��stica, sostenibilidad, impacto ambiental, planificaci��n de transportes y geograf��a; y 3) los pesos de las variables, para lo que se emplean las Redes Bayesianas lo que supone una importante aportaci��n metodol��gica al tratarse de una novedosa aplicaci��n de esta t��cnica. Los pesos se obtienen aprovechando la capacidad de clasificaci��n de las Redes Bayesianas, en concreto de una red dise��ada con un algoritmo de tipo greedy denominado K2 que permite priorizar cada variable en funci��n de las relaciones que se establecen en el conjunto de variables. La principal ventaja del empleo de esta t��cnica es la reducci��n de la arbitrariedad en la fijaci��n de los pesos de la cual suelen adolecer las t��cnicas de An��lisis Multicriterio. Como caso de estudio, se eval��a la sostenibilidad de los 10 puertos secos existentes en Espa��a. Los resultados del cuestionario DELPHI revelan una mayor importancia a la hora de buscar la localizaci��n de un Puerto Seco en los aspectos tenidos en cuenta en las teor��as cl��sicas de localizaci��n industrial, principalmente econ��micos y de accesibilidad. Sin embargo, no deben perderse de vista el resto de factores, cuesti��n que se pone de manifiesto a trav��s del cuestionario, dado que ninguno de los factores tiene un peso tan peque��o como para ser despreciado. Por el contrario, los resultados de la aplicaci��n de Redes Bayesianas, muestran una mayor importancia de las variables medioambientales, por lo que la sostenibilidad de las localizaciones exige un gran respeto por el medio natural y el medio urbano en que se encuadra. Por ��ltimo, la aplicaci��n pr��ctica refleja que la localizaci��n de los puertos secos existentes en Espa��a en la actualidad presenta una calidad modesta, que parece responder m��s a decisiones pol��ticas que a criterios t��cnicos. Por ello, deben emprenderse pol��ticas encaminadas a generar un modelo log��stico colaborativo-competitivo en el que se eval��en los diferentes factores tenidos en cuenta en esta investigaci��n. The global economic structure, with its decentralized production and the consequent increase in freight traffic all over the world, creates considerable problems and challenges for the freight transport sector. This situation has led shipping to become the most suitable and cheapest way to transport goods. Thus, ports are configured as nodes with critical importance in the logistics supply chain as a link between two transport systems, sea and land. Increase in activity at seaports is producing three undesirable effects: increasing road congestion, lack of open space in port installations and a significant environmental impact on seaports. These adverse effects can be mitigated by moving part of the activity inland. Implementation of dry ports is a possible solution and would also provide an opportunity to strengthen intermodal solutions as part of an integrated and more sustainable transport chain, acting as a link between road and railway networks. In this sense, implementation of dry ports allows the separation of the links of the transport chain, thus facilitating the shortest possible routes for the lowest capacity and most polluting means of transport. Thus, the decision of where to locate a dry port demands a thorough analysis of the whole logistics supply chain, with the objective of transferring the largest volume of goods possible from road to more energy efficient means of transport, like rail or short-sea shipping, that are less harmful to the environment. However, the decision of where to locate a dry port must also ensure the sustainability of the site. Thus, the main goal of this dissertation is to research the variables influencing the sustainability of dry port location and how this sustainability can be evaluated. With this objective, in this research we present a methodology for assessing the sustainability of locations by the use of Multi-Criteria Decision Analysis (MCDA) and Bayesian Networks (BNs). MCDA is used as a way to establish a scoring, whilst BNs were chosen to eliminate arbitrariness in setting the weightings using a technique that allows us to prioritize each variable according to the relationships established in the set of variables. In order to determine the relationships between all the variables involved in the decision, giving us the importance of each factor and variable, we built a K2 BN algorithm. To obtain the scores of each variable, we used a complete cartography analysed by ArcGIS. Recognising that setting the most appropriate location to place a dry port is a geographical multidisciplinary problem, with significant economic, social and environmental implications, we consider 41 variables (grouped into 17 factors) which respond to this need. As a case of study, the sustainability of all of the 10 existing dry ports in Spain has been evaluated. In this set of logistics platforms, we found that the most important variables for achieving sustainability are those related to environmental protection, so the sustainability of the locations requires a great respect for the natural environment and the urban environment in which they are framed.

Metodolog��a para la determinaci��n de los par��metros de planificaci��n de los nodos e infraestructuras log��sticas en un territorio, considerando en su planificaci��n el impacto sobre los diferentes componentes territoriales

Esta tesis presenta el dise��o y la aplicaci��n de una metodolog��a que permite la determinaci��n de los par��metros para la planificaci��n de nodos e infraestructuras log��sticas en un territorio, considerando adem��s el impacto de estas en los diferentes componentes territoriales, as�� como en el desarrollo poblacional, el desarrollo econ��mico y el medio ambiente, presentando as�� un avance en la planificaci��n integral del territorio. La Metodolog��a propuesta est�� basada en Miner��a de Datos, que permite el descubrimiento de patrones detr��s de grandes vol��menes de datos previamente procesados. Las caracter��sticas propias de los datos sobre el territorio y los componentes que lo conforman hacen de los estudios territoriales un campo ideal para la aplicaci��n de algunas de las t��cnicas de Miner��a de Datos, tales como los ��arboles decisi��n y las redes bayesianas. Los ��rboles de decisi��n permiten representar y categorizar de forma esquem��tica una serie de variables de predicci��n que ayudan al an��lisis de una variable objetivo. Las redes bayesianas representan en un grafo ac��clico dirigido, un modelo probabil��stico de variables distribuidas en padres e hijos, y la inferencia estad��stica que permite determinar la probabilidad de certeza de una hip��tesis planteada, es decir, permiten construir modelos de probabilidad conjunta que presentan de manera gr��fica las dependencias relevantes en un conjunto de datos. Al igual que con los ��rboles de decisi��n, la divisi��n del territorio en diferentes unidades administrativas hace de las redes bayesianas una herramienta potencial para definir las caracter��sticas f��sicas de alguna tipolog��a especifica de infraestructura log��stica tomando en consideraci��n las caracter��sticas territoriales, poblacionales y econ��micas del ��rea donde se plantea su desarrollo y las posibles sinergias que se puedan presentar sobre otros nodos e infraestructuras log��sticas. El caso de estudio seleccionado para la aplicaci��n de la metodolog��a ha sido la Rep��blica de Panam��, considerando que este pa��s presenta algunas caracter��sticas singulares, entra las que destacan su alta concentraci��n de poblaci��n en la Ciudad de Panam��; que a su vez a concentrado la actividad econ��mica del pa��s; su alto porcentaje de zonas protegidas, lo que ha limitado la vertebraci��n del territorio; y el Canal de Panam�� y los puertos de contenedores adyacentes al mismo. La metodolog��a se divide en tres fases principales: Fase 1: Determinaci��n del escenario de trabajo 1. Revisi��n del estado del arte. 2. Determinaci��n y obtenci��n de las variables de estudio. Fase 2: Desarrollo del modelo de inteligencia artificial 3. Construcci��n de los ��arboles de decisi��n. 4. Construcci��n de las redes bayesianas. Fase 3: Conclusiones 5. Determinaci��n de las conclusiones. Con relaci��n al modelo de planificaci��n aplicado al caso de estudio, una vez aplicada la metodolog��a, se estableci�� un modelo compuesto por 47 variables que definen la planificaci��n log��stica de Panam��, el resto de variables se definen a partir de estas, es decir, conocidas estas, el resto se definen a trav��s de ellas. Este modelo de planificaci��n establecido a trav��s de la red bayesiana considera los aspectos de una planificaci��n sostenible: econ��mica, social y ambiental; que crean sinergia con la planificaci��n de nodos e infraestructuras log��sticas. The thesis presents the design and application of a methodology that allows the determination of parameters for the planning of nodes and logistics infrastructure in a territory, besides considering the impact of these different territorial components, as well as the population growth, economic and environmental development. The proposed methodology is based on Data Mining, which allows the discovery of patterns behind large volumes of previously processed data. The own characteristics of the territorial data makes of territorial studies an ideal field of knowledge for the implementation of some of the Data Mining techniques, such as Decision Trees and Bayesian Networks. Decision trees categorize schematically a series of predictor variables of an analyzed objective variable. Bayesian Networks represent a directed acyclic graph, a probabilistic model of variables divided in fathers and sons, and statistical inference that allow determine the probability of certainty in a hypothesis. The case of study for the application of the methodology is the Republic of Panama. This country has some unique features: a high population density in the Panama City, a concentration of economic activity, a high percentage of protected areas, and the Panama Canal. The methodology is divided into three main phases: Phase 1: definition of the work stage. 1. Review of the State of the art. 2. Determination of the variables. Phase 2: Development of artificial intelligence model 3. Construction of decision trees. 4. Construction of Bayesian Networks. Phase 3: conclusions 5. Determination of the conclusions. The application of the methodology to the case study established a model composed of 47 variables that define the logistics planning for Panama. This model of planning established through the Bayesian network considers aspects of sustainable planning and simulates the synergies between the nodes and logistical infrastructure planning.