Probabilistic reasoning with a bayesian DNA device based on strand displacement

We present a computing model based on the DNA strand displacement technique which performs Bayesian inference. The model will take single stranded DNA as input data, representing the presence or absence of a specific molecular signal (evidence). The program logic encodes the prior probability of a disease and the conditional probability of a signal given the disease playing with a set of different DNA complexes and their ratios. When the input and program molecules interact, they release a different pair of single stranded DNA species whose relative proportion represents the application of Bayes? Law: the conditional probability of the disease given the signal. The models presented in this paper can empower the application of probabilistic reasoning in genetic diagnosis in vitro.

B2DI a bayesian BDI agent model with causal belief updating based on MSBN

In this paper, we introduce B2DI model that extends BDI model to perform Bayesian inference under uncertainty. For scalability and flexibility purposes, Multiply Sectioned Bayesian Network (MSBN) technology has been selected and adapted to BDI agent reasoning. A belief update mechanism has been defined for agents, whose belief models are connected by public shared beliefs, and the certainty of these beliefs is updated based on MSBN. The classical BDI agent architecture has been extended in order to manage uncertainty using Bayesian reasoning. The resulting extended model, so-called B2DI, proposes a new control loop. The proposed B2DI model has been evaluated in a network fault diagnosis scenario. The evaluation has compared this model with two previously developed agent models. The evaluation has been carried out with a real testbed diagnosis scenario using JADEX. As a result, the proposed model exhibits significant improvements in the cost and time required to carry out a reliable diagnosis.

Distributed Fault Diagnosis Using Bayesian Reasoning in MAGNETO

Many of the emerging telecom services make use of Outer Edge Networks, in particular Home Area Networks. The configuration and maintenance of such services may not be under full control of the telecom operator which still needs to guarantee the service quality experienced by the consumer. Diagnosing service faults in these scenarios becomes especially difficult since there may be not full visibility between different domains. This paper describes the fault diagnosis solution developed in the MAGNETO project, based on the application of Bayesian Inference to deal with the uncertainty. It also takes advantage of a distributed framework to deploy diagnosis components in the different domains and network elements involved, spanning both the telecom operator and the Outer Edge networks. In addition, MAGNETO features self-learning capabilities to automatically improve diagnosis knowledge over time and a partition mechanism that allows breaking down the overall diagnosis knowledge into smaller subsets. The MAGNETO solution has been prototyped and adapted to a particular outer edge scenario, and has been further validated on a real testbed. Evaluation of the results shows the potential of our approach to deal with fault management of outer edge networks.

A multi-agent system with distributed bayesian reasoning for network fault diagnosis

In this paper, an innovative approach to perform distributed Bayesian inference using a multi-agent architecture is presented. The final goal is dealing with uncertainty in network diagnosis, but the solution can be of applied in other fields. The validation testbed has been a P2P streaming video service. An assessment of the work is presented, in order to show its advantages when it is compared with traditional manual processes and other previous systems.

Bayesian model selection of structural explanatory models: Application to road accident data

Using the Bayesian approach as the model selection criteria, the main purpose in this study is to establish a practical road accident model that can provide a better interpretation and prediction performance. For this purpose we are using a structural explanatory model with autoregressive error term. The model estimation is carried out through Bayesian inference and the best model is selected based on the goodness of fit measures. To cross validate the model estimation further prediction analysis were done. As the road safety measures the number of fatal accidents in Spain, during 2000-2011 were employed. The results of the variable selection process show that the factors explaining fatal road accidents are mainly exposure, economic factors, and surveillance and legislative measures. The model selection shows that the impact of economic factors on fatal accidents during the period under study has been higher compared to surveillance and legislative measures.

Efficient random variable generation: ratio of uniforms and polar rejection sampling

Monte Carlo techniques, which require the generation of samples from some target density, are often the only alternative for performing Bayesian inference. Two classic sampling techniques to draw independent samples are the ratio of uniforms (RoU) and rejection sampling (RS). An efficient sampling algorithm is proposed combining the RoU and polar RS (i.e. RS inside a sector of a circle using polar coordinates). Its efficiency is shown in drawing samples from truncated Cauchy and Gaussian random variables, which have many important applications in signal processing and communications. RESUMEN. Método eficiente para generar algunas variables aleatorias de uso común en procesado de señal y comunicaciones (por ejemplo, Gaussianas o Cauchy truncadas) mediante la combinación de dos técnicas: "ratio of uniforms" y "rejection sampling".

Probabilistic reasoning with an enzyme-driven DNA device

We present a biomolecular probabilistic model driven by the action of a DNA toolbox made of a set of DNA templates and enzymes that is able to perform Bayesian inference. The model will take single-stranded DNA as input data, representing the presence or absence of a specific molecular signal (the evidence). The program logic uses different DNA templates and their relative concentration ratios to encode the prior probability of a disease and the conditional probability of a signal given the disease. When the input and program molecules interact, an enzyme-driven cascade of reactions (DNA polymerase extension, nicking and degradation) is triggered, producing a different pair of single-stranded DNA species. Once the system reaches equilibrium, the ratio between the output species will represent the application of Bayes? law: the conditional probability of the disease given the signal. In other words, a qualitative diagnosis plus a quantitative degree of belief in that diagno- sis. Thanks to the inherent amplification capability of this DNA toolbox, the resulting system will be able to to scale up (with longer cascades and thus more input signals) a Bayesian biosensor that we designed previously.

An experimental and numerical study of ductile failure under quasi-static and impact loadings of Inconel 718 nickel-base superalloy

A numerical and experimental study of ballistic impacts at various temperatures on precipitation hardened Inconel 718 nickel-base superalloy plates has been performed. A coupled elastoplastic-damage constitutive model with Lode angle dependent failure criterion has been implemented in LS-DYNA non-linear finite element code to model the mechanical behaviour of such an alloy. The ballistic impact tests have been carried out at three temperatures: room temperature (25 °C), 400 °C and 700 °C. The numerical study showed that the mesh size is crucial to predict correctly the shear bands detected in the tested plates. Moreover, the mesh size convergence has been achieved for element sizes on the same order that the shear bands. The residual velocity as well as the ballistic limit prediction has been considered excellent for high temperature ballistic tests. Nevertheless, the model has been less accurate for the numerical simulations performed at room temperature, being though in reasonable agreement with the experimental data. Additionally, the influence that the Lode angle had on quasi-static failure patterns such as cup-cone and slanted failure has been studied numerically. The study has revealed that the combined action of weakened constitutive equations and Lode angle dependent failure criterion has been necessary to predict the previously-mentioned failure patterns

Generalización de variables medioambientales mediante interpolación GIS

La investigación de esta tesis se centra en el estudio de técnicas geoestadísticas y su contribución a una mayor caracterización del binomio factores climáticos-rendimiento de un cultivo agrícola. El inexorable vínculo entre la variabilidad climática y la producción agrícola cobra especial relevancia en estudios sobre el cambio climático o en la modelización de cultivos para dar respuesta a escenarios futuros de producción mundial. Es información especialmente valiosa en sistemas operacionales de monitoreo y predicción de rendimientos de cultivos Los cuales son actualmente uno de los pilares operacionales en los que se sustenta la agricultura y seguridad alimentaria mundial; ya que su objetivo final es el de proporcionar información imparcial y fiable para la regularización de mercados. Es en este contexto, donde se quiso dar un enfoque alternativo a estudios, que con distintos planteamientos, analizan la relación inter-anual clima vs producción. Así, se sustituyó la dimensión tiempo por la espacio, re-orientando el análisis estadístico de correlación interanual entre rendimiento y factores climáticos, por el estudio de la correlación inter-regional entre ambas variables. Se utilizó para ello una técnica estadística relativamente nueva y no muy aplicada en investigaciones similares, llamada regresión ponderada geográficamente (GWR, siglas en inglés de “Geographically weighted regression”). Se obtuvieron superficies continuas de las variables climáticas acumuladas en determinados periodos fenológicos, que fueron seleccionados por ser factores clave en el desarrollo vegetativo de un cultivo. Por ello, la primera parte de la tesis, consistió en un análisis exploratorio sobre comparación de Métodos de Interpolación Espacial (MIE). Partiendo de la hipótesis de que existe la variabilidad espacial de la relación entre factores climáticos y rendimiento, el objetivo principal de esta tesis, fue el de establecer en qué medida los MIE y otros métodos geoestadísticos de regresión local, pueden ayudar por un lado, a alcanzar un mayor entendimiento del binomio clima-rendimiento del trigo blando (Triticum aestivum L.) al incorporar en dicha relación el componente espacial; y por otro, a caracterizar la variación de los principales factores climáticos limitantes en el crecimiento del trigo blando, acumulados éstos en cuatro periodos fenológicos. Para lleva a cabo esto, una gran carga operacional en la investigación de la tesis consistió en homogeneizar y hacer los datos fenológicos, climáticos y estadísticas agrícolas comparables tanto a escala espacial como a escala temporal. Para España y los Bálticos se recolectaron y calcularon datos diarios de precipitación, temperatura máxima y mínima, evapotranspiración y radiación solar en las estaciones meteorológicas disponibles. Se dispuso de una serie temporal que coincidía con los mismos años recolectados en las estadísticas agrícolas, es decir, 14 años contados desde 2000 a 2013 (hasta 2011 en los Bálticos). Se superpuso la malla de información fenológica de cuadrícula 25 km con la ubicación de las estaciones meteorológicas con el fin de conocer los valores fenológicos en cada una de las estaciones disponibles. Hecho esto, para cada año de la serie temporal disponible se calcularon los valores climáticos diarios acumulados en cada uno de los cuatro periodos fenológicos seleccionados P1 (ciclo completo), P2 (emergencia-madurez), P3 (floración) y P4 (floraciónmadurez). Se calculó la superficie interpolada por el conjunto de métodos seleccionados en la comparación: técnicas deterministas convencionales, kriging ordinario y cokriging ordinario ponderado por la altitud. Seleccionados los métodos más eficaces, se calculó a nivel de provincias las variables climatológicas interpoladas. Y se realizaron las regresiones locales GWR para cuantificar, explorar y modelar las relaciones espaciales entre el rendimiento del trigo y las variables climáticas acumuladas en los cuatro periodos fenológicos. Al comparar la eficiencia de los MIE no destaca una técnica por encima del resto como la que proporcione el menor error en su predicción. Ahora bien, considerando los tres indicadores de calidad de los MIE estudiados se han identificado los métodos más efectivos. En el caso de la precipitación, es la técnica geoestadística cokriging la más idónea en la mayoría de los casos. De manera unánime, la interpolación determinista en función radial (spline regularizado) fue la técnica que mejor describía la superficie de precipitación acumulada en los cuatro periodos fenológicos. Los resultados son más heterogéneos para la evapotranspiración y radiación. Los métodos idóneos para estas se reparten entre el Inverse Distance Weighting (IDW), IDW ponderado por la altitud y el Ordinary Kriging (OK). También, se identificó que para la mayoría de los casos en que el error del Ordinary CoKriging (COK) era mayor que el del OK su eficacia es comparable a la del OK en términos de error y el requerimiento computacional de este último es mucho menor. Se pudo confirmar que existe la variabilidad espacial inter-regional entre factores climáticos y el rendimiento del trigo blando tanto en España como en los Bálticos. La herramienta estadística GWR fue capaz de reproducir esta variabilidad con un rendimiento lo suficientemente significativo como para considerarla una herramienta válida en futuros estudios. No obstante, se identificaron ciertas limitaciones en la misma respecto a la información que devuelve el programa a nivel local y que no permite desgranar todo el detalle sobre la ejecución del mismo. Los indicadores y periodos fenológicos que mejor pudieron reproducir la variabilidad espacial del rendimiento en España y Bálticos, arrojaron aún, una mayor credibilidad a los resultados obtenidos y a la eficacia del GWR, ya que estaban en línea con el conocimiento agronómico sobre el cultivo del trigo blando en sistemas agrícolas mediterráneos y norteuropeos. Así, en España, el indicador más robusto fue el balance climático hídrico Climatic Water Balance) acumulado éste, durante el periodo de crecimiento (entre la emergencia y madurez). Aunque se identificó la etapa clave de la floración como el periodo en el que las variables climáticas acumuladas proporcionaban un mayor poder explicativo del modelo GWR. Sin embargo, en los Bálticos, países donde el principal factor limitante en su agricultura es el bajo número de días de crecimiento efectivo, el indicador más efectivo fue la radiación acumulada a lo largo de todo el ciclo de crecimiento (entre la emergencia y madurez). Para el trigo en regadío no existe ninguna combinación que pueda explicar más allá del 30% de la variación del rendimiento en España. Poder demostrar que existe un comportamiento heterogéneo en la relación inter-regional entre el rendimiento y principales variables climáticas, podría contribuir a uno de los mayores desafíos a los que se enfrentan, a día de hoy, los sistemas operacionales de monitoreo y predicción de rendimientos de cultivos, y éste es el de poder reducir la escala espacial de predicción, de un nivel nacional a otro regional. ABSTRACT This thesis explores geostatistical techniques and their contribution to a better characterization of the relationship between climate factors and agricultural crop yields. The crucial link between climate variability and crop production plays a key role in climate change research as well as in crops modelling towards the future global production scenarios. This information is particularly important for monitoring and forecasting operational crop systems. These geostatistical techniques are currently one of the most fundamental operational systems on which global agriculture and food security rely on; with the final aim of providing neutral and reliable information for food market controls, thus avoiding financial speculation of nourishments of primary necessity. Within this context the present thesis aims to provide an alternative approach to the existing body of research examining the relationship between inter-annual climate and production. Therefore, the temporal dimension was replaced for the spatial dimension, re-orienting the statistical analysis of the inter-annual relationship between crops yields and climate factors to an inter-regional correlation between these two variables. Geographically weighted regression, which is a relatively new statistical technique and which has rarely been used in previous research on this topic was used in the current study. Continuous surface values of the climate accumulated variables in specific phenological periods were obtained. These specific periods were selected because they are key factors in the development of vegetative crop. Therefore, the first part of this thesis presents an exploratory analysis regarding the comparability of spatial interpolation methods (SIM) among diverse SIMs and alternative geostatistical methodologies. Given the premise that spatial variability of the relationship between climate factors and crop production exists, the primary aim of this thesis was to examine the extent to which the SIM and other geostatistical methods of local regression (which are integrated tools of the GIS software) are useful in relating crop production and climate variables. The usefulness of these methods was examined in two ways; on one hand the way this information could help to achieve higher production of the white wheat binomial (Triticum aestivum L.) by incorporating the spatial component in the examination of the above-mentioned relationship. On the other hand, the way it helps with the characterization of the key limiting climate factors of soft wheat growth which were analysed in four phenological periods. To achieve this aim, an important operational workload of this thesis consisted in the homogenization and obtention of comparable phenological and climate data, as well as agricultural statistics, which made heavy operational demands. For Spain and the Baltic countries, data on precipitation, maximum and minimum temperature, evapotranspiration and solar radiation from the available meteorological stations were gathered and calculated. A temporal serial approach was taken. These temporal series aligned with the years that agriculture statistics had previously gathered, these being 14 years from 2000 to 2013 (until 2011 for the Baltic countries). This temporal series was mapped with a phenological 25 km grid that had the location of the meteorological stations with the objective of obtaining the phenological values in each of the available stations. Following this procedure, the daily accumulated climate values for each of the four selected phenological periods were calculated; namely P1 (complete cycle), P2 (emergency-maturity), P3 (flowering) and P4 (flowering- maturity). The interpolated surface was then calculated using the set of selected methodologies for the comparison: deterministic conventional techniques, ordinary kriging and ordinary cokriging weighted by height. Once the most effective methods had been selected, the level of the interpolated climate variables was calculated. Local GWR regressions were calculated to quantify, examine and model the spatial relationships between soft wheat production and the accumulated variables in each of the four selected phenological periods. Results from the comparison among the SIMs revealed that no particular technique seems more favourable in terms of accuracy of prediction. However, when the three quality indicators of the compared SIMs are considered, some methodologies appeared to be more efficient than others. Regarding precipitation results, cokriging was the most accurate geostatistical technique for the majority of the cases. Deterministic interpolation in its radial function (controlled spline) was the most accurate technique for describing the accumulated precipitation surface in all phenological periods. However, results are more heterogeneous for the evapotranspiration and radiation methodologies. The most appropriate technique for these forecasts are the Inverse Distance Weighting (IDW), weighted IDW by height and the Ordinary Kriging (OK). Furthermore, it was found that for the majority of the cases where the Ordinary CoKriging (COK) error was larger than that of the OK, its efficacy was comparable to that of the OK in terms of error while the computational demands of the latter was much lower. The existing spatial inter-regional variability between climate factors and soft wheat production was confirmed for both Spain and the Baltic countries. The GWR statistic tool reproduced this variability with an outcome significative enough as to be considered a valid tool for future studies. Nevertheless, this tool also had some limitations with regards to the information delivered by the programme because it did not allow for a detailed break-down of its procedure. The indicators and phenological periods that best reproduced the spatial variability of yields in Spain and the Baltic countries made the results and the efficiency of the GWR statistical tool even more reliable, despite the fact that these were already aligned with the agricultural knowledge about soft wheat crop under mediterranean and northeuropean agricultural systems. Thus, for Spain, the most robust indicator was the Climatic Water Balance outcome accumulated throughout the growing period (between emergency and maturity). Although the flowering period was the phase that best explained the accumulated climate variables in the GWR model. For the Baltic countries where the main limiting agricultural factor is the number of days of effective growth, the most effective indicator was the accumulated radiation throughout the entire growing cycle (between emergency and maturity). For the irrigated soft wheat there was no combination capable of explaining above the 30% of variation of the production in Spain. The fact that the pattern of the inter-regional relationship between the crop production and key climate variables is heterogeneous within a country could contribute to one is one of the greatest challenges that the monitoring and forecasting operational systems for crop production face nowadays. The present findings suggest that the solution may lay in downscaling the spatial target scale from a national to a regional level.

Dense Map Inference with User-Defined Priors: From Priorlets to Scan Eigenvariations

When mapping is formulated in a Bayesian framework, the need of specifying a prior for the environment arises naturally. However, so far, the use of a particular structure prior has been coupled to working with a particular representation. We describe a system that supports inference with multiple priors while keeping the same dense representation. The priors are rigorously described by the user in a domain-specific language. Even though we work very close to the measurement space, we are able to represent structure constraints with the same expressivity as methods based on geometric primitives. This approach allows the intrinsic degrees of freedom of the environment’s shape to be recovered. Experiments with simulated and real data sets will be presented

Automated attribute inference in complex service workflows based on sharing analysis

The properties of data and activities in business processes can be used to greatly facilítate several relevant tasks performed at design- and run-time, such as fragmentation, compliance checking, or top-down design. Business processes are often described using workflows. We present an approach for mechanically inferring business domain-specific attributes of workflow components (including data Ítems, activities, and elements of sub-workflows), taking as starting point known attributes of workflow inputs and the structure of the workflow. We achieve this by modeling these components as concepts and applying sharing analysis to a Horn clause-based representation of the workflow. The analysis is applicable to workflows featuring complex control and data dependencies, embedded control constructs, such as loops and branches, and embedded component services.

Adaptive neuro-fuzzy inference systems (ANFIS) for supervised classification of competence in MIS based on psychomotor skills

INTRODUCTION: Objective assessment of motor skills has become an important challenge in minimally invasive surgery (MIS) training.Currently, there is no gold standard defining and determining the residents' surgical competence.To aid in the decision process, we analyze the validity of a supervised classifier to determine the degree of MIS competence based on assessment of psychomotor skills METHODOLOGY: The ANFIS is trained to classify performance in a box trainer peg transfer task performed by two groups (expert/non expert). There were 42 participants included in the study: the non-expert group consisted of 16 medical students and 8 residents (< 10 MIS procedures performed), whereas the expert group consisted of 14 residents (> 10 MIS procedures performed) and 4 experienced surgeons. Instrument movements were captured by means of the Endoscopic Video Analysis (EVA) tracking system. Nine motion analysis parameters (MAPs) were analyzed, including time, path length, depth, average speed, average acceleration, economy of area, economy of volume, idle time and motion smoothness. Data reduction was performed by means of principal component analysis, and then used to train the ANFIS net. Performance was measured by leave one out cross validation. RESULTS: The ANFIS presented an accuracy of 80.95%, where 13 experts and 21 non-experts were correctly classified. Total root mean square error was 0.88, while the area under the classifiers' ROC curve (AUC) was measured at 0.81. DISCUSSION: We have shown the usefulness of ANFIS for classification of MIS competence in a simple box trainer exercise. The main advantage of using ANFIS resides in its continuous output, which allows fine discrimination of surgical competence. There are, however, challenges that must be taken into account when considering use of ANFIS (e.g. training time, architecture modeling). Despite this, we have shown discriminative power of ANFIS for a low-difficulty box trainer task, regardless of the individual significances between MAPs. Future studies are required to confirm the findings, inclusion of new tasks, conditions and sample population.

Contributions to Bayesian network learning with applications to neuroscience

Neuronal morphology is a key feature in the study of brain circuits, as it is highly related to information processing and functional identification. Neuronal morphology affects the process of integration of inputs from other neurons and determines the neurons which receive the output of the neurons. Different parts of the neurons can operate semi-independently according to the spatial location of the synaptic connections. As a result, there is considerable interest in the analysis of the microanatomy of nervous cells since it constitutes an excellent tool for better understanding cortical function. However, the morphologies, molecular features and electrophysiological properties of neuronal cells are extremely variable. Except for some special cases, this variability makes it hard to find a set of features that unambiguously define a neuronal type. In addition, there are distinct types of neurons in particular regions of the brain. This morphological variability makes the analysis and modeling of neuronal morphology a challenge. Uncertainty is a key feature in many complex real-world problems. Probability theory provides a framework for modeling and reasoning with uncertainty. Probabilistic graphical models combine statistical theory and graph theory to provide a tool for managing domains with uncertainty. In particular, we focus on Bayesian networks, the most commonly used probabilistic graphical model. In this dissertation, we design new methods for learning Bayesian networks and apply them to the problem of modeling and analyzing morphological data from neurons. The morphology of a neuron can be quantified using a number of measurements, e.g., the length of the dendrites and the axon, the number of bifurcations, the direction of the dendrites and the axon, etc. These measurements can be modeled as discrete or continuous data. The continuous data can be linear (e.g., the length or the width of a dendrite) or directional (e.g., the direction of the axon). These data may follow complex probability distributions and may not fit any known parametric distribution. Modeling this kind of problems using hybrid Bayesian networks with discrete, linear and directional variables poses a number of challenges regarding learning from data, inference, etc. In this dissertation, we propose a method for modeling and simulating basal dendritic trees from pyramidal neurons using Bayesian networks to capture the interactions between the variables in the problem domain. A complete set of variables is measured from the dendrites, and a learning algorithm is applied to find the structure and estimate the parameters of the probability distributions included in the Bayesian networks. Then, a simulation algorithm is used to build the virtual dendrites by sampling values from the Bayesian networks, and a thorough evaluation is performed to show the model’s ability to generate realistic dendrites. In this first approach, the variables are discretized so that discrete Bayesian networks can be learned and simulated. Then, we address the problem of learning hybrid Bayesian networks with different kinds of variables. Mixtures of polynomials have been proposed as a way of representing probability densities in hybrid Bayesian networks. We present a method for learning mixtures of polynomials approximations of one-dimensional, multidimensional and conditional probability densities from data. The method is based on basis spline interpolation, where a density is approximated as a linear combination of basis splines. The proposed algorithms are evaluated using artificial datasets. We also use the proposed methods as a non-parametric density estimation technique in Bayesian network classifiers. Next, we address the problem of including directional data in Bayesian networks. These data have some special properties that rule out the use of classical statistics. Therefore, different distributions and statistics, such as the univariate von Mises and the multivariate von Mises–Fisher distributions, should be used to deal with this kind of information. In particular, we extend the naive Bayes classifier to the case where the conditional probability distributions of the predictive variables given the class follow either of these distributions. We consider the simple scenario, where only directional predictive variables are used, and the hybrid case, where discrete, Gaussian and directional distributions are mixed. The classifier decision functions and their decision surfaces are studied at length. Artificial examples are used to illustrate the behavior of the classifiers. The proposed classifiers are empirically evaluated over real datasets. We also study the problem of interneuron classification. An extensive group of experts is asked to classify a set of neurons according to their most prominent anatomical features. A web application is developed to retrieve the experts’ classifications. We compute agreement measures to analyze the consensus between the experts when classifying the neurons. Using Bayesian networks and clustering algorithms on the resulting data, we investigate the suitability of the anatomical terms and neuron types commonly used in the literature. Additionally, we apply supervised learning approaches to automatically classify interneurons using the values of their morphological measurements. Then, a methodology for building a model which captures the opinions of all the experts is presented. First, one Bayesian network is learned for each expert, and we propose 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 is induced. Finally, a consensus Bayesian multinet which models the opinions of the whole group of experts is built. A thorough analysis of the consensus model identifies different behaviors between the experts when classifying the interneurons in the experiment. A set of characterizing morphological traits for the neuronal types can be defined by performing inference in the Bayesian multinet. These findings are used to validate the model and to gain some insights into neuron morphology. Finally, we study a classification problem where the true class label of the training instances is not known. Instead, a set of class labels is available for each instance. This is inspired by the neuron classification problem, where a group of experts is asked to individually provide a class label for each instance. We propose a novel approach for learning Bayesian networks using count vectors which represent the number of experts who selected each class label for each instance. These Bayesian networks are evaluated using artificial datasets from supervised learning problems. Resumen La morfología neuronal es una característica clave en el estudio de los circuitos cerebrales, ya que está altamente relacionada con el procesado de información y con los roles funcionales. La morfología neuronal afecta al proceso de integración de las señales de entrada y determina las neuronas que reciben las salidas de otras neuronas. Las diferentes partes de la neurona pueden operar de forma semi-independiente de acuerdo a la localización espacial de las conexiones sinápticas. Por tanto, existe un interés considerable en el análisis de la microanatomía de las células nerviosas, ya que constituye una excelente herramienta para comprender mejor el funcionamiento de la corteza cerebral. Sin embargo, las propiedades morfológicas, moleculares y electrofisiológicas de las células neuronales son extremadamente variables. Excepto en algunos casos especiales, esta variabilidad morfológica dificulta la definición de un conjunto de características que distingan claramente un tipo neuronal. Además, existen diferentes tipos de neuronas en regiones particulares del cerebro. La variabilidad neuronal hace que el análisis y el modelado de la morfología neuronal sean un importante reto científico. La incertidumbre es una propiedad clave en muchos problemas reales. La teoría de la probabilidad proporciona un marco para modelar y razonar bajo incertidumbre. Los modelos gráficos probabilísticos combinan la teoría estadística y la teoría de grafos con el objetivo de proporcionar una herramienta con la que trabajar bajo incertidumbre. En particular, nos centraremos en las redes bayesianas, el modelo más utilizado dentro de los modelos gráficos probabilísticos. En esta tesis hemos diseñado nuevos métodos para aprender redes bayesianas, inspirados por y aplicados al problema del modelado y análisis de datos morfológicos de neuronas. La morfología de una neurona puede ser cuantificada usando una serie de medidas, por ejemplo, la longitud de las dendritas y el axón, el número de bifurcaciones, la dirección de las dendritas y el axón, etc. Estas medidas pueden ser modeladas como datos continuos o discretos. A su vez, los datos continuos pueden ser lineales (por ejemplo, la longitud o la anchura de una dendrita) o direccionales (por ejemplo, la dirección del axón). Estos datos pueden llegar a seguir distribuciones de probabilidad muy complejas y pueden no ajustarse a ninguna distribución paramétrica conocida. El modelado de este tipo de problemas con redes bayesianas híbridas incluyendo variables discretas, lineales y direccionales presenta una serie de retos en relación al aprendizaje a partir de datos, la inferencia, etc. En esta tesis se propone un método para modelar y simular árboles dendríticos basales de neuronas piramidales usando redes bayesianas para capturar las interacciones entre las variables del problema. Para ello, se mide un amplio conjunto de variables de las dendritas y se aplica un algoritmo de aprendizaje con el que se aprende la estructura y se estiman los parámetros de las distribuciones de probabilidad que constituyen las redes bayesianas. Después, se usa un algoritmo de simulación para construir dendritas virtuales mediante el muestreo de valores de las redes bayesianas. Finalmente, se lleva a cabo una profunda evaluaci ón para verificar la capacidad del modelo a la hora de generar dendritas realistas. En esta primera aproximación, las variables fueron discretizadas para poder aprender y muestrear las redes bayesianas. A continuación, se aborda el problema del aprendizaje de redes bayesianas con diferentes tipos de variables. Las mixturas de polinomios constituyen un método para representar densidades de probabilidad en redes bayesianas híbridas. Presentamos un método para aprender aproximaciones de densidades unidimensionales, multidimensionales y condicionales a partir de datos utilizando mixturas de polinomios. El método se basa en interpolación con splines, que aproxima una densidad como una combinación lineal de splines. Los algoritmos propuestos se evalúan utilizando bases de datos artificiales. Además, las mixturas de polinomios son utilizadas como un método no paramétrico de estimación de densidades para clasificadores basados en redes bayesianas. Después, se estudia el problema de incluir información direccional en redes bayesianas. Este tipo de datos presenta una serie de características especiales que impiden el uso de las técnicas estadísticas clásicas. Por ello, para manejar este tipo de información se deben usar estadísticos y distribuciones de probabilidad específicos, como la distribución univariante von Mises y la distribución multivariante von Mises–Fisher. En concreto, en esta tesis extendemos el clasificador naive Bayes al caso en el que las distribuciones de probabilidad condicionada de las variables predictoras dada la clase siguen alguna de estas distribuciones. Se estudia el caso base, en el que sólo se utilizan variables direccionales, y el caso híbrido, en el que variables discretas, lineales y direccionales aparecen mezcladas. También se estudian los clasificadores desde un punto de vista teórico, derivando sus funciones de decisión y las superficies de decisión asociadas. El comportamiento de los clasificadores se ilustra utilizando bases de datos artificiales. Además, los clasificadores son evaluados empíricamente utilizando bases de datos reales. También se estudia el problema de la clasificación de interneuronas. Desarrollamos una aplicación web que permite a un grupo de expertos clasificar un conjunto de neuronas de acuerdo a sus características morfológicas más destacadas. Se utilizan medidas de concordancia para analizar el consenso entre los expertos a la hora de clasificar las neuronas. Se investiga la idoneidad de los términos anatómicos y de los tipos neuronales utilizados frecuentemente en la literatura a través del análisis de redes bayesianas y la aplicación de algoritmos de clustering. Además, se aplican técnicas de aprendizaje supervisado con el objetivo de clasificar de forma automática las interneuronas a partir de sus valores morfológicos. A continuación, se presenta una metodología para construir un modelo que captura las opiniones de todos los expertos. Primero, se genera una red bayesiana para cada experto y se propone un algoritmo para agrupar las redes bayesianas que se corresponden con expertos con comportamientos similares. Después, se induce una red bayesiana que modela la opinión de cada grupo de expertos. Por último, se construye una multired bayesiana que modela las opiniones del conjunto completo de expertos. El análisis del modelo consensuado permite identificar diferentes comportamientos entre los expertos a la hora de clasificar las neuronas. Además, permite extraer un conjunto de características morfológicas relevantes para cada uno de los tipos neuronales mediante inferencia con la multired bayesiana. Estos descubrimientos se utilizan para validar el modelo y constituyen información relevante acerca de la morfología neuronal. Por último, se estudia un problema de clasificación en el que la etiqueta de clase de los datos de entrenamiento es incierta. En cambio, disponemos de un conjunto de etiquetas para cada instancia. Este problema está inspirado en el problema de la clasificación de neuronas, en el que un grupo de expertos proporciona una etiqueta de clase para cada instancia de manera individual. Se propone un método para aprender redes bayesianas utilizando vectores de cuentas, que representan el número de expertos que seleccionan cada etiqueta de clase para cada instancia. Estas redes bayesianas se evalúan utilizando bases de datos artificiales de problemas de aprendizaje supervisado.

Learning Bayesian networks from data by the incremental compilation of new network polynomials

Probabilistic graphical models are a huge research field in artificial intelligence nowadays. The scope of this work is the study of directed graphical models for the representation of discrete distributions. Two of the main research topics related to this area focus on performing inference over graphical models and on learning graphical models from data. Traditionally, the inference process and the learning process have been treated separately, but given that the learned models structure marks the inference complexity, this kind of strategies will sometimes produce very inefficient models. With the purpose of learning thinner models, in this master thesis we propose a new model for the representation of network polynomials, which we call polynomial trees. Polynomial trees are a complementary representation for Bayesian networks that allows an efficient evaluation of the inference complexity and provides a framework for exact inference. We also propose a set of methods for the incremental compilation of polynomial trees and an algorithm for learning polynomial trees from data using a greedy score+search method that includes the inference complexity as a penalization in the scoring function.

Expressive power of binary relevance and chain classifiers based on Bayesian Networks for multi-label classification

Bayesian network classifiers are widely used in machine learning because they intuitively represent causal relations. Multi-label classification problems require each instance to be assigned a subset of a defined set of h labels. This problem is equivalent to finding a multi-valued decision function that predicts a vector of h binary classes. In this paper we obtain the decision boundaries of two widely used Bayesian network approaches for building multi-label classifiers: Multi-label Bayesian network classifiers built using the binary relevance method and Bayesian network chain classifiers. We extend our previous single-label results to multi-label chain classifiers, and we prove that, as expected, chain classifiers provide a more expressive model than the binary relevance method.