Combining Wireless Sensor Networks and Machine Learning for Flash Flood Nowcasting

This paper addresses an investigation with machine learning (ML) classification techniques to assist in the problem of flash flood now casting. We have been attempting to build a Wireless Sensor Network (WSN) to collect measurements from a river located in an urban area. The machine learning classification methods were investigated with the aim of allowing flash flood now casting, which in turn allows the WSN to give alerts to the local population. We have evaluated several types of ML taking account of the different now casting stages (i.e. Number of future time steps to forecast). We have also evaluated different data representation to be used as input of the ML techniques. The results show that different data representation can lead to results significantly better for different stages of now casting.

Learning an L1-regularized Gaussian Bayesian Network in the Equivalence Class Space

Learning the structure of a graphical model from data is a common task in a wide range of practical applications. In this paper, we focus on Gaussian Bayesian networks, i.e., on continuous data and directed acyclic graphs with a joint probability density of all variables given by a Gaussian. We propose to work in an equivalence class search space, specifically using the k-greedy equivalence search algorithm. This, combined with regularization techniques to guide the structure search, can learn sparse networks close to the one that generated the data. We provide results on some synthetic networks and on modeling the gene network of the two biological pathways regulating the biosynthesis of isoprenoids for the Arabidopsis thaliana plant

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¨ıve 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 motivación de la regularización. Podemos entender la regularización como una modificación del estimador de máxima verosimilitud, de tal manera que se pueda dar una respuesta cuando la configuración del problema es inestable. A modo de ejemplo, podemos mencionar el ajuste de modelos paramétricos o no paramétricos cuando hay más parámetros que casos en el conjunto de datos, o la estimación de grandes matrices de covarianzas. Se suele recurrir a la regularización, además, para mejorar el compromiso sesgo-varianza en una estimación. Por tanto, la definición de regularización es muy general y, aunque la introducción de una función de penalización es probablemente el método más popular, éste es sólo uno de entre varias posibilidades. En esta tesis se ha trabajado en aplicaciones de regularización para obtener representaciones dispersas, donde sólo se usa un subconjunto de las entradas. En particular, la regularización L1 juega un papel clave en la búsqueda de dicha dispersión. La mayor parte de las contribuciones presentadas en la tesis giran alrededor de la regularización L1, aunque también se exploran otras formas de regularización (que igualmente persiguen un modelo disperso). Además de presentar una revisión de la regularización L1 y sus aplicaciones en estadística y aprendizaje de máquina, se ha desarrollado metodología para regresión, clasificación supervisada y aprendizaje de estructura en modelos gráficos. Dentro de la regresión, se ha trabajado principalmente en métodos de regresión local, proponiendo técnicas de diseño del kernel que sean adecuadas a configuraciones de alta dimensionalidad y funciones de regresión dispersas. También se presenta una aplicación de las técnicas de regresión regularizada para modelar la respuesta de neuronas reales. Los avances en clasificación supervisada tratan, por una parte, con el uso de regularización para obtener un clasificador naive Bayes y, por otra parte, con el desarrollo de un algoritmo que usa regularización por grupos de una manera eficiente y que se ha aplicado al diseño de interfaces cerebromáquina. Finalmente, se presenta una heurística para inducir la estructura de redes Bayesianas Gaussianas usando regularización L1 a modo de filtro.

Performance of artificial neural networks and genetical evolved artificial neural networks unfolding techniques

With the Bonner spheres spectrometer neutron spectrum is obtained through an unfolding procedure. Monte Carlo methods, Regularization, Parametrization, Least-squares, and Maximum Entropy are some of the techniques utilized for unfolding. In the last decade methods based on Artificial Intelligence Technology have been used. Approaches based on Genetic Algorithms and Artificial Neural Networks have been developed in order to overcome the drawbacks of previous techniques. Nevertheless the advantages of Artificial Neural Networks still it has some drawbacks mainly in the design process of the network, vg the optimum selection of the architectural and learning ANN parameters. In recent years the use of hybrid technologies, combining Artificial Neural Networks and Genetic Algorithms, has been utilized to. In this work, several ANN topologies were trained and tested using Artificial Neural Networks and Genetically Evolved Artificial Neural Networks in the aim to unfold neutron spectra using the count rates of a Bonner sphere spectrometer. Here, a comparative study of both procedures has been carried out.

Improving web learning through model optimization using bootstrap for a tour-guide robot

We perform a review of Web Mining techniques and we describe a Bootstrap Statistics methodology applied to pattern model classifier optimization and verification for Supervised Learning for Tour-Guide Robot knowledge repository management. It is virtually impossible to test thoroughly Web Page Classifiers and many other Internet Applications with pure empirical data, due to the need for human intervention to generate training sets and test sets. We propose using the computer-based Bootstrap paradigm to design a test environment where they are checked with better reliability.

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.

Tool wear monitoring using neuro-fuzzy techniques: a comparative study in a turning process

Tool wear detection is a key issue for tool condition monitoring. The maximization of useful tool life is frequently related with the optimization of machining processes. This paper presents two model-based approaches for tool wear monitoring on the basis of neuro-fuzzy techniques. The use of a neuro-fuzzy hybridization to design a tool wear monitoring system is aiming at exploiting the synergy of neural networks and fuzzy logic, by combining human reasoning with learning and connectionist structure. The turning process that is a well-known machining process is selected for this case study. A four-input (i.e., time, cutting forces, vibrations and acoustic emissions signals) single-output (tool wear rate) model is designed and implemented on the basis of three neuro-fuzzy approaches (inductive, transductive and evolving neuro-fuzzy systems). The tool wear model is then used for monitoring the turning process. The comparative study demonstrates that the transductive neuro-fuzzy model provides better error-based performance indices for detecting tool wear than the inductive neuro-fuzzy model and than the evolving neuro-fuzzy model.

PUE attack detection in CWSNs using anomaly detection techniques

Cognitive wireless sensor network (CWSN) is a new paradigm, integrating cognitive features in traditional wireless sensor networks (WSNs) to mitigate important problems such as spectrum occupancy. Security in cognitive wireless sensor networks is an important problem since these kinds of networks manage critical applications and data. The specific constraints of WSN make the problem even more critical, and effective solutions have not yet been implemented. Primary user emulation (PUE) attack is the most studied specific attack deriving from new cognitive features. This work discusses a new approach, based on anomaly behavior detection and collaboration, to detect the primary user emulation attack in CWSN scenarios. Two non-parametric algorithms, suitable for low-resource networks like CWSNs, have been used in this work: the cumulative sum and data clustering algorithms. The comparison is based on some characteristics such as detection delay, learning time, scalability, resources, and scenario dependency. The algorithms have been tested using a cognitive simulator that provides important results in this area. Both algorithms have shown to be valid in order to detect PUE attacks, reaching a detection rate of 99% and less than 1% of false positives using collaboration.

On-line learning of a fuzzy controller for a precise vehicle cruise control system

Usually, vehicle applications require the use of artificial intelligent techniques to implement control methods, due to noise provided by sensors or the impossibility of full knowledge about dynamics of the vehicle (engine state, wheel pressure or occupiers weight). This work presents a method to on-line evolve a fuzzy controller for commanding vehicles? pedals at low speeds; in this scenario, the slightest alteration in the vehicle or road conditions can vary controller?s behavior in a non predictable way. The proposal adapts singletons positions in real time, and trapezoids used to codify the input variables are modified according with historical data. Experimentation in both simulated and real vehicles are provided to show how fast and precise the method is, even compared with a human driver or using different vehicles.

Analysis of pattern recognition and dimensionality reduction techniques for odor biometrics

In this paper, we analyze the performance of several well-known pattern recognition and dimensionality reduction techniques when applied to mass-spectrometry data for odor biometric identification. Motivated by the successful results of previous works capturing the odor from other parts of the body, this work attempts to evaluate the feasibility of identifying people by the odor emanated from the hands. By formulating this task according to a machine learning scheme, the problem is identified with a small-sample-size supervised classification problem in which the input data is formed by mass spectrograms from the hand odor of 13 subjects captured in different sessions. The high dimensionality of the data makes it necessary to apply feature selection and extraction techniques together with a simple classifier in order to improve the generalization capabilities of the model. Our experimental results achieve recognition rates over 85% which reveals that there exists discriminatory information in the hand odor and points at body odor as a promising biometric identifier.

Diseño del curso para operador "Coeficientes inherentes de reactividad" a plataforma e-Learning

Las centrales nucleares necesitan de personal altamente especializado y formado. Es por ello por lo que el sector de la formación especializada en centrales nucleares necesita incorporar los últimos avances en métodos formativos. Existe una gran cantidad de cursos de formación presenciales y es necesario transformar dichos cursos para utilizarlos con las nuevas tecnologías de la información. Para ello se necesitan equipos multidisciplinares, en los que se incluyen ingenieros, que deben identificar los objetivos formativos, competencias, contenidos y el control de calidad del propio curso. En este proyecto se utilizan técnicas de ingeniería del conocimiento como eje metodológico para transformar un curso de formación presencial en formación on-line a través de tecnologías de la información. En la actualidad, las nuevas tecnologías de la información y comunicación están en constante evolución. De esta forma se han sumergido en el mundo transformando la visión que teníamos de éste para dar lugar a nuevas oportunidades. Es por ello que este proyecto busca la unión entre el e-learning y el mundo empresarial. El objetivo es el diseño, en plataforma e-learning, de un curso técnico que instruya a operadores de sala de control de una central nuclear. El trabajo realizado en este proyecto ha sido, además de transformar un curso presencial en on-line, en obtener una metodología para que otros cursos se puedan transformar. Para conseguir este cometido, debemos preocuparnos tanto por el contenido de los cursos como por su gestión. Por este motivo, el proyecto comienza con definiciones básicas de terminología propia de e-learning. Continúa con la generación de una metodología que aplique la gestión de conocimiento para transformar cualquier curso presencial a esta plataforma. Definida la metodología, se aplicará para el diseño del curso específico de Coeficientes Inherentes de Reactividad. Finaliza con un estudio económico que dé viabilidad al proyecto y con la creación de un modelo económico que estime el precio para cualquier curso futuro. Abstract Nuclear power plants need highly specialized and trained personnel. Thus, nuclear power plant Specialized Training Sector requires the incorporation of the latest advances in training methods. A large array of face-to-face training courses exist and it has become necessary to transform said courses in order to apply them with the new information systems available. For this, multidisciplinary equipment is needed where the engineering workforce must identify educational objectives, competences and abilities, contents and quality control of the different courses. In this project, knowledge engineering techniques are employed as the methodological axis in order to transform a face-to-face training course into on-line training through the use of new information technologies. Nowadays, new information and communication technologies are in constant evolution. They have introduced themselves into our world, transforming our previous vision of them, leading to new opportunities. For this reason, the present Project seeks to unite the use of e-learning and the Business and Corporate world. The main objective is the design, in an e-learning platform, of a technical course that will train nuclear power plant control-room operators. The work carried out in this Project has been, in addition to the transformation of a face-to-face course into an online one, the obtainment of a methodology to employ in the future transformation of other courses. In order to achieve this mission, our interest must focus on the content as well as on the management of the various courses. Hence, the Project starts with basic definitions of e-learning terminology. Next, a methodology that applies knowledge management for the transformation of any face-to-face course into e-learning has been generated. Once this methodology is defined, it has been applied for the design process of the Inherent Coefficients of Reactivity course. Finally, an economic study has been developed in order to determine the viability of the Project and an economic model has been created to estimate the price of any given course

Advanced morphometric techniques applied to the study of human brain anatomy

El desarrollo de las técnicas de imágenes por resonancia magnética han permitido el estudio y cuantificación, in vivo, de los cambios que ocurren en la morfología cerebral ligados a procesos tales como el neurodesarrollo, el envejecimiento, el aprendizaje o la enfermedad. Un gran número de métodos de morfometría han sido desarrollados con el fin de extraer la información contenida en estas imágenes y traducirla en indicadores de forma o tamaño, tales como el volumen o el grosor cortical; marcadores que son posteriormente empleados para encontrar diferencias estadísticas entre poblaciones de sujetos o realizar correlaciones entre la morfología cerebral y, por ejemplo, la edad o la severidad de determinada enfermedad. A pesar de la amplia variedad de biomarcadores y metodologías de morfometría, muchos estudios sesgan sus hipótesis, y con ello los resultados experimentales, al empleo de un número reducido de biomarcadores o a al uso de una única metodología de procesamiento. Con el presente trabajo se pretende demostrar la importancia del empleo de diversos métodos de morfometría para lograr una mejor caracterización del proceso que se desea estudiar. En el mismo se emplea el análisis de forma para detectar diferencias, tanto globales como locales, en la morfología del tálamo entre pacientes adolescentes con episodios tempranos de psicosis y adolescentes sanos. Los resultados obtenidos demuestran que la diferencia de volumen talámico entre ambas poblaciones de sujetos, previamente descrita en la literatura, se debe a una reducción del volumen de la región anterior-mediodorsal y del núcleo pulvinar del tálamo de los pacientes respecto a los sujetos sanos. Además, se describe el desarrollo de un estudio longitudinal, en sujetos sanos, que emplea simultáneamente distintos biomarcadores para la caracterización y cuantificación de los cambios que ocurren en la morfología de la corteza cerebral durante la adolescencia. A través de este estudio se revela que el proceso de “alisado” que experimenta la corteza cerebral durante la adolescencia es consecuencia de una disminución de la profundidad, ligada a un incremento en el ancho, de los surcos corticales. Finalmente, esta metodología es aplicada, en un diseño transversal, para el estudio de las causas que provocan el decrecimiento tanto del grosor cortical como del índice de girificación en adolescentes con episodios tempranos de psicosis. ABSTRACT The ever evolving sophistication of magnetic resonance image techniques continue to provide new tools to characterize and quantify, in vivo, brain morphologic changes related to neurodevelopment, senescence, learning or disease. The majority of morphometric methods extract shape or size descriptors such as volume, surface area, and cortical thickness from the MRI image. These morphological measurements are commonly entered in statistical analytic approaches for testing between-group differences or for correlations between the morphological measurement and other variables such as age, sex, or disease severity. A wide variety of morphological biomarkers are reported in the literature. Despite this wide range of potentially useful biomarkers and available morphometric methods, the hypotheses and findings of the grand majority of morphological studies are biased because reports assess only one morphometric feature and usually use only one image processing method. Throughout this dissertation biomarkers and image processing strategies are combined to provide innovative and useful morphometric tools for examining brain changes during neurodevelopment. Specifically, a shape analysis technique allowing for a fine-grained assessment of regional thalamic volume in early-onset psychosis patients and healthy comparison subjects is implemented. Results show that disease-related reductions in global thalamic volume, as previously described by other authors, could be particularly driven by a deficit in the anterior-mediodorsal and pulvinar thalamic regions in patients relative to healthy subjects. Furthermore, in healthy adolescents different cortical features are extracted and combined and their interdependency is assessed over time. This study attempts to extend current knowledge of normal brain development, specifically the largely unexplored relationship between changes of distinct cortical morphological measurements during adolescence. This study demonstrates that cortical flattening, present during adolescence, is produced by a combination of age-related increase in sulcal width and decrease in sulcal depth. Finally, this methodology is applied to a cross-sectional study, investigating the mechanisms underlying the decrease in cortical thickness and gyrification observed in psychotic patients with a disease onset during adolescence.

Methods and Techniques for Segmentation of Consumers in Social Media

Los medios sociales han revolucionado la manera en la que los consumidores se relacionan entre sí y con las marcas. Las opiniones publicadas en dichos medios tienen un poder de influencia en las decisiones de compra tan importante como las campañas de publicidad. En consecuencia, los profesionales del marketing cada vez dedican mayores esfuerzos e inversión a la obtención de indicadores que permitan medir el estado de salud de las marcas a partir de los contenidos digitales generados por sus consumidores. Dada la naturaleza no estructurada de los contenidos publicados en los medios sociales, la tecnología usada para procesar dichos contenidos ha menudo implementa técnicas de Inteligencia Artificial, tales como algoritmos de procesamiento de lenguaje natural, aprendizaje automático y análisis semántico. Esta tesis, contribuye al estado de la cuestión, con un modelo que permite estructurar e integrar la información publicada en medios sociales, y una serie de técnicas cuyos objetivos son la identificación de consumidores, así como la segmentación psicográfica y sociodemográfica de los mismos. La técnica de identificación de consumidores se basa en la huella digital de los dispositivos que utilizan para navegar por la Web y es tolerante a los cambios que se producen con frecuencia en dicha huella digital. Las técnicas de segmentación psicográfica descritas obtienen la posición en el embudo de compra de los consumidores y permiten clasificar las opiniones en función de una serie de atributos de marketing. Finalmente, las técnicas de segmentación sociodemográfica permiten obtener el lugar de residencia y el género de los consumidores. ABSTRACT Social media has revolutionised the way in which consumers relate to each other and with brands. The opinions published in social media have a power of influencing purchase decisions as important as advertising campaigns. Consequently, marketers are increasing efforts and investments for obtaining indicators to measure brand health from the digital content generated by consumers. Given the unstructured nature of social media contents, the technology used for processing such contents often implements Artificial Intelligence techniques, such as natural language processing, machine learning and semantic analysis algorithms. This thesis contributes to the State of the Art, with a model for structuring and integrating the information posted on social media, and a number of techniques whose objectives are the identification of consumers, as well as their socio-demographic and psychographic segmentation. The consumer identification technique is based on the fingerprint of the devices they use to surf the Web and is tolerant to the changes that occur frequently in such fingerprint. The psychographic profiling techniques described infer the position of consumer in the purchase funnel, and allow to classify the opinions based on a series of marketing attributes. Finally, the socio-demographic profiling techniques allow to obtain the residence and gender of consumers.

Distance Education and the Evolution of Online Learning in the United States

Online education is no longer a trend, rather it is mainstream. In the Fall of 2012, 69.1% of chief academic leaders indicated online learning was critical to their long-term strategy and of the 20.6 million students enrolled in higher education, 6.7 million were enrolled in an online course (Allen & Seaman, 2013; United States Department of Education, 2013). The advent of online education and its rapid growth has forced academic institutions and faculty to question the current styles and techniques for teaching and learning. As developments in educational technology continue to advance, the ways in which we deliver and receive knowledge in both the traditional and online classrooms will further evolve. It is necessary to investigate and understand the progression and advancements in educational technology and the variety of methods used to deliver knowledge to improve the quality of education we provide today and motivate, inspire, and educate the students of the 21st century. This paper explores the atioevolution of distance education beginning with correspondence and the use of parcel post, to radio, then to television, and finally to online education.

An exercises editor for an e-learning environment

One of the most important tenets of e-learning is that it bridges work and learning. A great e-learning experience brings learning into the work environment. This is a key point, the capacity to construct a work environment when the student can develop proper tasks to complete the learning process. This paper describes a work environment based on the development of two tools, an exercises editor and an exercises viewer. Both tools are able to manage color images where, because of the implementation of basic steganographic techniques, it is possible to add information, exercises, questions, and so on. The exercises editor allows to decide which information must be visible or remain hidden to the user, when the image is loaded in the exercises viewer. Therefore, it is possible to hide the solutions of the proposed tasks; this is very useful to complete a self-evaluation learning process. These tools constitute a learning architecture with the final objective that learners can apply and practice new concepts or skills.