Efficient learning of decomposable models with a bounded clique size

The learning of probability distributions from data is a ubiquitous problem in the fields of Statistics and Artificial Intelligence. During the last decades several learning algorithms have been proposed to learn probability distributions based on decomposable models due to their advantageous theoretical properties. Some of these algorithms can be used to search for a maximum likelihood decomposable model with a given maximum clique size, k, which controls the complexity of the model. Unfortunately, the problem of learning a maximum likelihood decomposable model given a maximum clique size is NP-hard for k > 2. In this work, we propose a family of algorithms which approximates this problem with a computational complexity of O(k · n^2 log n) in the worst case, where n is the number of implied random variables. The structures of the decomposable models that solve the maximum likelihood problem are called maximal k-order decomposable graphs. Our proposals, called fractal trees, construct a sequence of maximal i-order decomposable graphs, for i = 2, ..., k, in k − 1 steps. At each step, the algorithms follow a divide-and-conquer strategy based on the particular features of this type of structures. Additionally, we propose a prune-and-graft procedure which transforms a maximal k-order decomposable graph into another one, increasing its likelihood. We have implemented two particular fractal tree algorithms called parallel fractal tree and sequential fractal tree. These algorithms can be considered a natural extension of Chow and Liu’s algorithm, from k = 2 to arbitrary values of k. Both algorithms have been compared against other efficient approaches in artificial and real domains, and they have shown a competitive behavior to deal with the maximum likelihood problem. Due to their low computational complexity they are especially recommended to deal with high dimensional domains.

Towards a Comprehensive Approach to Quantifying and Mapping Ecosystem Services Flows

25 p.

Efficient electronic implementations of adaptive systems for ambient intelligence environments

183 p.

Frame-level features conveying phonetic information for language and speaker recognition

150 p.

Un electrocardiógrafo inteligente de bajo coste

Castellano: A lo largo de este proyecto se ha desarrollado un sistema de bajo coste para la tomade electrocardiogramas y posterior visualización de los mismos en un dispositivo Android. Además se ha creado un módulo inteligente capaz de realizar un diagnóstico de manera automática y razonada sobre los datos recogidos. El proyecto se ha realizado principalmente sobre tecnologías abiertas: Arduino como componente central del sistema electrónico, Android para visualizar datos en una plataforma móvil y CLIPS como motor sobre el cual se ha desarrollado el sistema experto que realiza el diagnóstico.

ANÁLISIS DEL EFECTO DEL CAMBIO DE LA DISTRIBUCIÓN DE CLASES MEDIANTE MÉTODOS DE REMUESTREO INTELIGENTE EN ÁRBOLES DE CLASIFICACIÓN SIMPLES Y CONSOLIDADOS

En los trabajos expuestos en esta memoria de tesis, hemos analizado elefecto que tienen sobre la capacidad de aprendizaje de diferentes algoritmosde clasificación los cambios en la distribución de clases, teniendo encuenta para ello, diferentes métodos de remuestreo de datos.En concreto se ha analizado este efecto en el conocido algoritmo deconstrucción de árboles de clasificación propuesto por Quinlan, el algoritmoC4.5, y en el algoritmo de construcción de árboles consolidados, elalgoritmo CTC, propuesto por el grupo de investigación ALDAPA de laUniversidad del País Vasco que, basado en el mismo C4.5, obtiene un árbol declasificación pero basado en un conjunto de muestras.Así mismo, planteamos cómo encontrar la distribución de clases más adecuadapara un algoritmo de clasificación y método de remuestreo concretos.

Instances of combinatorial optimization problems: complexity and generation

138 p.

Contributions on distance-based algorithms, multi-classifier construction and pairwise classification

179 p.

Contributions to the efficient use of general purpose coprocessors: kernel density estimation as case study

142 p.

Combination of web usage, content and structure information for diverse web mining applications in the tourism context and the context of users with disabilities

188 p.

Contributions to learning Bayesian network models from weakly supervised data: Application to Assisted Reproductive Technologies and Software Defect Classification

162 p.

Contributions to time series data mining departing from the problem of road travel time modeling

194 p.