CIG Activities

Meetings


Santiago Gil Begué

Propuesta de un nuevo clasificador multi-dimensional de redes Bayesianas en árbol.

CIG seminar. November 15th, 2018, 12:00-13:00. Hemiciclo H-1002, Bloque 1.

Los clasificadores multi-dimensionales de redes Bayesianas (conocidos por sus siglas en inglés, MBCs) son modelos gráficos probabilísticos hechos a medida para resolver problemas de clasificación multi-dimensional, en los que una instancia se debe asignar a múltiples variables clase. En este trabajo proponemos un nuevo clasificador multi-dimensional, el cual consiste en un árbol de clasificación con MBCs en los nodos hoja. También presentamos una aproximación wrapper para aprender este clasificador desde un conjunto de datos. Un estudio experimental llevado a cabo sobre datos sintéticos generados de manera aleatoria muestra resultados alentadores en términos de precisión predictiva.



Esteban Puerto Santana

Asymmetric hidden Markov models with continuous variables.

CIG seminar. November 8th, 2018, 12:00-13:00. Hemiciclo H-1002, Bloque 1.

Hidden Markov models have been successfully applied to model signals and dynamic data. However, when dealing with many variables,traditional hidden Markov models do not take into account asymmetric dependencies, leading to models with overfitting and poor problem insight. To deal with the previous problem, asymmetric hidden Markov models were recently proposed, whose emission probabilities are modified to follow a state-dependent graphical model. However, only discrete models have been developed. We introduce asymmetric hidden Markov models with continuous variables using state-dependent linear Gaussian Bayesian networks. We propose a parameter and structure learning algorithm for this new model. We run experiments with real data from bearing vibration. Since vibrational data is continuous, with the proposed model we can avoid any variable discretization step and perform learning and inference in an asymmetric information frame.



Fernando Rodríguez Sánchez

Multidimensional clustering with Bayesian networks

CIG seminar. October 25th, 2018

El objetivo de esta charla es presentar el clustering probabilistico multidimensional como alternativa al enfoque tradicional. Si bien los metodos tradicionales de clustering asumen la premisa de que existe una forma única de agrupar las instancias, ésta no suele
cumplirse cuando trabajamos con datos complejos (alta dimensionalidad, atributos pertenecientes a varios dominios, etc.). Nuestro objetivo es por tanto presentar las carencias de este tipo de metodos, el trabajo realizado actualmente y como podemos mejorarlo. Todo ello en el marco de las redes Bayesianas.



Irene Córdoba Sánchez

A partial orthogonalization method for generating covariance and concentration graph matrices

CIG seminar. October 11th, 2018

Structure learning methods for covariance and concentration graphs are often validated on synthetic models, usually obtained by randomly generating: (i) an undirected graph, and (ii) a compatible symmetric positive definite (SPD) matrix. In order to ensure positive definiteness in (ii), a dominant diagonal is usually imposed. However, the link strengths in the resulting graphical model, determined by off-diagonal entries in the SPD matrix, are in many scenarios extremely weak. Recovering the structure of the undirected graph thus becomes a challenge, and algorithm validation is notably affected. In this paper, we propose an alternative method which overcomes such problem yet yielding a compatible SPD matrix. We generate a partially row-wise-orthogonal matrix factor, where pairwise orthogonal rows correspond to missing edges in the undirected graph. In numerical experiments ranging from moderately dense to sparse scenarios, we obtain that, as the dimension increases, the link strength we simulate is stable with respect to the structure sparsity. Importantly, we show in a real validation setting how structure recovery is greatly improved for all learning algorithms when using our proposed method, thereby producing a more realistic comparison framework.



Juan Antonio Fernandez del Pozo

Magerit

CIG Meeting. September 15th, 2014



Bojan Mihaljevic

Multi-expert multi-dimensional classification of GABAergic interneurons with label Bayesian networks

CIG Meeting. July 14th, 2014



Gherardo Varando

Decision Boundary for Discrete Bayesian Network Classifier

CIG Meeting. May 26th, 2014



Laura Antón-Sánchez

Modeling replicated 3D spatial point patterns of cerebral cortex synapses

CIG Meeting. May 19th, 2014



Gherardo Varando

Conditional Density Approximations with Mixtures of Polynomials

CIG Meeting. May 12th, 2014



Bojan Mihaljevic

Classifying GABAergic interneurons with semi-supervised projected model-based clustering

CIG Meeting. March 17th, 2014



Juan Antonio Fernández del Pozo

Representación de datos Multidimensionales: descripción de tablas de reglas y conjuntos de datos

CIG Meeting. May 17th, 2013



Laura Antón-Sanchez

Optimal Neuronal Wiring through EDAs with Permutations Domains

CIG Meeting. April 25th, 2013



Luis Guerra

Clustering of dendritic spines on Prezi

CIG Meeting. April 11th, 2013



Rubén Armañanzas

Advices on Job Seeking

CIG Meeting. April 4th, 2013



Pedro L. López-Cruz

How to build an R package

CIG Meeting. March 21st, 2013



Hossein Karshenas

Multi-objective feature subset selection with EDAs

CIG Meeting. March 14th, 2013



Bojan Mihaljevic

BayesClass: an R package for learning Bayesian network classifiers

CIG Meeting. March 7th, 2013.



Juan A. Fernández del Pozo (2011)

Parameter Control of Genetic Algorithms by Learning and Simulation of Bayesian Networks

CIG Meeting. 9th June, 2011



L. Guerra (2011)

Partially labelled data: classification and discovery of unknown labels using subspaces of features

CIG Meeting. February, 2011



Linda C. van der Gaar (2010)

When in Doubt ... Be Indecisive

CIG Meeting. 11th November, 2010



D. Vidaurre (2010)

L1-Regularization for supervised learning data

CIG Meeting. 25th September, 2010