banner

Site menu:

Principal >> Curso 2007-2008 >> Conferencias

Conferencias del Curso Académico 2007-2008

 

Data Streams: conceptos y técnicas

Día Hora Lugar Imparte
5 de junio de 2008 11:00 Salón de Grados de la ETSIIT

José Riquelme, Dpto. LSI, Universidad de Sevilla

Los data streams (expresión inglesa equivalente a datos en flujo continuo) representan un nuevo reto para las técnicas tradicionales de Minería de Datos. En la actualidad existen múltiples sistemas que están produciendo continuamente datos con gran celeridad: mercados económicos, servicios telefónicos, pagos con tarjetas, redes de sensores, etc. Extraer un conocimiento útil de este tipo de datos implica principalmente dos limitaciones: los datos no se pueden guardar, es decir, sólo podrán ser procesados una vez y el sistema debe tener un modelo aprendido en todo momento. La conferencia estará dividida en tres partes. En la primera, se introducirán los conceptos claves en el campo de los data streams, definiendo el problema, cuáles son los objetivos que se pretende con su análisis y ejemplos reales de aplicación. En la segunda parte se hará un recorrido por las principales técnicas existentes en la literatura para tratar los data streams. Así se repasarán las metodologías más difundidas para la obtención de consultas (query), reglas de asociación, clustering o clasificación en este entorno. Para finalizar, se expondrá una técnica propia del grupo de la Universidad de Sevilla para obtener un modelo de clasificación para data streams con atributos continuos. Este modelo utiliza un sistema de clasificación híbrida usando reglas y vecinos más cercanos.

Intelligent robots and Mecahtronic Systems

Día Hora Lugar Imparte
8 de febrero de 2008 11:45 Salón de Actos de la ETSIIT Toshio Fukuda
Professor of Dept. of Micro System Engineering and Dept. of Mechano-Informatics and Systems, Nagoya University, Japan
 

There are so many growing demands on automatic systems using robotic and mechatronic systems to assist us in our daily life. Most of them require the autonomous and/or semi-autonomous systems using computational intelligence for designing mechanical structure, control system and human interface. This lecture will provide the principles and methods for those design problems with the demonstration of the trajectory generation and control methods for autonomous mobile robots, brachiation robots, multiple robot coordination and human interface.

Logical and Statistical AI: A Unified View

Día Hora Lugar Imparte
8 de febrero de 2008 10:00 Salón de Actos de la ETSIIT Pedro Domingos
Associate Professor of Computer Science and Engineering at the University of Washington
 
Intelligent agents must be able to handle the complexity and uncertainty of the real world. Logical AI has focused mainly on the former, and statistical AI on the latter. In this talk, I will present a unified view of the two, including logical inference, probabilistic inference, statistical learning, and inductive logic programming. In recent years, a number of representations that combine logic and probability have been proposed. As an example, I will use Markov logic, which attaches weights to first-order formulas and views them as templates for constructing Markov networks. The algorithms covered are available in the open-source Alchemy package. I will illustrate the concepts with applications to information extraction, robot mapping, social network analysis, computational biology, and others, and conclude with a discussion of open problems and exciting research directions.

Rule based fuzzy models and identification

Día Hora Lugar Imparte
7 de febrero de 2008 11:45 Salón de Actos de la ETSIIT L. T. Koczy
Professor and Dean of Engineering at the Szechenyi Istvan University in Gyor (Hungary)
Fuzzy rule based models have been as far the far most wide spread practical applications of fuzzy sets/logic, mainly applied for control and reasoning/decision support. The evolution of rule based models started from Zadeh's Compositional Rule of Inference, and continued with Mamdani's projection based practical approach which later had many variations, such as Larsen's method, etc. There is the alternative Takagi-Sugeno approach as well.

The speaker's group had further developed the models towards sparse models with applying a multitude of rule interpolation techniques, Sugeno's group introduced the hierarchical structured rule base approach and later the speaker and colleagues combined the two into interpolative sparse models. In the talk all these models will be referred to.

The first successful I/O data based automatic rule base identification method was published also by Sugeno (with Yasukawa); this was based on Bezdek's Fuzzy c-mean clustering algorithm. The talk will go through several alternative approaches done by the speaker's group, partly continuing and extending the ideas of this latter approach, for interpolative hierarchical fuzzy models (with Chong and Gedeon), partly using evolutionary (bacterial) algorithm, traditional second order gradient approach (Levenberg-Marquardt) and the combination of these two (called bacterial memetic algorithm, these latter with Botzheim, Ruano, Cabrita and recently with Gal). All these will be described in some detail and some examples will be given, constructed, real life and benchmark data sets used for demonstration.

Expressing preferences in flexible queries: An overview of fuzzy set and possibility theory - based methods

Día Hora Lugar Imparte
7 de febrero de 2008 10:00 Salón de Actos de la ETSIIT Henri Prade
"Directeur de Recherche" at C.N.R.S., and works at IRIT (Institut de Recherche en Informatique de Toulouse)
Expressing preferences may be done in a variety of formats, such as soft constraints, conditional preferences, prioritized goals, examples of satisfactory cases The talk will provide a structured overview of different lines of research dealing with the handling of preferences in database queries, but also in semi-structured data or in documents retrieval. It includes the handling of fuzzy queries whose components may be conditionally weighted in terms of importance, the representation of preferences in a possibilistic logic manner using symbolic weights (which leaves the freedom for the user to specify or not priorities among preferences), database 'Best' operator-like methods and the CP-net approach developed in artificial intelligence.

The talk will also emphasize the interest of distinguishing between negative preferences expressing rejection, and positive preferences stating that some options are really satisfactory if they are available, thus leaving room for alternatives to which the agent is indifferent. This bipolar nature of preferences can be accommodated in the framework of possibility theory.

Soft Computing: Perspectivas de Investigación en España

Día Hora Lugar Imparte
21 de noviembre de 2007 17:00 Salón de Grados de la ETSIIT Pedro Larrañaga Múgica
Gestor del Plan Nacional de I+D+i del Ministerio de Educación y Ciencia en el área de las Tecnologías Informáticas, y Catedrático de Universidad en el área de Ciencias de la Computación e Inteligencia Artificial en la Universidad del País Vasco

 

Máster: Soft Computing y Sistemas Inteligentes
Mapa Web
sugerencias
Dpto. Ciencias de la Computación e Inteligencia Artificial
E.T.S. Ingeniería Informática y de Telecomunicaciones
Universidad de Granada
 
¡CSS Válido!   Valid XHTML 1.1   AAA validation