Invited talks





Information Theoretic Signal Processing and Machine Learning

Dr. Jose C. Principe
Distinguished Professor of Electrical Engineering
University of Florida

Jueves 21 de mayo de 2009, 11:00,
Facultad de Informática, Aula de Grados

Organiza: LIDIA (Laboratorio para la Investigación y Desarrollo en Inteligencia Artificial)
Análisis de sensibilidad en problemas de optimización y de fiabilidad

Dr. Enrique Castillo Ron
Departamento de Matemática Aplicada
Universidad de Cantabria

Martes 10 de marzo de 2009, 12:30,
Facultad de Informática, Aula de Grados

Organiza: LIDIA (Laboratorio para la Investigación y Desarrollo en Inteligencia Artificial)
Sistemas Multiagente Avanzados

David Camacho Fernández
Departamento de Informática, Universidad Autónoma de Madrid

Lunes 19 de mayo de 2008, 13:00,
Facultad de Informática, Aula 2.6

Resumen

Partiendo de las facilidades de diseño, implementación y prueba que proporcionan las plataformas de múltiples agentes, la utilización de técnicas avanzadas provenientes de áreas como la Inteligencia Artificial, la Teoría de la Información, o la Neurociencia permiten la exploración de nuevas técnicas de resolución de problemas, así como el diseño de entornos de simulación apropiados para este tipo de algoritmos. En particular este seminario pretende mostrar cómo se están aplicando algunas de estas técnicas, relacionadas principalmente con la Inteligencia Artificial, la Teoría de la Información o la Neurociencia sobre problemas concretos donde la utilización de Sistemas Multiagente permiten obtener una serie de ventajas derivadas de la utilización de los fundamentos tecnológicos en los que estos se basan.

ÍNDICE:
1. Introducción y Estado de la cuestión
2. Desarrollo de motores de búsqueda basados en SMA y compresión.
3. Desarrollo de sistemas de preguntas frequentes (Question Answering Systems) dinámicos.
4. Desarrollo de SMA basados en firmas neuronales.

Organiza: LIDIA (Laboratorio para la Investigación y Desarrollo en Inteligencia Artificial)
Desarrollo de Sistemas Adaptativos sobre Internet

David Camacho Fernández
Departamento de Informática, Universidad Autónoma de Madrid

Lunes 19 de mayo de 2008, 12:00,
Facultad de Informática, Aula 2.5

Resumen

Los Sistemas Multiagente (SMA) permiten el diseño e implementación de sistemas computacionales complejos formados por un grupo distribuido de elementos o agentes. En particular, este tipo de sistemas permite coordinar a los posibles agentes de manera distribuida de tal forma que mantengan un comportamiento global coherente. Los agentes que forman estos sistemas disponen de características como la autonomía, que les permite dotar al sistema no sólo de una mayor flexibilidad y robustez, sino de la posibilidad de que diferentes partes del sistema traten de alcanzar sus objetivos desde diferentes perspectivas, es decir, los sistemas Multiagente permiten la existencia de múltiples comportamientos entre los agentes siempre que existan los mecanismos adecuados que regulen las posibles interacciones entre los mismos. La combinación de las tecnologías basadas en agentes, junto con técnicas como la Planificación o el Aprendizaje Automático, permiten el desarrollo de sistemas flexibles, autónomos y con capacidad de razonamiento sobre el entorno en el que se hayan envueltos. Este seminario pretende mostrar cómo pueden aplicarse diversas técnicas relacionadas con la Inteligencia Artificial para permitir el desarrollo de Sistemas Adaptativos en entornos complejos y dinámicos
como Internet.

Organiza: LIDIA (Laboratorio para la Investigación y Desarrollo en Inteligencia Artificial)
Constructive Negations and Paraconsistency

Dr. Sergei Odintsov
Sobolev Institute of Mathematics,
Novosibirsk, Russia

Martes 4 de marzo de 2008, 11:30,
Facultad de Informática, Aula de Grados

Abstract

This talk presents Dr. Odintsov’s recent investigations of the two main concepts of negation developed in the constructive logic: the negation as reduction to absurdity (L.E.J. Brouwer) and the strong negation (D. Nelson) are studied in the setting of paraconsistent logic. The paraconsistent logics are those, which admit inconsistent but non-trivial theories, i.e., the logics which allow making inferences in non-trivial fashion from an inconsistent set of hypotheses. Logics in which all inconsistent theories are trivial are called explosive. In the intuitionistic logic Li, the negation is defined as reduction to absurdity. The concept of strong negation is realized in the Nelson logic N3. Both logics are explosive and have paraconsistent analogs: Johansson’s logic Lj and paraconsistent Nelson’s logic N4. It will be shown that refusing the explosion axiom "contradiction implies everything" does not lead to decrease of the expressive power of a logic. To understand, which new expressive possibilities have the logics Lj and N4 as compared to the explosive logics Li and N3, we study the lattices of extensions of the logics Lj and N4. This is the first case when lattices of paraconsistent logics are systematically investigated. The study is based on algebraic methods, demonstrates the remarkable regularity and the similarity of structures of both lattices of logics, and gives essential information on the paraconsistent nature of logics Lj and N4.

ORGANIZA: Pedro Cabalar, Departamento de Computación.   
Normal Forms for Answer Set Programming

Dr. Alessandro Provetti
Università degli Studi di Messina, Italia

Martes 24 de septiembre de 2007, 12:30,
Facultad de Informática, Aula de Grados

Abstract

This talk introduces to the author's work, in collaboration with S. Costantini, on normal forms for logic programs under stable/answer set semantics. These forms may simplify the study of program consistency, and suggest new algorithms for finding answer sets. Also, the graph-based representation of Answer Set Programs will be introduced and discussed.

ORGANIZA: Pedro Cabalar, Departamento de Computación.

Prototypal Applications of automated reasoning with defaults and preferences

Dr. Alessandro Provetti
Università degli Studi di Messina, Italia

Martes 24 de septiembre de 2007, 11:00,
Facultad de Informática, Aula de Grados


Abstract

This seminar presents the author's line of research on automated reasoning with rules in challenging scenarios. In the first part, the Answer Set Programming, its semantics of rules and it mathematical framework are briefly presented. In the second part, the PPDL specification language, built on Answer Set Programming with preferences, is presented and some proof-of-concept applications in (i) interactive computer games, (ii) Web service connection, and(iii) Web data collection and annotation is discussed. Informations and articles are available from the author's research group Web page: http://mag.dsi.unimi.it/

ORGANIZA: Pedro Cabalar, Departamento de Computación.
Answer-set programming – solving search problems with logic

Dr. Miroslaw Truszczynski
University of Kentucky

Jueves 17/09/2007, 17:00
Facultad de Informática, Aula de grados

Abstract

Answer-set programming is a way of using logic to solve search problems. A general approach is as follows. Given a search problem, one first designs a theory in some logic so that models of this theory represent solutions to the problem. Next, one computes computes a model of the corresponding theory and, finally, one reads back the solution encoded by the model.  Several answer-set programming formalisms have been developed on the basis of logic programming with the semantics of stable models. The logic of predicate calculus, and the closely related formalism of model expansion, also gives rise to effective implementations of the ASP paradigm, similar in the spirit to logic programming with stable model semantics and with a similar scope of applicability. In the talk we will present and contrast the two approaches, emphasizing their fundamental aspects: high-level modeling, rooted in nonmonotonic reasoning and declarative programming, and low-level search, related to satisfiability. We will discuss basic results concerning the complexity and expressive power. We will discuss some of the current research directions and open problems.


ORGANIZA: Pedro Cabalar, Departamento de Computación.
Tableaux methods for Equilibrium Logic

Agustín Valverde
Universidad de Málaga

Lunes 05/03/2007, 12:30
Facultad de Informática, Seminar 4.1
The Query-Vector Document Model

Fabrizio Silvestri
ISTI-CNR, Pisa, Italia

Friday 24/11/2006, 12:00
Facultad de Informática, Seminar 4.1

Abstract

We present a novel strategy to partition a document  collection onto several servers and to perform effective collection  selection. The method is based on the analysis of query logs. We proposed a novel document representation called query-vectors  model. Each document is represented as a list recording the  queries for which the document itself is a match, along with their  ranks. To both partition the collection and build the collection  selection function, we co-cluster queries and documents. The  document clusters are then assigned to the underlying IR servers,  while the query clusters represent queries that return similar  results, and are used for collection selection. We show that this  document partition strategy greatly boosts the performance of  standard collection selection algorithms, including CORI, w.r.t.  a round-robin assignment. Secondly, we show that performing  collection selection by matching the query to the existing query  clusters and successively choosing only one server, we reach an  average precision-at-5 up to 1.74 and we constantly improve  CORI precision of a factor between 11% and 15%. As a side  result we show a way to select rarely asked-for documents. Separating these documents from the rest of the collection allows the indexer to produce a more compact index containing only  relevant documents that are likely to be requested in the future. In our tests, around 52% of the documents (3,128,366) are not  returned among the first 100 top-ranked results of any query.

Organized by: Álvaro Barreiro (barreiro [at] udc.es), Dept. Computación.
Origen y evolución de la cibernética

Roberto Moreno Díaz
Universidad de Las Palmas de Gran Canaria

Friday 28/4/2006, 12:30
Facultad de Informática, Room 3.3

Organized by: LFCIA.
Walsh - Functions for signal processing and code division multiplexing

    Franz Pichler
    Universidad Johannes Kepler, Austria

Thursday 27/4/2006, 12:30
Facultad de Informática, Room 3.3

Organized by: LFCIA.
From meta-grammars to factorized grammars

Éric Villemonte de la Clergerie
INRIA, France

Wednesday 15/3/2006, 11:30
Facultad de Informática, aula de grados

Abstract

Meta-Grammars [MGs] allow modular and factorized descriptions of syntactic phenomena. I will show how the factorization provided by MGs coupled with the use of factorization operators provided by DyALog system may be used to generate very compact Tree Adjoining Grammars [TAGs]. I will complete the presentation by a description of our development environment for MGs and of FRMG, a large coverage French MG. Evaluation results will also be provided for FRMG.

Organized by: Miguel A. Alonso Pardo & Jesús Vilares Ferro, Dept. Computación.
Regular Sound Changes for Cross-Language Information Retrieval

Michael P. Oakes,
Sunderland University,
United Kingdom

Tuesday 14/3/2006, 11:30
Facultad de Informática, aula de grados

Abstract

Cognate words are vocabulary items which occur in two or more historically related languages, such that they have similar meanings, and one can be transformed into the other by a predictable series of phonological changes. The discovery of such sound changes is potentially relevant to Cross-Language Information Retrieval: the automatic conversion of query terms in one language into their equivalents in a second, historically related, language will enable documents in that second language can be retrieved. We will describe one computer program called JAKARTA, that given a sample word list from two related languages, extracts the probable rules for predicting any word of one language from that of the other. As a case study we will examine vocabulary lists for Galician and Spanish.

Organized by: Miguel A. Alonso Pardo & Jesús Vilares Ferro, Dept. Computación.
Ontology and rule languages for the Web

Axel Polleres
Univ. Rey Juan Carlos, Móstoles

Viernes 3/3/2006, 12:30,
Facultad de Informática, Aula de grados

Abstract

The Ontology Layer in the Semantic Web architecture has reached a more or less stable state by the W3C recommendations of RDF/RDFS and OWL. Next to tackle in this architecture is the rules layer. However, adding rules on top of OWL and RDF does not turn out to be straightforward. This talk will summarize current efforts from research and standardization bodies towards "reasonable" rules extentions for the Semantic Web and report on ongoing efforts such as the W3C's rule Interchange Format (RIF) working  group on the one hand and I will also outline my own ongoing work in a focus subarea, dealing with contextually scoped negation in rules describing Web metadata.

Organized by: Pedro Cabalar, Dept. Computación.
Some remarks on the history of constructive negation

Sergei Odintsov
Sobolev Institute of Mathematics,
Novosibirsk, Russia

Jueves 2/3/2006, 19:00,
Facultad de Informática, Aula de grados

Abstract

During 20th century the concept of constructive negation arose several and under different names. Among them are "strong negation", "explicit negation" and even "classical negation". Namely under the name "classical" a negation of this kind was introduced into the logic programming by M. Gelfond and V. Lifschitz. In my report I review the most important points of this history.

Organized by: Pedro Cabalar, Dept. Computación.