Document Actions
Invited speakers
Norwegian University of Science and Technology, Trondheim, Norway
Nature Inspired Hybrid Intelligent Systems
Mother nature has been a continuous source of inspiration for many innovative successful ideas, techniques, algorithms and
computational metaphors. In this talk, we focus on how such nature inspired algorithms could be used for the construction of hybrid intelligent systems. The integration of different learning and adaptation techniques, to overcome individual limitations and achieve synergistic effects through hybridization or fusion of these techniques, has in recent years contributed to a large number of new intelligent system designs. These ideas have led to the emergence of several different kinds of hybrid intelligent system architectures.
We present some of the generic modular hybrid architectures which have evolved over the past decade in the hybrid intelligent systems community. We further attempt to discuss the importance of these architectures with an emphasis on the best practices for selection and combination of intelligent methods. Two application examples will be presented to demonstrate how such systems could be used for solving real world problems.
______________________________________________________________________________
Charles Chidume
International Centre for Theoretical Physics, Trieste, Italy
chidume@ictp.it
_______________________________________________________________________________
Research Institute for Symbolic Computation,
Johannes Kepler University, Linz, Austria
Tudor.Jebelean@risc.uni-linz.ac.at
University of Eger, Hungary
Multi-Domain Logic and its Applications to SAT
We describe a new formalism and special proof methods for a novel generalization of propositional logic, which is especially suitable for solving the satisfiability problem (SAT).
A Multi Domain Logic (MDL) formula is a formula in first order logic, containing only atoms of the form "x belongs to A", where each "x" is a variable existentially quantified over the whole formula, and each "A" is a constant set. All logical connectives are allowed, but no other quantifiers. Note that classical propositional logic corresponds to the special case when each set is either {T} or {F}.
Similarly to propositional logic, the conjunctive normal form of a MDL formula is a conjunctive set of MDL clauses. Each MDL clause is a disjunction of literals abbreviated as "Ax", with at most one occurrence of each variable. The union of the sets corresponding to a certain variable can be seen as "the domain" of that variable, thus MDL is also a generalization of multi-valued logic, were each variable may have a different domain. The MDL SAT problem consists in finding solutions (that is, domain values for the variables) which satisfy all clauses.
A classical propositional SAT problem can be transformed in a MDL SAT problem by merging several variables in a MDL variable. For instance, the clause "A or B or (not C) or D" becomes "{TF, FT, TT}x or {FF, FT, TT}y", where "x" represents the combined values of "AB" and "y" represents "CD". (Any number of variables can be merged.) In contrast to the well known unit propagation (which is the basis of modern SAT solvers), this representation allows the simultaneous propagation of the combined values of several classical variables. The basic idea of MDL originates from the earlier work of the second author on “hyper-unit” propagation (that is simultaneous propagation of several unit clauses) [1] and on the representation and propagation of “k-literals” (generalized literals containing information on several propositional variables) [2].
The notions of resolution, subsumption, as well as the basic steps of the DPLL method generalize in an elegant and straightforward way, and open very promising research directions towards very efficient SAT solvers.
[1] G. Kusper. Solving the resolution-free SAT Problem by hyper-unit propagation in linear time. Annals of Mathematics and Artificial Intelligence 43(1-4), 2005, pp. 129-136.
[2] G. Kusper, L. Csoke. Better test results for the Graph Coloring and the Pigeonhole Problems using DPLL with k-literal representation. ICAI-2007, Volume II, pp. 127-135, University of Eger, Hungary, 2007.
______________________________________________________________________________
Matthias Hoelzl
Institute of Computer Science
Ludwig-Maximilians-University, Munich, Germany
hoezl@informatik.uni-muenchen.de
Constraints for Decisions with Multiple Objectives
The IST-FET Integrated Project SENSORIA is developing a comprehensive approach to the engineering of service-oriented software systems where foundational theories, techniques and methods are fully integrated into pragmatic software engineering processes.
To address problems like dynamic composition of services, we have developed a new soft-constraint formalism, called monoidal soft
constraints, that allows us to easily specify optimization problems in domains with multiple competing objectives and user preferences.
We give an overview of the general SENSORIA approach, present the theory of monoidal soft constraints and show how these constraints can be applied to various problems, such as service composition or the optimization or wireless radio networks.
______________________________________________________________________________
Laboratoire d'Informatique de Paris 6, UPMC-CNRS, F-75016 Paris, France
LMIB - School of Science, Beihang University, Beijing 100191, China
Triangular Decomposition for Algebraic and Geometric Computing
In this talk, we present several algorithms for decomposing systems of multivariate polynomials into triangular systems of various kinds. The algorithms have been efficiently implemented and successfully applied to numerous problems of scientific computing, ranging over computational polynomial algebra, automated geometric reasoning, solving systems of nonlinear equations, qualitative analysis of biological systems, and computer aided geometric design. We discuss some of the applications with a number of illustrative examples.
_______________________________________________________________________________
Czestochowa University of Technology, Poland
An Overview of Cell/B.E. Technologies
Power consumption, heat dissipation and other physical limitations are pushing the microprocessor industry towards multicore design patterns. One such example is the Cell Broadband Engine (or the Cell/B.E.), a novel architectural solution jointly developed by IBM, Sony, and Toshiba. It is an innovative heterogeneous multicore chip that is significantly different from conventional multiprocessor or multicore architectures. The initial target of the Cell/B.E. was Sony's PlayStation 3 game console, which boasts the chip with nine cores for faster and more realistic video gaming. Soon it became evident that the impressive computational power of the Cell/B.E. processor makes it potentially well suited for other applications such as visualization, multimedia processing, and various scientific and technical workloads. However, another lessons learned is that exploiting spectacular capabilities built-in the Cell/B.E. architecture will require new tools, new algorithms, and a new way of looking at its programming.
The goal of the talk is to discuss challenges, and present state-of-the-art and in-progress research of key aspects in Cell/B.E.
technologies. In particular, this talk explores experience from application developers in the use of the Cell/B.E. chip and performance of real applications, as well as results in the implementation of tools supporting the development and parallelization of applications.
______________________________________________________________________________