Tadeusz Dyduch
, Krzysztof Cetnarowicz
Abstract. These papers are aimed at showing the basic virtues of the agent approach to the management system. We will show the logical skeleton of the management system and reasons why the multiagent environment is suitable to implement it. Our research concentrates on creating a complete closed-loop MRP class application. Especially, we would like to show modification and developing perspectives. The following article also gives a short introduction to the M-agent theory. Finally, we will show the features of this approach like calibration, system topology and cooperation with different database systems.
Full version:
Iwona Karcz-Dulęba
Abstract. Distributions of traits in a population provide important information about evolution of the population itself. In this paper an analysis of traits distributions in a phenotypic evolution is presented. A very simple model of evolution is under consideration - infinite populations evolve in one dimension space of a bimodal fitness function. The analysis of dynamic behavior of a population yields an interesting result concerning generation of the subsequent distributions. It appears that every normal distribution generates two offspring normal distributions. The evolutionary process initialized with a single normal distribution grows up to 2t normal distributions after t generations. The evolution of normal distributions is described equi\-valently by evolution of their parameters: means and variances. The evolution of distributions' means resembles fractals generated by an Iterated Function System (IFS). Equations describing the location of distributions' means in the next generation define contractive affine transformations. The defined iterative system maps the interval [0,1] into the Cantor set after infinite number of iterations.
Full version:
Marek Kisiel-Dorohinicki
Abstract. The concept of decentralised evolutionary computation realised as evolutionary multi-agent system (EMAS) is described in the paper. Also agent-based evolutionary approach to neural network architecture optimisation is presented. Then the problem of time-series prediction and a general idea of evolutionary neural multi-agent predicting system is introduced. Selected design issues together with preliminary simulation results conclude the work.
Full version:
Jacek Kitowski
Abstract. In this paper we present some methods and algorithms for large scale computing which cover different areas of computational and computer sciences. They concern particle models, CFD computing, animation, monitoring and predicting of application performance as well as scientific visualization and scientific data storing and retrieving.
Full version:
Przemysław Magiera, Andrzej Karbowski, Infinite Time Horizon Optimal Control with Expected Value-Variance Criteria Made Possible by Decomposed Optimization, pp.69-81.
Abstract. In the paper, a method of determining optimal control with the expected value-variance criteria is presented. The method enables the reduction in risk of getting some undesirable values of a performance index. As in most multicriteria infinite horizon control problems we apply transformation into linear programming problem. Unfortunately, the complexity of the obtained optimization problem is very high. Decomposition methods, which base on a specific form of a problem seem to be the only rescue. A practical problem of investing in some assets at the Warsaw Stock-Exchange is considered. It has been transformed into a linear programming problem. Some trials have been undertaken to solve the problem with the aid of a software package HOPDM (Higher Order Primal-Dual Method) which is used for solving large scale linear programming problems.
Full version:
Jerzy Martyna,
Abstract. In this paper we present a model for coalition formation in a Multi-Agent System with incomplete information. Our approach is partitioned into two levels: the social level and the strategic level. The social level is focused on some strategies for the interaction in a given environment. The strategic level consists of strategies for individual agent to act in an uncertain environment for maximization of its own expected payoff.
Full version:
Jerzy Martyna,
Abstract. Properties of actual and future languages for Multi-Agent System designing are outlined. This approach includes principles of Multi-Agent System programming: description of agent mental state, synchronous and asynchronous communication mechanisms, agent interpreter cycle, etc. An alternative approach is to adopt the used programming languages to design the Multi-Agent System. This approach is obviously more difficult or almost impossible to implement. Nevertheless, in this paper a concept of adopting language to Multi-Agent System designing is presented.
Full version:
Jan J. Mulawka, Tomasz Janczak, Artur Malinowski, Robert Nowak, DNA Computing - Promise for Information Processing, pp.113-130.
Abstract. Molecular computing is a new multidisciplinary area of research concerned with design, and theoretical and empirical understanding of computing gleaned from nature. It aims at the implementation of algorithms in molecular hardware, e.g. using DNA molecules and enzymes. The main advantages of the molecular computing lie in the possibility to obtain a massive parallelism of computation and associative memory of high capacity. The primary objectives of this paper are to discuss fundamental concepts of molecular computing and basic processes of standard genetic engineering used for DNA computing. In particular, such processes like hybridization, ligation, PCR, affinity separation, restriction enzyme digestion, and electrophoresis will be considered. Next, we provide computation models and algorithms applied in this area. We show that biochemical reactions may be used for performing molecular inference according to the algorithm of either forward or backward chaining. Finally, we concentrate on realization issues. Basic properties of DNA chips as well as data flow microreactors will be considered.
Full version:
Lech Polkowski, Maria Semeniuk-Polkowska, Towards Usage of Natural Language in Approximate Computation: A Granular Semantics Employing Formal Languages over Mereological Granules of Knowledge, pp.131-146.
Abstract. Granularity of Knowledge [15] has been proposed as a metaphor with which to capture the basic phenomenon of Reasoning under Uncertainty viz. the presence in the reasoning process and in its symbolic versions of constructs representing vaguely defined collections of objects drawn together by some similarity relations, e.g. fuzzy ones (cf. [14]). Similarly, the idea of Computing with Words has been proposed [14] as a computational paradigm in which certain words/phrases of Natural Language are labels for granules of knowledge and semantics of a chosen subset of Natural Language is realized via a calculus of granules of knowledge, e.g. by means of Fuzzy Logic. In this work, we would like to make a further step towards this end (cf. [9]) and we modify here a scheme for reasoning under uncertainty based on approximate mereological calculus in distributed systems proposed earlier (cf. [10]) . Vague specifications coding synthesis problems are rendered as phrases of Natural Language and interpreted as approximate formulae in logics for approximate reasoning over distributed systems of intelligent agents endowed with knowledge, in particular with approximate mereological predicates for constructing granules of knowledge. Calculi of those granules and the resulting formal grammars and languages denoting semantically the chosen phrases of Natural Language are invoked here (cf. [12]). The results presented here constitute a step towards a certain fulfillment of Computing with Words program. It is expected that those languages will be applicable in problems of Intelligent Control, e.g. in Mobile Robotics and in problems of Approximate Reasoning in Natural Language (cf. our earlier work [8]).
Full version:
Joël Quinqueton, Emergent Problem Solving in Multi-Agent Systems, pp.147-163.
Abstract. Problem solving with multi-agent systems appears to be a very active research area nowadays. We present in this talk an overview of some works around distributed Constraint Satisfaction Problems and we point out some recent results about phase tarnsition and emergence in multi-agents resolution of such problems. We also point out the links and differences with distributed or randomized algorithms.
Full version:
Maria Staniszkis
Abstract. The study of multiagent distributed systems through the optics of the theory of formal grammars and languages is presented. Models based on NLC and BNLC graph grammars are mainly considered. The basic intuition of a grammar model of distributed systems follows the blackboard model for problem solving and is the following: component grammars forming the distributed system execute rewritings on a shared sentential form/graph. The order of participation on rewriting, the start and stop conditions of participation are determined by the strategies for the cooperation of components. The strategies of cooperation considered include the ones based on chains or linear, partial and arbitrary orders control relations. The introduced transformations on graphs representing the control permit one to obtain results regarding the synergy created by the cooperation.
Full version: