IIS'98

Seventh International Symposium on
Intelligent Information Systems
(Former WIS Series)
Malbork, Poland, 15-19 June, 1998

<iis@ipipan.waw.pl>
http://www.ipipan.waw.pl/~iis/

Accepted Papers / Lista przyjetych referato`w i komunikato`w.

Proceedings: Table of Contents / Spis tres`ci

Publisher: Institute of Computer Science PAS, Warsaw, 1998

ISBN 83-900820-9-8

• 1, Gryz Jarek ,
  "York University, Toronto",
  An algorithm for query folding with functional dependencies ,R
  Query folding is a technique for determining how a query may be answered using a given set of resources, which may include materialized views, cached results of previous queries, or queries answerable by other databases. Its applications include query optimization in centralized databases, query processing in distributed databases, and query answering in federated databases. The power of query folding can be considerably enhanced by taking into account integrity constraints that are known to hold on base relations. This paper describes an extension of query folding that utilizes functional dependencies to find foldings of queries that would otherwise be overlooked. We describe a strategy for finding foldings in the presence of functional dependencies and present a basic algorithm that implements that strategy.
  keywords Databases, Query Optimization and Evaluation, Materialized Views.
• 129, Kersten G.E., Szpakowicz S. , "Carleto University, University of Ottawa, Ottawa, Canada", Modeling business negotiations for electronic commerce ,R
• 63, Ras Zbigniew W. *) , Bergmann Anna **),"*)
  Univ. of North Carolina, Dept. of Comp. Science, **) University of Dortmund, Dept. of Comp. Science",
  Maintaining soundness of rules in distributed knowledge-based systems ,R

• 41, Kaufman K. A., Michalski Ryszard S. * , George Mason University, * also IPI-PAN, Multistrategy data mining via the KGL metalanguage ,R
• 106, Hippe Zdzislaw S ,,"Katedra Informatyki Chemicznej, Politechnika Rzeszowska", From research on searching hidden regularities in data , R
• 4, Michalewicz Maciej , Kl`opotek Mieczysl`aw A. , Wierzchon` Sl`awomir T. ",IPI-PAN," A concept of standard internet tools for integration of databases, knowledge bases and reasoning systems ",R,,,,
• 119, Jelonek Jacek, Krawiec Krzysztof, Stefanowski Jerzy , "Institute of Computing Sciences, Poznan University of Technology", Comparative study of feature subset selection techniques for machine learning tasks ,R
• 122, Ziarko Wojciech ,
  Department of Computer Science, University of Regina
  Rough sets methodology in data mining ,R
  Data mining is considered to be one of the most promising directions spanning database and AI research. The primary objective of database mining methodologies and systems is to help the user in discovering potentially significant facts or data patterns which are frequently "buried" in the mass of irrelevant information . The analitical methodology used for data mining presented in this article is based on the mathematical theory of rough sets.The theory of rough sets can be seen as a methodology for reasoning about relationships occuring in data. This makes it very well suited for data mining applications. As opposed to other techniques concerned primarly with acquisition of decision trees and rules from data, the methodology of rough sets provides a consistent set of methods aimed at identification of relationships in data, their representation and analysis. The relationships identified in data are represented in the form of a decision table which subsequently is analyzed using rough sets-based methods. The original model of rough sets is limited to the analysis of functional, or partial functional relationships. The probabilistic extension of the original model, called variable precision model, can also handle probabilistic relationships. The review of analytical capabilities of the variable precision model in the context of data mining applications is the main theme of the article. Keywords: data mining, decision tables, rough sets, data dependencies, reduct, core, variable precision rough sets
• 61, Duch Wl`odzisl`aw , Adamczak Rafal` , Gra`bczewski Krzysztof , Z|al Grzegorz ,
  Department of Computer Methods, Nicolaus Copernicus University,
  Hybrid neural-global minimization logical rule extraction method for medical diagnosis support ,R
• 116, Gl`owin`ski Cezary ,,IAIS PW, Discovering rules with various form of selectors ,K
• 57, Bobrowski Leon ,,Instytut Informatyki Polit. Bialostockiej + IBIB PAN, Rules extraction from data sets by using dipolar criterion ,K
• 139, De`bska Barbara ,,Pollitechnika Rzeszowska, The methodology of an automatic generation of a rule knowledge base from a set of cases ,R
• 118, Michalski Artur ,,Instytut Informatyki Politechnika Poznanska , First results with using rough set -based decision rules in reinforcement learning ,K
• 2, Bogdan Sergiusz , Kl`opotek Mieczysl`aw , PW ,IPI-PAN, A system searching for functional dependencies in continuous data. K,,,,
• 138, Matuszewski Andrzej ,, Instytut Podstaw Informatyki PAN, Characterization of dependence between two multivalued variables through the Rasch model - application to construction data-mining algorithms for survey data """,K
• 54, Ciupke Krzysztof, Maniak Piotr ,"Katedra PKM, Politechnika Slaska, Gliwice", Application of machine learning methods to diagnostics of machinery ,K,,,,

• 111, Przybyszewski Andrzej W. 12, Rubin Mark A. 2,
  1Dept. of Neurology, UMass Medical Center, Worcester, MA 01655, 2Dept. of Cognitive and Neural Systems, Boston University, Boston, MA 02215,
  Gating input to visual cortex by feedback to LGN , R
  Anatomical studies have documented massive back-projections from higher to lower visual cortices and to the lateral geniculate nucleus (LGN). The large number of synapses from these sources suggest that they should have a profound influence on the information carried by feed-forward inputs to these cells. However, the functional role of these connections is unclear. In order to explore the role of the feedback connections, we have recorded spike trains from electrodes placed in LGN in the macaque monkey under sufenta anesthesia, and have compared LGN cells' activity with and without suppression by cooling of feedback from primary visual cortex (V1). Normally, magno and parvo LGN cells show a wide range over which their responses are proportional to stimulus contrast. Inactivation of V1 feedback causes LGN cells to become more nonlinear and less sensitive to high contrast than during normal conditions. Responses during V1 inactivation have a similar shape to those of retinal ganglion cells. We have also tested the properties of the so-called extended surround as they relate to cortical activity and to influences on responses to LGN stimulation. A model of this data suggests an interpretation in terms of two functional components of feedback, a contrast-dependent component which dominates at high input contrast, and a constant baseline level of inhibitory feedback. We also show that the influence of the extended surround on the classical center mechanism is more complicated than a simple integration model.
  
  keywords lateral geniculate nucleus, feedback, top-down, bottom-up, cortical cooling
• 121, Jelonek Jacek , Krawiec Krzysztof, Sl`owinski Roman ,
  "Institute of Computing Sciences, Poznan University of Technology",
  Construction of textural features for classification of histological images ,K,,,,
• 42, Pacholczyk D. 1, Desmontils E. 2,"1 University of Angers, 2 Univ. of Nantes", A linguistic interpretation of affirmative or negative information in declarative modeling in image analysis ,R
• 136, Alferes Jose Julio, Joao A. Leite Lus Moniz Pereira, Przymusinska Halina , Przymusinski Teodor , College of Engineering, Department of Computer Science University of California , Updates of logic programs by logic programs ,R
• 51, Choras` Ryszard ,,Univ. of Technology and Agriculture Bydgoszcz, Computer vision techniques ,R,,,,
• 47, Buller Andrzej 1, de Garis Hugo 2, "1Politechnika Gdan`ska, 2 ATR Japonia", Brain-building strategy: some remarks and questions ,R
• 124, Kl`opotek Mieczysl`aw , Nowak Tomasz , ,"Wydzial` Fizyki Technicznej i Matematyki Stosowanej, Politechnika Warszawska, *IPI PAN", A simulation environment for testing behavior of a population of home-shopping agents. ,K
• 49, Daciuk Jan ,,Politechnika Gdan`ska, Finite-state automata in document retrieval ,K

• 133, Siminski Roman, Wakulicz-Deja Alicja ,Uniwersytet Slaski, Principles and practice in knowledge bases verification ,R
• 53, Moczulski Wojciech ,,"Politechnika /Sl/aska, Katedra Podstaw Konstrukcji Maszyn", Methods of knowledge acquisition in machinery diagnostics ,R
• 120, Kott Grzegorz, Wieleba Ryszard ,Centrum Informatyki Akademiii Obrony Narodowej, Expert system model to support decisions in command ,K
• 110, Lukaszewski Tomasz ,
  ,Poznan University of Technology Institute of Computing Science
  On reasoning under uncertainty in adaptive, distributed and multiagent expert systems ,K
  In this paper we discuss reasoning under uncertainty in distributed, adaptive and multiagent rule-based expert systems. It was shown that a multiple activation of rules is a main property of these systems. On the contrary to traditional systems such a multiple activation may lead to false results. Some solutions of this problem designed for the certainty factors model and the Dempster-Schafer theory are presented.
  Keywords: reasoning under uncertainty; adaptive, distributed and multiagent systems
• 55, Wylez|ol` Marek ,,"Katedra PKM, Politechnika Slaska, Gliwice", Electronic form - editor for acquisition of empirical rules from experts for the purpose of diagnostics ,K
• 59, Kostka Pawel` ,,Politechnika s`la`ska, Optimization of representation space for knowledge acquisition on rotating machinery using machine learning methods ,K

• 112, Mikol`ajczak Bolesl`aw ,
  Computer and Information Science Department, University of Massachusetts,
  High-level Petri Nets in development of intelligent parallel and distributed information systems ,R
• 62, Kacprzyk Janusz, Zadroz|ny Sl`awomir ,Systems Research Institute PAS, On summarization of large data sets via a fuzzy-logic-based querying add-on to Microsoft Access ,R
• 43, Koczkodaj W.W. ,.Herman M.W., Orlowsk M.i ,"Computer Science Laurentian University ", Knowledge elicitation, orthogonal basis and consistency driven pairwise comparisons ,R
• 48, Bartkowiak Anna, Szustalewicz Adam ,
  Institute of Computer Science, University of Wroclaw,
  Predictions in presence of outliers: Which tool to choose? K
  We consider two sets of data known to be benchmarks for regression in presence of outliers. These are: the $stackloss$ data and the Hawkins-Bradu-Kass data. The statistical methodology stresses the importance of detecting and accounting for the outlying and{/}or influential data vectors. The neural network methodology hopes that the problem will be tackled automatically by itself. We investigate in more detail the two approaches. We obtain the same goodness of approximation provided that we know about the outliers.
  Keywords: Regression, Prediction, Outliers, Radial basis neural network
• 5, Czech Zbigniew J , ,"Politechnika Slaska, Gliwice Instytut Informatyki", Solving the set-partitioning problem using tabu search ,K
• 126, Nowak Graz|yna, Hippe Zdzislaw S ,"Katedra Informatyki Chemicznej, Politechnika Rzeszowska", Logical-heuristic strategy for searching optimal transformations of molecular graphs ,K
• 3, Gomolin`ska Anna ,,
  Institute of Mathematics University of Bial`ystok,
  Belief change, nonmonotonic inference, and the Scott information systems , K
  In our approach belief change and nonmonotonic inference are modelled by means of Scott's information systems.
  keywordsbelief change, nonmonotonic inference, Scott's information system
• 137, Boryczka Urszula
  Institute of Computer Science, University of Silesia
  Optimization of the bus--service and the Ant Colony System , K
  This paper reports on one of a number of parallel investigations being undertaken into alternative approaches to solve a bus--scheduling problem. We use a parallel, multi--agent system known as the Ant Colony System, created by Dorigo et al. In the bus--scheduling problem it is very difficult to deduce which solution/ schedule is really optimal and the optimality of this solution is often subjective. We must define an accepted measure of quality of solutions given by the Ant Colony System. To solve this problem, we propose two algorithms in two versions: stochastic and deterministic. We tested these proposals for the fragment of the Silesian bus--system. The whole problem concerns over 120 bus--lines with about 1000 bus stops. keywords bus--service, bus--system, ant colony system
• 60, Duch Wl`odzisl`aw , Grudzin`ski Karol ,
  Department of Computer Methods, Nicolaus Copernicus University,",
  A framework for distance-based adaptive classification methods ,K

• 27, Cetnarowicz Krzysztof, Dobrowolski Grzegorz, Kisiel-Dorohinicki Marek, Nawarecki Edward ,
  Institute of Computer Science, University of Mining and Metallurgy, Krakow,
  Mechanisms of self-regulation in multi-agent systems , R
• 52, Obuchowicz Andrzej, Korbicz Jo`zef ,TU Zielona Gora, Evolutionary search with soft selection and forced direction of mutation ,R
• 125, Nowostawski Mariusz , Sloman Aaron ,
  School of Computer Science, University of Birmingham, UK
  Emergent thought - Automated agent design ,R
  The following report presents the preliminary results and theoretical backgrounds of a multi-layer evolutionary mechanism. The project aims at developing a mechanism for automated agent design within a pre-designed multi-layer architecture. The core mechanism used in the experiments is a Genetic Programming algorithm working in a specially designed condition-action rules system. Described multi-layer evolution paradigm is inspired by notion of the niche \& design spaces. There are presented plans for experiments with the multi-layer co-evolutionary mechanism for parallel distributed search within restrained search space. Attention is paid to evolutionary mechanisms in general, with no commitments to any particular problem. In theory the multi-layer evolutionary process can be applied to a number of systems and problems, e.g. to Genetic Programming.
  keywords agents, genetic algorithms, evolutionary mechanisms, architectures
• 66, Kosin`ski Witold 1), Michalewicz Zbigniew 2), Weigl Martyna 1), Koles`nik Roman 3), 1)Instytut Podstawowych Problemów Techniki PAN, 2) Univ. of North Carolina, Dept. of Comp. Science, 3) Politechnika Warszawska Genetic algorithms for preprocessing of data for universal Approximators ,R
• 114, Kwas`nicka Halina ,,"Wydzialowy Zaklad Informatyki, Politechnika Wroclawska", Efficiency of genetic algorithms with plejotropy and polygene effect -Simulation Study ,K
• 123, Zelechowska Joanna (1) , (2) Buller Andrzej ,
  "(1) Institute of Mathematics University of Gdansk, Poland, (2) Faculty of Electronics, Telecommunications and Informatics Technical University of Gdansk, Poland",
  Neural network-based system for determining and classifying malignant and benign lesions in ultrasonic images of the breast K
• 131, Balicki Jerzy, Kitowski Zygmunt ,,Akademia Marynarki Wojennej , Relaxation methods of artificial neural networks for solving NP-hard module assignment optimization problem in two-processor systems with constrained resources ,K
• 40, Seredyn`ski Franciszek ,
  IPI PAN
  Genetic algorithms-based intelligent decision support system to Task Scheduling Problem ,K
  An approach to develop parallel and distributed algorithms of scheduling tasks in parallel computers with use of learning machines is proposed. Coevolutionary multi-agent system with game theoretical model of interaction between agents serves as a theoretical framework of the approach. Genetic-algorithms based learning machines called classifier systems are used as players in a game. Experimental study of such a system shows its self-organizing futures and the ability of emergent behavior. Following this approach a parallel and distributed scheduler is described. Following this approach a parallel and distributed scheduler is described.Results of the experimental study of the scheduler are presented.
• 46, Kwedlo Wojciech, Kre`towski Marek ,
  Politechnika Bial`ostocka,
  Learning decision rules using an evolutionary algorithm and entropy-based discretization ,K
  Keywords: learning decision rules, evolutionary algorithms, discretization
• 58, Pawlak Marek ,,Politechnika Lubelska, Application of heuristic algorithms and an evolutionary algorithm to production scheduling ,K
• 65, Skowron`ski Krzysztof ,,Politechnika Warszawska, Genetic rules induction based on MDL principle ,K,,,,

• 108, Sikora Marek 2, Mro`zek Adam 1
  "1 IITiS Gliwice, 2Pol.S`la`ska Gliwice.",
  Semantic oriented synthesis of a rule knowledge base integrated inference engine. ,R
  In this article we present the proposition of connection between a rule knowledge base and a sequential automata for the purpose of modelling an observed behaviour of a man solving defined problem. For decision rules, connection of knowledge base and sequential automata is introduced as of an extended decision table.
  Key words: decision table, sequential automata, modelling behaviour, program synthesis.
• 109, Grzymal`a-Busse Jerzy W. , Soe Than ,University of Kansas, Inducing simpler rules from reduced data ,R
• 128, Onis`ko Agnieszka 1, Druz|dz|el Marek J. 2., Wasyluk Hanna 3 ,
  "1Politechnika Bial`ostocka, 2 University of Pitsburg, 3 CMKP",
  A probabilistic model for diagnosis of liver disorders ,R,,,,
• 107, Stepaniuk Jarosl`aw , Urban Miroslawa , Baszun-Stepaniuk Elz|bieta , "12nd Dept. of children's diseases, university medical school, Bialystok, Inst. Comp. Sc., Bialystok University of Technology", The application of rough set based data mining technique in the prognostician of the diabetic nephropathy prevalence ,K
• 115, Drwal Grzegorz 1 , Mro`zek Adam 2 ,"1Politechnika S`la`ska, 2IITiS PAN", System RClass - software implementation of a rough classifier ,K
• 132, Marszal`-Paszek Barbara, Paszek Piotr , Uniwersytet Slaski, Extract strong relationship between data from decision table ,K,,,,

Back to IIS'98