|
Research Institute of Intelligent Computer Systems Ternopil National Economic University |
|
2005, Vol. 4, Issue 1 |
|
Contents and abstracts
COMBINING BAYESIAN NETWORKS AND ROUGH SETS: Janusz Zalewski 1), Sławomir T. Wierzchoń 2), Henry L. Pfister 3)
1) Florida Gulf Coast University, Ft. Myers, FL 33965, USA, zalewski@fgcu.edu, http://www.fgcu.edu/zalewski/ This paper discusses a combination of Bayesian belief networks and rough sets for reasoning about uncertainty. The motivation for this work is the problem with assessment of properties of software used in real-time safety-critical systems. A number of authors applied Bayesian networks for this purpose, however, their approach suffers from problems related to calculating the conditional probability distributions, when there is scarcity of experimental data. The current authors propose enhancing this method by using rough sets, which do not require knowledge of probability distributions and thus are helpful in making preliminary evaluations, especially in real-time decision making. The combination of Bayesian network and rough sets tools, Netica and Rosetta, respectively, is used to demonstrate the applicability of this method in a case study of the Australian Navy exercise.
TEXTURE CLUSTERING OF SATELLITE IMAGES Lukashevich M.M. 1), Sadykhov R.Kh. 2)
1) BSUIR, Minsk, P.Brovka str., 6, Minsk, Belarus kafevm@bsuir.by The goal of this paper is to present a texture clustering system for remote sensing image data. Texture information is useful for image data browsing and retrieval. Authors present the results of self-organizing neural network design for solving the clustering task of gray scale remote sensing image data. The architecture of neural network and the learning algorithms for this network such as: algorithm WTA (Winner Takes All), algorithm CWTA (Winner Takes All with Conscience) and classic Kohonen algorithm WTM (Winner Takes Most - the Winner receives more) are considered. Some experimental results using textures of the Brodatz album, multi-spectral and radar images are also represented.
COMPARISON OF SEVERAL METHODS FOR PARETO SET Ingrida Radziukyniene 1), Antanas Žilinskas 2)
1) University of Florida, 303 Weil Hall, Gainesville, FL 32611-6595, USA, ingridar@ufl.edu, Pareto set generation methods are considered with respect to their application for multi criteria portfolio selection. Several such methods were compared experimentally including some recently proposed evolutionary methods and the method of adjustable weights. Test problems were based on standard portfolio quality criteria and data on stocks of 10 Lithuanian companies. The experimental data on the performance of the considered algorithms in different metrics are presented and discussed.
NEURAL-NETWORK SEGMENTATION OF ELECTROENCEPHALOGRAM Svetlana Bezobrazova 1), Vladimir Golovko 2)
1) Brest State Technical University, Moskovskaya str. 267, 224017, Brest, Belarus, Svetilka@gmail.com A goal of EEG signals analysis is not only human psychologically and functionality states definition but also pathological activity detection. In this paper we present an approach for epileptiform activity detection by artificial neural network technique for EEG signal segmentation and for the highest Lyapunov’s exponent computing. The EEG segmentation by the neural network approach makes it possible to detect an abnormal activity in signals. We examine our system for segmentation and anomaly detection on the EEG signals where the anomaly is an epileptiform activity. UNIVERSAL EMBEDDED RECONFIGURABLE HARDWARE PLATFORM FOR MULTIMEDIA APPLICATIONS IN REAL-TIME
Michael Livshitz, Alexey Petrovsky, Andrey Stankevich,
Computer Engineering Department, This paper deals with reconfigurable hardware platform for different purposes real-time speech and audio signal processing. A design conception and turnkey solution are described. Much attention is paid to reconfigurable peripheral processor meant for external interface realization, pre- and post- data processing as well as digital signal processing algorithms implementation with the object of the DSP unloads. Moreover, three applications implemented on the considered platform are demonstrated.
AHP METHOD APPLICATION FOR THE EVALUATION OF INTERNET Witold Chmielarz
Faculty of Management, Warsaw University The main objective of this paper is an analysis of possibilities of using Saaty’s AHP method in the evaluation of internet websites. The evaluation range is limited to selected websites of the computer shops which are most frequently visited by customers. In the beginning of the article basic assumptions of AHP method are presented. Next, the author shows AHP application in the dimensioning of websites as well as implications resulting from this approach. The last part of this article contains conclusions related to the analysed approach and claims for further research.
MULTIPLE VIEWS IN PEER DATA MANAGEMENT M.B.Al-Mourad 1), Rozalina Mohamed 2), Yaser M. A. Khalifa 3)
1) College of Information Technology
This document presents the required layout of papers to be submitted for publication in the “Computing”
International journal. The abstract may not be longer than 150 words. Peer-to-peer (P2P) systems are revival
paradigm for information sharing among distributed nodes in the network. A P2P network is a network that relies
primarily on the computing power and bandwidth of the participants in the network rather than concentrating it in a
relatively low number of servers. P2P software systems like Kazaa and Napster rank amongst the most popular
software applications ever. Numerous web businesses and sites have promoted "peer to peer" technology as the future
of Internet networking for E-commerce. Multiple views for data are created for mediating between data sources on the
Semantic Web. Our goal is to support users’ different needs. This is due to the fact that different users have different
needs for joining the P2P community and their requirements may change over time as new information become
available. Hence the same information may participate in many different ways in multiple data sources’ mapping
efforts.
IMPLEMENTATION OF NEURAL NETWORKS AND BOOSTING Vladimir Golovko 1), Leanid Vaitsekhovich 2)
Brest State Technical University, In this article the classification task in the domain of intrusion detection is considered. Often a chosen algorithm is not good enough for practical use. So the question arises how is it possible to improve the performance? In this case we can employ so-called Committee Machines that increase accuracy and reliability of the base classification model. These advantages are the result of dividing complex computational problems among several experts. The knowledge of each expert influences on the general conclusion of Committee Machine.
HIERARCHICAL CLUSTERING ALGORITHM FOR DETECTING Rachid Beghdad
Faculty of sciences, 12 boulevard Bouaouina, Béjaïa 06000, Algeria. We introduce a new intrusion detection method based on the Hierarchical Clustering Algorithm (HCA), to detect anomalous user’s profiles. In the Unix system, a simple user has only some privileges (can access to some resources), but the root user has more privileges. So, we can speak here about hierarchy of users. By the same way, we can use a hierarchy of users in intrusion detection field, to distinguish between the normal user and suspicious user. Many data mining methods were already used in previous works in intrusion detection. Even if some of them led to interesting results, but they still suffer from some weaknesses. This is the reason why we focused in this study on the use of the HCA to detect anomalous profiles. A survey of intrusion detection methods is presented. The HCA procedure is described in detail. Our simulation results demonstrate the robustness of our approach in comparison to some previous used methods.
CHC ALGORITHM FOR ANTENNA ARRAY Bogdan Artyushenko, Galina Shilo, Volodymyr Krischuk
Zaporizhzhya National Technical University A new CHC based method for optimization of antenna array, with failed elements, is developed. Performances of CHC and canonical genetic algorithm are compared. Numerical results show that CHC gives better results with the same working time. To improve computational efficiency parallelization possibilities are studied.
AN NUMERICAL METHOD BASED ITERATIVE PROCESS Mohamed Tellache 1), Youcef Lamhene 1), Brahim Haraoubia 1), Henri Baudrand 2)
1) Laboratory of Instrumentation (LINS), Faculty of Electronics and Computers, In the present work, the modeling of microwaves planar circuits is proposed with an original method based on the Waves Concept Iterative Process (WCIP). It consists in the development of simulation software based on an iterative method. The iterative method is developed from the fast modal transform on a two-dimensional fast Fourier transform (FFT) algorithm. The method has been applied to the characterization and the modeling of patch antennas with notches in microstrip and coplanar technology and the quarter wavelength directive coupler. The obtained results are very powerful and successfully compared to others methods in term of time and reliability of convergence and particularly the accuracy of the results obtained in comparison with previous works. ACTIVE SYSTEM MANAGEMENT UNDER UNCERTAINTY Shakah G. 1), Krasnoproshin V.V. 2), Valvachev A.N. 3)
1) Irbed National University: The paper describes the use of fuzzy set theory and theory of active systems for constructing systems that manage geographically distributed organizations under uncertainty. Unification algorithms for fuzzy data and their use for choosing management of distant objects are presented. A NEW APPROACH TO SHAPE-BASED IMAGE RETRIEVAL Mehdi Chehel Amirani 1), Zahra Sadeghi Gol 2), Ali Asghar Beheshti Shirazi 3)
1) Iran University of Science and Technology, Narmak, Tehran, Iran, amirani@ee.iust.ac.ir Content-based image retrieval (CBIR) is very active research topic in recent years. This paper introduces a new approach to shape-based image retrieval. At first, feature points are determined at the boundary of the shape as the extremums of a new version of the curvature function and the initial features are calculated at these points. The proposed method utilizes a supervised system for nonlinear combination of initial features for extraction of efficient and low dimensional feature vector for each shape. The retrieval performance of the approach is illustrated using the MPEG-7 shape database. Our experiments show that the proposed method is well suited for object indexing and retrieval in large databases.
RELIABLE AND EFFICIENT PARALLEL COMPUTING Aleksej Otwagin
United Institute of Informatics Problems, National Academy of Sciences of Belarus, Basic principles of reliable parallel computations are considered. The parallel program is represented as a graph computation schema, that executed by unreliable computing system with possible node faults. For organization and optimization of parallel processing a kind of multiagent architecture was used. The proposed solution uses the principles of runtime evolutionary optimization to increase performance characteristics.
NEGOTIATION AGENT BEHAVIORS BASED ON REINFORCEMENT Amine Chohra, Arash Bahrammirzaee, and Kurosh Madani
Images, Signals, and Intelligent Systems Laboratory (LISSI / EA 3956), Behaviors, in which the characters conciliatory, neutral, or aggressive define a ‘psychological’ aspect of human personality, play an important role for negotiation agent. Elsewhere, learning in negotiation is fundamental for understanding human behaviors and developing new concepts. In this paper, a negotiation strategy essentially based on such human personality behaviors is suggested for SISINE project which aims to develop innovative teaching methodology of negotiation skills. For this purpose, first, reinforcement learning (Q-learning and Sarsa-Learning) approaches are developed, analyzed, and compared in order to acquire the strategy negotiation behaviors. Second, a Fuzzy ArtMap Neural Network (FAMNN) is developed to acquire this strategy. Third, a Field Programmable Gate Array (FPGA) architecture is suggested for the FAMNN integration. The suggested strategy displays the ability to provide agents, through a basic buying strategy, with a first intelligence level in a social and cognitive system for learning negotiation strategies (human-agent and agent-agent).
Neurocomputer Based Complexity Estimator Optimizing Ivan Budnyk, El khier Bouyoucef, Abdennasser Chebira, Kurosh Madani
Image, Signal and Intelligent Systems Laboratory (LISSI / EA 3956), This paper presents application of ZISC© IBM® neurocomputer based approach for estimating task complexity within T-DTS framework. T-DTS (Tree-like Divide To Simplify) is Hybrid Multiple Neural Networks software platform which constructs a neural tree structures of a complex problem following the paradigm “divide” and “conquer”. Complexity estimator modules are the core of this framework. One of them is ZISC© IBM® complexity estimator that has been recently applied to T-DTS. The global aim of this research work is to increase T-DTS performance in terms of generalization and learning abilities. In this paper we demonstrate matchless ZISC© IBM® based neurocomputer complexity estimator effect on database decomposition and searching for optimal T-DTS adjustment of complexity threshold.
FUZZY QUERIES IN BLACK-SCHOLES MODEL Arkadiusz Banasik
Silesian University of Technology, An approach how to use fuzzy queries in field of financial options is described. It provides the key aspect of information for investor expressed in natural language or queries in natural language. Presented paper indicates the first step in knowledge base creation for investor showing how to apply the corresponding mathematical apparatus to cope with natural language statements in Black-Scholes Model for option pricing. |