Computer Science
Loading...
ISSN 1508-2806
e-ISSN: 2300-7036
Call number
Volume
Tom 2
Date
2000
Description
Journal
Computer Science
AGH University Press (2004-)
ISSN: 1508-2806 e-ISSN: 2300-7036
ISSN: 1508-2806 e-ISSN: 2300-7036
journal.volume.project
Contains
Journal Issues
Articles
Zastosowanie metod naturalnych w problemach poszukiwania optymalnego rozwiązania
(Wydawnictwa AGH, 2000) Jasińska-Suwada, Anna; Dzwinel, Witold; Rozmus, Krzysztof; Sołtysiak, Jacek
In the paper we present a new method, which can be used as a natural solver for searching the best solution in the multidimensional and multimodal parameter space. The method is based on a well-known simulation technique, i.e., molecular dynamics. To show advantages and disadvantages of the particle method in comparison to the standard genetic algorithm, we analyse efficiency of the methods in finding the global minimum of multi-dimensional and multi-modal test-bed functions and we calculate the evaluation indices. We analyse also the ways the solution space is explored and the parameters of algorithms adjusted. The optimal heuristics are proposed. The tests carried out show that the choice of the most appriopriate optimization method depends on type of a problem considered. We show that the particle method is more efficient for finding the optimal solution for multi-modal problems with distinct global extreme, while the genetic algorithm is better for deceptive functions with several locals extreme, which are placed far away from the global optimum. This comes from the different ways in which the particle method and genetic algorithm explore the solution space. The particle method can be used for initial analysis of functions, which character is unknown.
Wykorzystanie sieci neuronowych i analizy portfelowej w prognozowaniu kursów akcji na giełdzie papierów wartościowych
(Wydawnictwa AGH, 2000) Stochel, Przemysław
The article presents the use of neural networks in decision making process on the capital market. The author tried to show the efficiency of established solution in Polish reality which features different conditions in comparison with the markets of higher developed countries. The aim of the paper was to prove that neural networks are flexible tools which on one hand might be adjusted to investor's requirements and on the other, can reduce equirements to his experience. The article is based on the author's own research carried out by modelling neural network operation with a simulation program. The established solutions are input which employs stocks portfolio computed on the basis of Markowitz portfolio theory and Sharpe's model. According to the established propositions, the portfolio created in such a way is modified by neutral network in order to optimise a criterion which maximises the income of such a modified portfolio. A detailed genesis of the established input vector and network structure are presented. It allows the reader to carry out his own research and create his own attitude towards applied values. The research results based on a real stock market database with the use of one-output networks predicting the price of a single company - Agros as well as networks predicting the desirable structure of the whole portfolio are presented. The effect of the network structure learning parameters, input vector (not only as to the input quantity but also as to period of time they were collected) was examined. The dependence between the factors mentioned above such as input vector and network structure were discussed. It seems that the presented paper has proved that some not widely spread methods with neural networks can become a competitive tool to optimisation methods.
Dopasowanie krzywej tonalnej i równoważenie poziomów szarości przy kontrastowaniu i wstępnej wizualizacji struktur w analizie zdjęć medycznych
(Wydawnictwa AGH, 2000) Ogiela, Marek
The article presents a group of transformations which medical pictures can be subjected to, in the aim of enhancing their contrast and better distinguishing and diagnosis of organs shown in this pictures. In this work the analysis was based on images acquired by various techniques, such as roentgenography (RTG), magnetic resonance (MRI), classic computer tomography (CT) and ultrasonography (USG). Contrast expansion in these images was performed using the histogram method of balancing of gray levels and the method based on the modification of tonal curve. Such transformations were necessary during preliminary processing of these images, especially to adjust their quality for next steps of the analysis in case of automatic medical diagnosis, and also to enable their visual evaluation by specialists and diagnostic interpretation of images of organs and their structural features. The studies show that the method based on the modification of tonal curve yields better results in case of images acquired by roentgenography and magnetic resonance imaging. On the other hand, histogram method is more suitable for contrast expansion in CT and USG images.
Obliczenia metodą elementów skończonych na sieci stacji roboczych i na komputerze równoległym
(Wydawnictwa AGH, 2000) Chrobak, Renata
This paper presents three parallel algorithms for simulation of axisymmetrical metal forging processes. The algorithms are based on the non-overlapping domain decomposition method where a mesh of elements is divided into layers assigned to different processes. Parallel programs implementing these algorithms are written in C using PVM calls. The efficiency of the parallel programs is verified on networked IBM RS/6000, Sun SPARCstations and HP 9000 workstations as well as on HP/Convex Exemplar SPP1000.
Formalna weryfikacja wymagań systemów czasu rzeczywistego
(Wydawnictwa AGH, 2000) Szpyrka, Marcin
The methodology of system requirements verification presented in this paper is a proposition of a practical procedure for reducing some negatives of the specification of requirements. The main problem that is considered is to create a complete description of the system requirements without any negatives. Verification of the initially defined requirements is based on the coloured Petri nets. Those nets are useful for testing some properties of system requirements such as completeness, consistency and optimality. An example of the lift controller is presented.