|
|
Multilayer hierarchical models
Béger, Michal ; Štanclová, Jana (advisor) ; Petříčková, Zuzana (referee)
This diploma thesis deals with hierarchical associative memories (HAM), which have been experimentally analysed only in the case of two layer hierarchy so far. The aim of this thesis is to study existing hierarchical models and evaluate experimentally their performance for more than two layer hierarchy. We show, that existing hierarchical model HAM is not suitable for three or more layer hierarchy. For that reason, we propose a new version of hierarchical model (called HAM-N), which enables utilization of any number of layers. The new model HAM-N uses the structure of the HAM model. However, due to modied learning and recall process, the HAM-N model eliminates the above-mentioned drawbacks of the HAM model. Finally, the HAM-N model is experimentally studied with respect to processing of large amounts of correlated patterns. Thesis also includes analysis of experiment results.
|
|
|
Multilayer Perceptron Learning
Brambora, Tomáš ; Rydvan, Pavel (advisor) ; Štanclová, Jana (referee)
Multilayer perceptron networks are interesting alternative to the classical von neuman computational models. This thesis summarizes theoretical basis of their learning and describes the implemenatation of a multilayer pereptron network, designed as an agent for multiagent system BANG3. At the end of the thesis, we summarize the results of testing the implemented multilayer perceptron network on two data sets using different learning algorithms with various parameter configurations.
|
|
|
Using Neural Networks in the Hierachical Routing
Straka, Martin ; Štanclová, Jana (advisor) ; Lokoč, Jakub (referee)
The objective of the diploma thesis is to show possible utilization of the neural networks for the partitioning needs of the hierarchical routing algorithm. The work proposes a hierarchical approach, which can be useful for the optimal path searching process. We focus on the application of several neural network models to extract hierarchical information from the transportation network data. Introduced models are based on the energy minimization principle and we demonstrate an employment of the deterministic annealing methods (MFA) as quite ambitious in the partitioning process. The experimental part of this work makes use of our findings and propose several suggestions on the proper parametrization of employed neural network models. In the experimental tests, we demonstrate capabilities of the MFA to provide a partitioning task also in case of lack of the global information.
|
|
|
Recognition of trafic signs in pictures
Hrinčár, Matej ; Patočka, Mikuláš (advisor) ; Štanclová, Jana (referee)
The aim of the following document is to create a program that will be able (preferably in the real time) to recognize the traffic signs on the pictures. The program should recognize the sign on the condition that no part of it is covered by any object and the sign is turned in a reasonable angle. Any background or objects of other colors should not confuse the program. The document solves two problems. The first one is finding the sign on the input image, the second one is recognizing the sign. For solving the first problem the color segmentation is used, which means the extraction of the needed color from the image. In this new image the program finds shapes similar to the traffic signs. Artificial neural network is used to recognize the sign.
|
|
|
Optical Character Recognizer
Klaučo, Matej ; Lánský, Jan (advisor) ; Štanclová, Jana (referee)
The aim of this work is to implement a conversion from the scanned text document to the editable form. In this work we closely analyse this problem, its division into smaller parts and their solutions. An important part of the work is also a creation of a set of complex applications with a sophisticated graphical user interface, which allow accomodation of conversion to the user's requirements. During the conversion we specialize in images, which may contain insignificiant noise caused by a scanner of poor quality and which can be rotated.
|
|
|
Programming language and IDE for Lego Mindstorms NXT
Pelc, Jan ; Štanclová, Jana (advisor) ; Bureš, Tomáš (referee)
Title: Programming language and IDE for Lego Mindstorms NXT Author: Jan Pelc Department: Department of Software Engineering Supervisor: RNDr. Jana Štanclová, Ph.D. Supervisor's e-mail address: Jana.Stanclova@ruk.cuni.cz Abstract: The work is focused on design and implementation of a classical programming language for the standard firmware of robotics platform Lego Mindstorms NXT, with respect to less experienced programmers. The work contains a brief description of the target platform and the communication interface between the controller unit and a PC, and an overview of available tools for creating programs for this platform. In the last part, our library for communication with the controller unit is described, and possibilities for remote debugging of programs running on this unit are analysed. Keywords: robotics, Lego Mindstorms NXT, programming language design, compiler implementation, remote debugging
|
|
|
Hand Recognition
Malovec, Róbert ; Štanclová, Jana (advisor) ; Hoksza, David (referee)
The research object of this diploma thesis is hand recognition problematics. The first part of this work is dedicated to palm photograph processing and extraction of characteristic features. Extracted features are used as inputs of investigated classification methods. The second part of this work provides detailed description of explored methods and their modifications, which are the research objects. Diploma thesis includes description and experimental results of minimum distance classifier, Bayes naive classifier and method based on neural networks. The aim of this work is the comparison of the particular methods with respect to classification success and time needed for classifier construction and classification of all available patterns. The main approach is to choose the best method, which solves the problem of hand recognition according to our experimental results.
|
|
|
Image Viewing and Manipulation Tool
Hauzar, David ; Kalibera, Tomáš (advisor) ; Štanclová, Jana (referee)
Image processing comprises many useful techniques for fixing and correcting of digital photographs, such as noise filtering, sharpening of images, color balancing, and many others. The aim of the work is to design and implement a portable tool that would allow easy integration of existing implementations of such techniques, providing its users with a unified and easy to use interface. The tool offers basic functions for image browsing, viewing, and processing. The advanced functions include image bending - manual combining of photographs of the same object taken with different exposures into a single photograph with higher dynamic range. The tool makes it possible to apply some of the operations to a group of images. These operations are image rotation, image rescaling, and median filtering. The extension mechanism of the tool includes support for adding new image processing operations, applying operations to a group of images, extending the range of supported image formats, modifying and extending user interface of the program.
|
|
|
Feed-forward neural networks and their application in data mining
Civín, Lukáš ; Mrázová, Iveta (advisor) ; Štanclová, Jana (referee)
The goal of data mining is to solve various problems dealing with knowledge extraction from huge amounts of real-world data, the quality of which might be disputable. Neural networks can help with the solution due to their generalization capabilities. While working on data mining projects, we have essentially the following two objectives in real-world applications of feed-forward neural networks. To obtain applicable results, it is crucial to provide the networks with well-prepared data. However, it is equally important to choose the right training strategy for the networks themselves - including network architecture, parameter settings or the training algorithm. One of the most important ideas behind these steps is namely to prevent "over-training". The final network should recall unknown examples as well as possible. There are plenty of techniques with different approaches to the solution. It is possible to modify the data, these comprises modifying the range of the data or its dimension, adding noise to the data, etc. Yet another way is the modification of the neural network by structural learning with forgetting, weight decay or early stopping. These techniques are analyzes both theoretically and experimentally in this thesis. With regard to the results achieved in a number of experimental tests we have...
|
|
|
An information system for sport schools
Mach, Jiří ; Bednárek, David (advisor) ; Štanclová, Jana (referee)
The most important part of this work is to implement an information system so that it meets all the basic needs for the management of sports schools. The system shall contain the catalogues of students, lecturers, facilities, arranged courses and applications. Three groups of users: administrators, teachers and pupils have an access to the entire program. For this reason, the work is created as a web application. This text contains especially an user guide and a programmer documentation.
|
guest ::
