|
|
Image Reconstruction in Electrical Impedance Tomography through Multilayer Perceptron
Kouakouo Nomvussi, Serge Ayme ; Mikulka, Jan
This study introduces a novel image reconstruction algorithm designed to excel in challenging scenarios with noisy datasets. Comparative evaluations against established methods, the Total Variation technique and the Gauss-Newton algorithm, are conducted using key performance metrics including the correlation coefficient and structural similarity index. The Results demonstrate that the proposed algorithm displays variable performance in noise-free data compared to Total Variation but consistently outperforms it in the presence of noise. Furthermore, when contrasted with the Gauss-Newton algorithm, the proposed method consistently exhibits superior outcomes, particularly in scenarios involving noisy datasets, where the Gauss-Newton algorithm faces limitations. This study underscores the robustness of the proposed algorithm in noisy conditions, suggesting its potential for applications where accurate image reconstruction is critical.
|
|
|
Automated Representation Learning for Cartesian Genetic Programming Using Neural Networks
Koči, Martin ; Mrázek, Vojtěch (referee) ; Sekanina, Lukáš (advisor)
This master's thesis addresses the integration of neural networks and Cartesian Genetic Programming (CGP). It explores the use of neural networks for automated representation creation for CGP and their application to improve the evolutionary process in CGP. The study covers basic concepts of machine learning, including various types of learning and neural network models. It also touches on evolutionary algorithms with an emphasis on their basic principles, general algorithms, and types of representations. This work also includes principles of representation learning and two fundamental architectures for their creation. It describes the subsequent use of representation learning in genetic programming. The solution design includes data acquisition and preprocessing, representation creation processes, and the utilization of the resulting representations. The thesis also implements two new approaches for creating representations for Cartesian genetic programs. It further explores their use in two new mutation operators, where one is based on direct modification of the vector representation and the other on the selection of genes for mutation based on their similarity. The last of the explored areas is predicting the suitability of candidate solutions using newly emerged representations.
|
|
|
Neural Network Based Edge Detection
Křepský, Jan ; Grézl, František (referee) ; Švub, Miroslav (advisor)
Utilization of artificial neural networks in digital image processing is nothing new. The aim of this work is to design and implement neural network based edge detector and learn how effective this approach is for edge detection in images and to compare these results with common detectors. In theoretical part of my work I describe some methods of image pre-processing, common approach to edge detection and their thinning and I try to introduce basics for understanding artificial neural networks theory.
|
|
|
Elliot Wave Detection
Kaleta, Marek ; Šperka, Svatopluk (referee) ; Petřík, Patrik (advisor)
This work deals with Elliott wave detection, which are statistical tool used to describe financial makret cycles and predict market trends. The work proposes methods to detect Elliott Waves and evaluetes them. From several methods of Elliott wave detection, Committee machines of multilayer perceptrons are used. Result of this work is a program, which detect Elliott impulse waves on input signal and builds hierarchy of Elliott waves.
|
|
|
Comparison of Libraries of Artificial Neural Networks
Dohnal, Zdeněk ; Zbořil, František (referee) ; Dalecký, Štěpán (advisor)
This thesis is about comparison of libraries of artificial neural networks. Basic theory of neuron, neural networks and their learning algorithms are explained here. Multilayer perceptron, Self organizing map and Hopfield net are chosen for experiments. Criteria of comparison such as licence, community or last actualization are designed. Approximation of function, association and clustering are chosen as task for experiments. After that, there is implementation of applications using chosen libraries. At the end, result of comparison and experiment are evaluated.
|
|
|
Artificial neural network for modeling electromagnetic fields in a car
Kostka, Filip ; Škvor, Zbyněk (referee) ; Raida, Zbyněk (advisor)
The project deals with artificial neural networks. After designing and debugging the test data set and the training sample set, we created a multilayer perceptron network in the Neural NetworkToolbox (NNT) of Matlab. When creating networks, we used different training algorithms and algorithms improving the generalization of the network. When creating a radial basis network, we did not use the NNT, but a specific source code in Matlab was written. Functionality of neural networks was tested on simple training and testing patterns. Realistic training data were obtained by the simulation of twelve monoconic antennas operating in the frequency range from 2 to 6 GHz. Antennas were located inside a mathematical model of Octavia II. Using CST simulations, electromagnetic fields in a car were obtained. Trained networks are described by regressive characteristics andthe mean square error of training. Algorithms improving generalization are applied on the created and trained networks. The performance of individual networks is mutually compared.
|
|
|
Genetic Programming in Prediction Tasks
Machač, Michal ; Mrázek, Vojtěch (referee) ; Sekanina, Lukáš (advisor)
This thesis introduces various machine learning algorithms which can be used in prediction tasks based on regression. Tree genetic programming and linear genetic programming are explained more thoroughly. Selected machine learning algorithms (linear regression, random forest, multilayer perceptron and tree genetic programming) are compared on publicly available datasets with the use of scikit-learn and gplearn libraries. A core part of this project is a new implementation of linear genetic programming which was developed in C++, tested on common symbolic regression problems and then evaluated on real datasets. Results obtained with the proposed system are compared with the results obtained with gplearn.
|
|
|
Recurrent Neural Networks in Computer Vision
Křepský, Jan ; Řezníček, Ivo (referee) ; Španěl, Michal (advisor)
The thesis concentrates on using recurrent neural networks in computer vision. The theoretical part describes the basic knowledge about artificial neural networks with focus on a recurrent architecture. There are presented some of possible applications of the recurrent neural networks which could be used for a solution of real problems. The practical part concentrates on face recognition from an image sequence using the Elman simple recurrent network. For training there are used the backpropagation and backpropagation through time algorithms.
|
|
|
Genetic Programming in Prediction Tasks
Machač, Michal ; Mrázek, Vojtěch (referee) ; Sekanina, Lukáš (advisor)
This thesis introduces various machine learning algorithms which can be used in prediction tasks based on regression. Tree genetic programming and linear genetic programming are explained more thoroughly. Selected machine learning algorithms (linear regression, random forest, multilayer perceptron and tree genetic programming) are compared on publicly available datasets with the use of scikit-learn and gplearn libraries. A core part of this project is a new implementation of linear genetic programming which was developed in C++, tested on common symbolic regression problems and then evaluated on real datasets. Results obtained with the proposed system are compared with the results obtained with gplearn.
|
|
|
Neural Network Library and Editor
Rouček, Martin ; Ježek, Pavel (advisor) ; Pešková, Klára (referee)
Neural network models are more often used in desktop applications given the increasing speed of computers. A very widespread platform for writing desktop applicatons is .NET Framework. Nevertheless, there is no neural networks library for the .NET Framework platform with a simple API and the possibility to work with library objects in a graphical interface. The author decided to create such a library. The main part of the thesis is a neural networks library GNNL that is initially limited to implementing two frequently used neural networks models which are a multilayer perceptron and self- organizing map together with learning algorithms of backpropagation and competitive learning. Graphical support of the library GNNL consists of a library GNNLV and neural network editor. The Library GNNLV contains the controls that allow working with GNNL library objects and a programmer can use them in his or hers application. The Neural network editor enables the programmer to create a neural network in a graphical interface, train it, analyze it, save it, and later use it in different applications. Text of the thesis focuses on analyzing and describing the implementation of the library with its graphical support. A major component of the text is a summary of neural networks theory for laics or programmers using library...
|
guest ::
