|
|
Image registration of ultrasound sequences using evolutionary algorithms
Hnízdilová, Bohdana ; Odstrčilík, Jan (referee) ; Mézl, Martin (advisor)
This master´s thesis deals with the registration of ultrasound sequences using evolutionary algorithms. The theoretical part of the thesis describes the process of image registration and its optimalization using genetic and metaheuristic algorithms. The thesis also presents problems that may occur during the registration of ultrasonographic images and various approaches to their registration. In the practical part of the work, several optimization methods for the registration of a number of sequences were implemented and compared.
|
|
|
Ant colony
Hart, Pavel ; Hynčica, Ondřej (referee) ; Honzík, Petr (advisor)
First part of the thesis is about literature research of optimization algorithms. Three of the algorithms were implemented and tested, concretely the ant colony algorithm, tabu search and simulated annealing. All three algorithms were implemented to solve the traveling salesman problem. In second part of the thesis the algorithms were tested and compared. In last part the influence of the ant colony parameters was evaluated.
|
|
|
Evolutionary algorithms
Haupt, Daniel ; Polách, Petr (referee) ; Honzík, Petr (advisor)
The first part of this work deals with the optimization and evolutionary algorithms which are used as a tool to solve complex optimization problems. The discussed algorithms are Differential Evolution, Genetic Algorithm, Simulated Annealing and deterministic non-evolutionary algorithm Taboo Search.. Consequently the discussion is held on the issue of testing the optimization algorithms through the use of the test function gallery and comparison solution all algorithms on Travelling salesman problem. In the second part of this work all above mentioned optimization algorithms are tested on 11 test functions and on three models of placement cities in Travelling salesman problem. Firstly, the experiments are carried out with unlimited number of accesses to the fitness function and secondly with limited number of accesses to the fitness function. All the data are processed statistically and graphically.
|
|
| |
|
|
Video stabilization using global optimization algorithms
Bartoš, Patrik ; Říha, Kamil (referee) ; Kříž, Petr (advisor)
This bachelor thesis focuses on video stabilization using CRS (Controlled Random Search) and GA (Genetic Algorithm) optimization algorithms. It describes registration process, geometrical transformations, interpolation methods, similarity criteria and optimization algorithms. It also briefly describes structure of the program created in MATLAB. Finally it contains results of achieved stabilization.
|
|
|
Evolutionary Optimization of Convolutional Neural Networks
Čoupek, Vojtěch ; Mrázek, Vojtěch (referee) ; Sekanina, Lukáš (advisor)
This thesis deals with the problem of neural network weights compression using the technique of Weight-Sharing and parameter optimization of this technique by unconventional optimization algorithms. The reason for the optimization is decreasing the memory or energy demands of the neural network response calculation. The aim is to design a system that accepts a neural network and reduces its memory demands. Its functionality is demonstrated with the help of several experiments. The thesis investigates the use of various optimization algorithms, additional compression using the quantization above the Weight-Sharing technique, and proposes the quantization results tuning method to improve accuracy. These procedures are first tested on the Le-Net-5 network and then applied for the MobileNet\_v2. network compression.
|
|
| |
|
|
The Role of Advanced Option Pricing Techniques Empirical Tests on Neural Networks
Brejcha, Jiří ; Baruník, Jozef (advisor) ; Vošvrda, Miloslav (referee)
This thesis concerns with a comparison of two advanced option-pricing techniques applied on European-style DAX index options. Specifically, the study examines the performance of both the stochastic volatility model based on asymmetric nonlinear GARCH, which was proposed by Heston and Nandi (2000), and the artificial neural network, where the conventional Black-Scholes-Merton model serves as a benchmark. These option-pricing models are tested with the use of the dataset covering the period 3rd July 2006 - 30th October 2009 as well as of its two subsets labelled as "before crisis" and "in crisis" data where the breakthrough day is the 17th March 2008. Finding the most appropriate option-pricing method for the whole periods as well as for both the "before crisis" and the "in crisis" datasets is the main focus of this work. The first two chapters introduce core issues involved in option pricing, while the subsequent third section provides a theoretical background related to all of above-mentioned pricing methods. At the same time, the reader is provided with an overview of the theoretical frameworks of various nonlinear optimization techniques, i.e. descent gradient, quassi-Newton method, Backpropagation and Levenberg-Marquardt algorithm. The empirical part of the thesis then shows that none of the...
|
|
|
On the predictibility of Central European stock returns: Do Neural Networks outperform modern economic techniques?
Baruník, Jozef ; Žikeš, Filip (advisor) ; Vošvrda, Miloslav (referee)
In this thesis we apply neural networks as nonparametric and nonlinear methods to the Central European stock markets returns (Czech, Polish, Hungarian and German) modelling. In the first two chapters we define prediction task and link the classical econometric analysis to neural networks. We also present optimization methods which will be used in the tests, conjugate gradient, Levenberg-Marquardt, and evolutionary search method. Further on, we present statistical methods for comparing the predictive accuracy of the non-nested models, as well as economic significance measures. In the empirical tests we first show the power of neural networks on Mackey-Glass chaotic time series followed by real-world data of the daily and weekly returns of mentioned stock exchanges for the 2000:2006 period. We find neural networks to have significantly lower prediction error than classical models for daily DAX series, weekly PX50 and BUX series. The lags of time-series were used, and also cross-country predictability has been tested, but the results were not significantly different. We also achieved economic significance of predictions with both daily and weekly PX-50, BUX and DAX with 60% accuracy of prediction. Finally we use neural network to learn Black-Scholes model and compared the pricing errors of...
|
|
|
Image registration of ultrasound sequences using evolutionary algorithms
Hnízdilová, Bohdana ; Odstrčilík, Jan (referee) ; Mézl, Martin (advisor)
This master´s thesis deals with the registration of ultrasound sequences using evolutionary algorithms. The theoretical part of the thesis describes the process of image registration and its optimalization using genetic and metaheuristic algorithms. The thesis also presents problems that may occur during the registration of ultrasonographic images and various approaches to their registration. In the practical part of the work, several optimization methods for the registration of a number of sequences were implemented and compared.
|
guest ::
