|
|
Neuroevolution Principles and Applications
Herec, Jan ; Strnadel, Josef (referee) ; Bidlo, Michal (advisor)
The theoretical part of this work deals with evolutionary algorithms (EA), neural networks (NN) and their synthesis in the form of neuroevolution. From a practical point of view, the aim of the work is to show the application of neuroevolution on two different tasks. The first task is the evolutionary design of the convolutional neural network (CNN) architecture that would be able to classify handwritten digits (from the MNIST dataset) with a high accurancy. The second task is the evolutionary optimization of neurocontroller for a simulated Falcon 9 rocket landing. Both tasks are computationally demanding and therefore have been solved on a supercomputer. As a part of the first task, it was possible to design such architectures which, when properly trained, achieve an accuracy of 99.49%. It turned out that it is possible to automate the design of high-quality architectures with the use of neuroevolution. Within the second task, the neuro-controller weights have been optimized so that, for defined initial conditions, the model of the Falcon booster can successfully land. Neuroevolution succeeded in both tasks.
|
|
|
Hearthstone Counter-Deck Builder
Stachura, Šimon ; Gemrot, Jakub (advisor) ; Pilát, Martin (referee)
1 Title: Hearthstone Counter-Deck Builder Author: Šimon Stachura Department: Katedra softwaru a výuky informatiky Supervisor: Mgr. Jakub Gemrot, Ph. D. Abstract: Collecting cards and building decks out of them is the basic principle of collectible card games (such as Hearthstone). This task is usually very complex and requires players to think about a lot of factors, such as stability of deck's results or interactions among cards. The goal of this work is to try to make deckbuilding for Hearthstone automatic. Hill-climbing algorithm was used for this task. Generated decks were evaluated based on their winrate against chosen human-built actual decks from the game. Usage of hill-climbing brought a lot of questions - for instance, how to restrict the huge space of possible decks, what artifical intelligence to use for games' simulation, or how to make the simulation stable enough in such a non- deterministic environment. We have also tried to apply a new approach to a few of these problems. We have conducted two experiments to test our approach. Both experimentally created decks reached about 80 percent winrate against human-made decks. The results proved that even in such a nondeterministic environment hill- climbing is able to find interesting solutions. However, these solutions are highly dependent on...
|
|
|
Evolutionary algorithms
Bortel, Martin ; Karásek, Jan (referee) ; Lambertová, Petra (advisor)
Thesis describes main attributes and principles of Evolutionary and Genetic algorithms. Crossover, mutation and selection are described as well as termination options. There are examples of practical use of evolutionary and genetic algorithms. Optimization of distribution routes using PHP&MySQL and Google Maps API technologies.
|
|
|
Solution of Continuous Systems by Evolutionary Computational Techniques
Lang, Stanislav ; Šeda, Miloš (referee) ; Olehla, Miroslav (referee) ; Matoušek, Radomil (advisor)
The thesis deals the issue of solution of continuous systems by evolutionary computational techniques. Evolutionary computing techniques fall into the field of softcomputing, an advanced metaheuristics optimization that is becoming more and more a method of solving complicated optimization problems with the gradual increase in computing performance of computers. The solution of continuous systems, or the synthesis of continuous control circuits, is one of the areas where these advanced algorithms find their application. When dealing with continuous systems we will focus on regulatory issues. Evolutionary computing can then become a tool not only for optimization of controller parameters but also to design its structure. Various algorithms (genetic algorithm, differential evolution, etc.) can be used to optimize the parameters of the controller, for the design of the controller structurewe usually encounter so called grammatical evolution. However, the use of grammatical evolution is not necessary if appropriate coding is used, as suggested in the presented thesis. The thesis presents a method of designing the structure and parameters of a general linear controller using the genetic algorithm. A general linear regulator is known also as so called polynomial controller, if we encounter the polynomial theory of control. The method of encoding the description of the general linear controller into the genetic chain is crucial, it determines a set of algorithms that are usable for optimization and influence the efficiency of the calculations. Described coding, effective EVT implementation, including multi-criteria optimization, is a key benefit of this work.
|
|
|
Optimization Methods for SIMLIB/C++ Simulation Library
Chlebík, Jakub ; Janoušek, Vladimír (referee) ; Peringer, Petr (advisor)
This thesis addresses the topic of parametric optimization of simulation models. It introduces theoretical foundation of optimization and its uses in simulation analysis. Furthermore, it suggests the extension of SIMBLI/C++ library by module for optimization methods. Some of the chosen methods are then theoretically described, implemented in C++ language, demonstrates its uses and evaluates their success.
|
|
|
Evolutionary Design of Structures Using Rewriting Systems
Dobeš, Michal ; Zachariášová, Marcela (referee) ; Bidlo, Michal (advisor)
This work focuses on application of rewriting systems in the context of biology-inspired development of structures in evolutionary algorithms. As a part of this work, a method has been proposed for design of general structures that uses the aforementioned principle. This document includes an introduction into the theory of sorting networks and their design, the problem of which is later shown to be solvable using the proposed method. In addition, the theoretical background of evolutionary algorithms and rewriting systems is discussed. In the field of rewriting systems, an extension to the class of IL-systems has been created for use in the proposed method. The proposed method has been successfuly applied in the design of growing sorting networks. The results produced by the proposed method significantly outperform the results of the currently best-known comparable method.
|
|
|
Evolutionary Optimisation of Analogue Circuits
Mihulka, Tomáš ; Zachariášová, Marcela (referee) ; Bidlo, Michal (advisor)
The aim of this work was to create a system for optimisaton of specific analog circuits by evolution using multiple fitness functions . A set of experiments was run, and the results analyzed to evaluate the feasibility of evolutionary optimisation of analog circuits . A requirement for this goal is the study and choice of certain types of analog circuits and evolutionary algorithms . For the scope of this work , amplifiers and oscillators were chosen as target circuits , and genetic algorithms and evolutionary strategies as evolutionary algorithms . The motivation for this work is the ongoing effort to automate the design and optimisation of analog circuits , where evolutionary optimisation is one of the options .
|
|
|
Evolutionary algorithms for TV commercial planning
Vytasil, Jiří ; Pilát, Martin (advisor) ; Neruda, Roman (referee)
This thesis deals with the problem of TV commercials planning. This problem is getting more difficult as the number of different TV stations grows and naive algorithms become unable to scale with this number. In this work, we deal with the possibility to use evolutionary algorithms to solve this problem. The work also contains an implementation of a software, which is capable of testing the various versions of the algorithm and comparing them to a naive one. The results indicate that evolutionary algorithms are a suitable technique to solve the problem at hand. Powered by TCPDF (www.tcpdf.org)
|
|
|
Melody generation using a genetic algorithm
Helikar, Matouš ; Maršík, Ladislav (advisor) ; Křen, Tomáš (referee)
Music composition, as all other creative activities, requires original inspiration, which can also come from melodies generated by a computer. This thesis describes generation of music tracks represented by tree structures with pluggable modules that create or alter individual musical motives. The trees can subsequently be combined by a crossbreeding algorithm driven by user ratings. This results in music tracks evolving on multiple levels, such as the selected instruments or musical motives, rhythm and overall structure. Appropriate settings of parameters for the generator and constituent modules can then produce varied tracks for inspiration or relaxation. The thesis is accompanied by a complete application using these techniques for music generation and a user study of satisfaction with the resulting tracks. Powered by TCPDF (www.tcpdf.org)
|
|
|
Combination of Evolutionary Algorithms and Constraint Programming for Scheduling
Štola, Miroslav ; Pilát, Martin (advisor) ; Klíma, Martin (referee)
Scheduling problems and constraint satisfaction problems are generally known to be extremely hard. This thesis proposes a new evolutionary al- gorithm approach to solve a constrained-based scheduling problem. In this approach, variable orderings are evolved. The variable ordering serves as a parameter for the constraint solver. Its purpose is to determine the order in which variables are labelled by the solver. Hence the evolving individuals may be encoded as permutations. Therefore, our approach can be applied to a wider range of constraint satisfaction problems. Methods for generating the initial population of individuals based on the analysis of the precedence constraints graph are proposed. New genetic operators are presented and successfully applied. Our approach succeeded in finding a range of diverse schedules with the optimal makespan. Furthermore, multi-objective opti- mization was successfully attempted with the NSGA-II. 1
|
guest ::
