|
|
Improving evolutionary algorithms using probability mixture models
Bajer, Lukáš ; Holeňa, Martin (advisor) ; Kovářík, Oleg (referee)
Evolutionary, and especially genetic algorithms have become one of the most successful methods for the optimization of empirical objective functions. However, in many engineering applications, evaluation of the empirical fitness function can be very time consuming or cost a considerable amount of money. In this article, we employ a surrogate model of the original fitness function which serves as a fast approximation whenever needed. First, we intended to use finite mixture models, but radial basis function networks was finally used as a particular surrogate model because of implementability. With this method, much larger populations or several generations can be simulated without waiting for expensive objective function evaluation. As a result, faster convergence in terms of the number of the original empirical fitness evaluations is achieved.
|
|
|
Procházky v grafech a genetické algoritmy
Szépe, Peter ; Pangrác, Ondřej (advisor) ; Bajer, Lukáš (referee)
Title: Walks in graphs and genetic algorithms Author: Peter Szépe Department: Department of Applied Mathematics Supervisor: RNDr. Ondřej Pangrác, Ph.D. Supervisor's e-mail address: pangrac@kam.mff.cuni.cz Abstract: We solve the problem of finding a maximal walk from the starting vertex to the target vertex with an upper bound of the length of the walk. It is an NP-hard problem where not even the approximation algorithms do guarantee a quick and nice solution. Hence it is important to develop heuristics for practical applications. Keywords: optimization, evolutionary algorithms, genetic algorithms, graphs, walks in graphs
|
|
|
Movement in project ENTs
Bajer, Lukáš ; Brom, Cyril (advisor) ; Šerý, Ondřej (referee)
The ENTS project is a simulator of an environment which is similar to our common world. In this environment, there live autonomous agents called ents. They take care of their world. To fulfil their goals and satisfy their daily needs, they often have to look for a path around the world. This work is focused on scripts which are responsible for this pathfinding. The ents' movements are improved by hierarchical version of the A* algorithm. Thanks to this, demands on CPU during looking for longer paths are considerably decreased. In addition, ents' scripts are enhanced by better movement around a room, other ents following and avoiding, and "lazy" picking up objects.
|
|
|
Model-based evolutionary optimization methods
Bajer, Lukáš ; Holeňa, Martin (advisor) ; Brockhoff, Dimo (referee) ; Pošík, Petr (referee)
Model-based black-box optimization is a topic that has been intensively studied both in academia and industry. Especially real-world optimization tasks are often characterized by expensive or time-demanding objective functions for which statistical models can save resources or speed-up the optimization. Each of three parts of the thesis concerns one such model: first, copulas are used instead of a graphical model in estimation of distribution algorithms, second, RBF networks serve as surrogate models in mixed-variable genetic algorithms, and third, Gaussian processes are employed in Bayesian optimization algorithms as a sampling model and in the Covariance matrix adaptation Evolutionary strategy (CMA-ES) as a surrogate model. The last combination, described in the core part of the thesis, resulted in the Doubly trained surrogate CMA-ES (DTS-CMA-ES). This algorithm uses the uncertainty prediction of a Gaussian process for selecting only a part of the CMA-ES population for evaluation with the expensive objective function while the mean prediction is used for the rest. The DTS-CMA-ES improves upon the state-of-the-art surrogate continuous optimizers in several benchmark tests.
|
|
|
Pathophysiology of inflammatory bowel disease. Relation to primary scklerosing cholangitis, liver transplantation and carcinogenesis.
Bajer, Lukáš ; Drastich, Pavel (advisor) ; Živný, Jan (referee) ; Procházka, Vlastimil (referee)
Inflammatory bowel disease (IBD) represents a group of multifactorial illnesses with increasing incidence worldwide. Crohn's disease (CD) and ulcerative colitis (UC) are the two most thoroughly defined phenotypes of IBD. IBD associated with primary sclerosing cholangitis (PSC) - a progressive biliary disease leading to cirrhosis and liver failure - is considered as specific IBD phenotype (also referred to as 'PSC - IBD') due to its clinical and pathophysiological characteristics. The aim of the experimental part of this thesis was to define specific features of PSC - IBD in the key areas of IBD pathogenesis. These are: microbiota composition, gut - barrier failure, genetic predisposition and aberrant cellular and antibody immune response. Furthermore, the other goals were to describe relation of IBD status and activity to liver transplantation (LTx) and carcinogenesis based on thorough analysis of clinical data in patients under surveillance at the liver transplantation unit. Using the next-generation parallel sequencing technology, we discovered specific bacterial and mycobial features of gut microbiota composition in PSC - IBD which significantly differed from UC and healthy controls recruited from Czech general population. Moreover, we identified numerous seral biomarkers distinguishing CD, UC...
|
|
|
Model-based evolutionary optimization methods
Bajer, Lukáš ; Holeňa, Martin (advisor) ; Brockhoff, Dimo (referee) ; Pošík, Petr (referee)
Model-based black-box optimization is a topic that has been intensively studied both in academia and industry. Especially real-world optimization tasks are often characterized by expensive or time-demanding objective functions for which statistical models can save resources or speed-up the optimization. Each of three parts of the thesis concerns one such model: first, copulas are used instead of a graphical model in estimation of distribution algorithms, second, RBF networks serve as surrogate models in mixed-variable genetic algorithms, and third, Gaussian processes are employed in Bayesian optimization algorithms as a sampling model and in the Covariance matrix adaptation Evolutionary strategy (CMA-ES) as a surrogate model. The last combination, described in the core part of the thesis, resulted in the Doubly trained surrogate CMA-ES (DTS-CMA-ES). This algorithm uses the uncertainty prediction of a Gaussian process for selecting only a part of the CMA-ES population for evaluation with the expensive objective function while the mean prediction is used for the rest. The DTS-CMA-ES improves upon the state-of-the-art surrogate continuous optimizers in several benchmark tests.
|
|
|
Movement in project ENTs
Bajer, Lukáš ; Šerý, Ondřej (referee) ; Brom, Cyril (advisor)
The ENTS project is a simulator of an environment which is similar to our common world. In this environment, there live autonomous agents called ents. They take care of their world. To fulfil their goals and satisfy their daily needs, they often have to look for a path around the world. This work is focused on scripts which are responsible for this pathfinding. The ents' movements are improved by hierarchical version of the A* algorithm. Thanks to this, demands on CPU during looking for longer paths are considerably decreased. In addition, ents' scripts are enhanced by better movement around a room, other ents following and avoiding, and "lazy" picking up objects.
|
|
|
Improving evolutionary algorithms using probability mixture models
Bajer, Lukáš ; Kovářík, Oleg (referee) ; Holeňa, Martin (advisor)
Evolutionary, and especially genetic algorithms have become one of the most successful methods for the optimization of empirical objective functions. However, in many engineering applications, evaluation of the empirical fitness function can be very time consuming or cost a considerable amount of money. In this article, we employ a surrogate model of the original fitness function which serves as a fast approximation whenever needed. First, we intended to use finite mixture models, but radial basis function networks was finally used as a particular surrogate model because of implementability. With this method, much larger populations or several generations can be simulated without waiting for expensive objective function evaluation. As a result, faster convergence in terms of the number of the original empirical fitness evaluations is achieved.
|
|
|
Procházky v grafech a genetické algoritmy
Szépe, Peter ; Bajer, Lukáš (referee) ; Pangrác, Ondřej (advisor)
Title: Walks in graphs and genetic algorithms Author: Peter Szépe Department: Department of Applied Mathematics Supervisor: RNDr. Ondřej Pangrác, Ph.D. Supervisor's e-mail address: pangrac@kam.mff.cuni.cz Abstract: We solve the problem of finding a maximal walk from the starting vertex to the target vertex with an upper bound of the length of the walk. It is an NP-hard problem where not even the approximation algorithms do guarantee a quick and nice solution. Hence it is important to develop heuristics for practical applications. Keywords: optimization, evolutionary algorithms, genetic algorithms, graphs, walks in graphs
|
|
| |
guest ::
