Topics for PhD Theses
Abstract Algebraic Logic(Petr Cintula)
Substructural Logics and Residuated Lattices(Petr Cintula Rostislav Horčík, Zuzana Haniková)
Deep and shallow multilayer neural networks(Věra Kůrková)The goal of the work is to compare learning algorithms and complexity of neural networks having one and more than one hidden layer.
Generalization in neural-network learning(Věra Kůrková)The goal of the work is theoretical and experimental comparison of various approaches to generalization in neural network learning. For example, generalization modelled using various types of regularization and generalization based on output-weight minimization.
Classification Trees and Forests(Petr Savický)Classification trees and their ensembles, which are called classification forests, are a suitable method of a prediction of an unknown class on the basis of known numerical or categorial attributes for complex distributions, if a sufficient amount of training cases is available. The typical methods of this type are CART, C4.5, Random Forests. The goal of the work is to investigate the properties of the methods and their modifications.
Complexity Measures of Neural Networks(Jiří Šíma)
Modern Methods of Evolutionary Optimization(Martin Holeňa)
Modern Regression Methods(Martin Holeňa)