Pravděpodobnostní modely v hudební kompozici

Abstract

This thesis deals with the foundations of homogeneous Markov chains with discrete time and finite number of states. The text also discusses the Bayesian approach, in particular the Bayesian prediction of a categorical random variable with Dirichlet prior distribution and its extension to homogeneous indecomposable Markov chains. The described apparatus is then applied to the generation of a simple melodic line as a Markov chain.