Aplikace/Elektronický dokument

Jiroušek R., Kratochvíl V.

Rozpoznávání a modelování vícerozměrných dat

2010

MUDIM is a system for MUlti DImensional compositional Models. This new approach for probability distribution representation and processing is based on the idea that a multidimensional distribution is computed - composed - from a system of oligodimensional distributions by iterative application of a special operator of composition.The purpose of compositional models is similar to graphical Markov models, namely Bayesian networks. However, compositional models are not so much connected with a danger of misconstruction inflicted by misleading (causal) interpretation of directed acyclic graphs.