National Repository of Grey Literature 1 records found  Search took 0.00 seconds. 
Approximate solution of Unconstrained influence diagrams
Fried, Vojtěch ; Vomlelová, Marta (advisor) ; Studený, Milan (referee)
We give an introduction to the theory of probabilistic graphical models and describe several types of them (Bayesian Networks - BN, Influence Diagrams - ID, Unconstrained Influence Diagrams - UID). Unconstrained Influence Diagrams support the possibility for the user to choose the ordering of decisions based on observations. This increases the expressive power of UIDs compared to IDs but makes it harder to find an optimal solution. It is often impossible to find an optimal solution because of exponential complexity increase compared to IDs. Therefore we design and investigate several approximate methods to solve UIDs. The result of these methods is an ordinary ID created from the former UID by adding edges. The optimal solution of the ID should be as close to the original UID as possible. Heuristical methods represent one type of the methods investigated. Heuristical methods use a simplification of the optimal algorithm. During the run of the algorithm heuristics are used to cut off the branches that are not perspective for further calculation. Another type of methods is to create the ID directly. We evaluate our methods experimentally based on randomly generated UIDs of three types and compare their performance namely to the optimal solution and to equally complex random methods.

Interested in being notified about new results for this query?
Subscribe to the RSS feed.