Oct. 13 2009 2:00PM

Oct. 13 2009 4:00PM

One of major benefits of Bayesian decision making is its well understood and consistent theory. Direct application of the theory is often impossible to to implied analytical and computational difficulties. Therefore, many approximations has been developed to address this problem. The presence of approximations breaks the original theory into many branches and sub-variants, and thus cluttering the field with many new "methods" and "improvements".

BDM is an attempt how to unify this clutter in in software using a systematic way. The mathematical elements from the theory (such as probability densities, the Bayes rule, decision designers) are root classes of object oriented design. Each offspring of this root-class corresponds to one level of abstraction or approximation. Advanced estimation and control algorithms arise by appropriate combination of these elements. Since the elements are mutually interchangeable, wide range of methods can be designed and tested systematically.

Introduction to the system will be presented and its use will be demonstrated on simple examples.

Pravidelný seminář DAR na Katedře kybernetiky, FAV, ZČU v Plzni. (Kontaktní osoba: Miroslav Flídr)