Divide and conquer techniques are widespread in computer science, but have not been explored thoroughly in graphical models. A method for splitting up a Bayesian network into components that can be treated independently with respect to many of the common operations has been constructed . This method forms the foundation for various tasks such as distributed triangulation and hybrid inference. Special emphasis has been directed towards the development of algorithms for incremental compilation.
|Effective start/end date||19/05/2010 → …|
- <ingen navn>