Stuart Russell and Peter Norvig [Russell and Norvig 95] describe an algorithm
for updating a time-sliced belief network using two slices of the network.
The algorithm performs three operations prediction, rollup, and estimation.

I was wondering if a more detailed description or an implementation of this
algorithm is available. I would like also to hear from people who might have
tried implementing the algorithm and faced difficulties. We seem  to run into 
problems whenever there are common causes in an earlier slice. We have also 
problems when temporal arcs connect dependent nodes.

Thanks,
Ahmed Tawfik

[Russell and Norvig 95] Artificial Intelligence: A Modern Approach, Section 
17.5, Prentice Hall, 1995.

Reply via email to