Why CVE Does More Than Representing Factors As TreesContextual Variable EliminationOnesMulti-Valued Variables

Multi-Valued Variables

We have presented an algorithm that allows for multi-valued variables, where the splitting operation creates the same number of confactors as there are values in the domain of the variable being split.

There is an alternate method for using multi-valued variables. This is to extend the notion of a context to allow for membership in a set of values. That is, a context could be a conjunction of terms of the form X in S where S is a set of values. The original version of equality of the form X=v is the same as X in {v}. Splitting can be done more efficiently as there is only one residual confactor for each split. Effectively we treat a multiset of confactors as a unit.

There are examples where this representation can be much more efficient, but it makes the algorithm much more complicated to explain. There are also examples where the binary splitting is less efficient as it needs more splits to get the same result.


David Poole and Nevin Lianwen Zhang,Exploiting Contextual Independence In Probabilistic Inference, Journal of Artificial Intelligence Research, 18, 2003, 263-313.

Why CVE Does More Than Representing Factors As TreesContextual Variable EliminationOnesMulti-Valued Variables