The Algorithm

nFor
each potential query **q
**

nCompute
DRisk(**X**|**q**)

nChoose
**q** for which DRisk(**X**|**q**)
is greatest

nCost
of computing DRisk(**X**|**q**):

nCost of
Bayesian network inference

nComplexity:
*O *(|**Q**|. Cost of inference)