The three steps are executed repeatedly until a TLC with certainty level higher than the preset threshold h so that we can claim the TLC is reached, or all UOAs are asked and no TLC reaches the threshold.
Recall that [A.sub.1], [A.sub.2], and [A.sub.3] are UOAs, and [A.sub.4] and [A.sub.5] are TLCs.
The two influential UOAs on [A.sub.4] are [A.sub.1] and [A.sub.2] (referring to Fig.
An assertion is called unconfirmed observable assertion (UOA) if it is observable and its value is not yet known.
If [A.sub.t] is a TLC and [A.sub.u] is a UOA, then [CI.sub.U[right arrow]T] shows that in case the UOA [A.sub.u]'s value is true, how much it would contribute to inferring that TLC [A.sub.t] is true.
That is just the property we want the IG criterion to possess, --It would select such a UOA that once its value is known it would help confirm a TLC to a large extent.
It first calculates the IG-index for each UOA, [A.sub.i], associated with the TLC [A.sub.t] selected in Step 1.
The value of P([A.sub.i]) is given in the network since [A.sub.i] is a UOA. P([A.sub.t]) was calculated in Step 1.
The UOA with the largest IG-index absolute value is picked as the key missing data, and its value is pursued from an information source such as a sensor or the user.