Input: A BN

Output: All the MI weights of the MI-QPN

Symbol Description:

●: an orphans node in a BN or a non-orphans one with parents

●: the prior probability distribution set of nodes

●: the CPT of with some of its parents

●: the set of conditional probability sets

Steps:

1. Obtain a topological sequence of all the nodes in the BN

2. for each node B₀ in the topological sequence do

3. if B₀ is a non-orphans node then

4. CT(B₀|B₁,…,B_n)¬get conditional probability (B₀, TP(B₁, …, B_n), CT(B₀|B₁, …, B_n))

5. Compute PT(B₀) with TC(B₀|B₁) and PT(B₁)

6. for i¬1 to n do

7. Compute MI(B₀→B_i) and MI(B_i®B₀) with TP(B₀|B₁,…,B_n) and CT(B₀|B₁,…, B_n).

8. end for

9. end if

10. end for

Get conditional probability (B₀, TP(B_k,…,B_m), CT(B₀|B₁, …, B_n))

1. if m

2. mid ¬ém-k/2ù

3. CT(B₀|B_k, …, B_mid)¬{P(B₀|B_k = s_k, …, B_mid = s_mid)|1 ≤ s₀ ≤ S₀, k ≤ i ≤ mid, 1 ≤ s_i ≤ S_i}

4. CT(B₀|B_mid+1, …, B_m)¬{P(B₀|B_mid+1=s_mid+1,…, B_n = s_m)|1 ≤ s₀ ≤ S₀, mid+1 ≤ i ≤ n, 1 ≤ s_i ≤ S_i}

5. TC(B₀|B_k, …, B_mid)¬get conditional probability(B₀, TP(B_k, …, B_mid), CT(B₀|B₁, …, B_mid))

6. TC(B₀|B_mid+1, …, B_m)¬get conditional probability(B₀, TP(B_mid+1, …, B_mid), CT(B₀|B_mid+1, …, B_m))

7. return TC(B₀|B_k, …, B_mid)ÈTC(B₀|B_mid+1, …, B_m)

8. end if

9. return CT(B₀|B_k)