Algorithm 3.

RB-I (D, n)

Input: dataset D, number of rounds n

Output: the most probable PAG PAG-I

1: Let $\mathcal{G}$ and r be empty sets

2: $\mathcal{G}$, r ← RB(D, n)

3: $\mathit{PAG}-I\leftarrow \arg \underset{G_i\in \mathcal{G}}{max}$ BSC-I(Gi, r, D)

4: return PAG-I