Algorithm 3.

OR-SGBN: SGBN from a single class

Input: data X ∈ ℝn×m

Initialize Θ(0) by least square fitting.

Initialize o(0) and ϒ(0) by solving Eqn. (4.2) with Θ = Θ(0).

Let T = 1.

repeat

Fixing ϒ = ϒ(T−1) and o = o(T−1).

Let t = 1, Θ(T−1,t=0) = Θ(T−1).

for ${\lambda }_{\mathit{dag}}={\lambda }_{\mathit{dag}}^{(1)}$ to ${\lambda }_{\mathit{dag}}^{(M)}$ do

Optimize Eqn. (4.2) with the initial solution Θ(T−1,t−1) to obtain Θ(T−1,t).

Let t=t+1.

end for

Let Θ(T) = Θ(T−1,M).

Fixing Θ(T), optimize Eqn. (4.2) to update o(T) and ϒ(T) to enforce DAG.

Let T = T + 1.

until convergence/max number of iterations

Output: Θ★ = Θ(T)