Working with networkx graphs

networkx.DiGraph instances are accepted by bnm.to_graphlike, which converts them to the canonical (n_vars, endpoints, var_names) representation. Edge orientation is taken from the directed-edge convention; undirected (CPDAG-style) edges are encoded by setting graph[u][v]["type"] = "undirected" on the networkx side.

import networkx as nx
import bnm

g = nx.DiGraph()
g.add_edges_from([("A", "B"), ("A", "C"), ("B", "D"), ("C", "D")])
gl = bnm.to_graphlike(g)
bnm.shd(gl, other_graph)

Note

This page is currently a stub. A worked example covering both DAG and CPDAG-style networkx inputs will land in v0.x.x.