All the relevant objects (graphs, factors, Bayesian networks, etc) can be saved/loaded using the pickle format.
These objects can be saved using directly pickle.dump <pickle.dump> and pickle.load <pickle.load>. For example:
>>> import pickle >>> from pybnesian import Dag >>> g = Dag(["a", "b", "c", "d"], [("a", "b")]) >>> with open("saved_graph.pickle", "wb") as f: ... pickle.dump(g, f) >>> with open("saved_graph.pickle", "rb") as f: ... lg = pickle.load(f) >>> assert lg.nodes() == ["a", "b", "c", "d"] >>> assert lg.arcs() == [("a", "b")]
import os os.remove('saved_graph.pickle')
We can reduce some boilerplate code using the save methods: Factor.save() <pybnesian.Factor.save>, UndirectedGraph.save() <pybnesian.UndirectedGraph.save>, DirectedGraph.save() <pybnesian.DirectedGraph.save>, BayesianNetworkBase.save() <pybnesian.BayesianNetworkBase.save>, etc... Also, the load <pybnesian.load> can load any saved object:
>>> import pickle >>> from pybnesian import load, Dag >>> g = Dag(["a", "b", "c", "d"], [("a", "b")]) >>> g.save("saved_graph") >>> lg = load("saved_graph.pickle") >>> assert lg.nodes() == ["a", "b", "c", "d"] >>> assert lg.arcs() == [("a", "b")]
pybnesian.load