def test_acyclic_generators():
for mechanism in mechanisms:
agg, data = AcyclicGraphGenerator(mechanism,
points=200,
nodes=10,
parents_max=3).generate()
assert type(agg) == nx.DiGraph
assert nx.is_directed_acyclic_graph(agg)
def test_noises():
for noise in [normal_noise, uniform_noise]:
agg, data = AcyclicGraphGenerator("linear",
points=200,
nodes=10,
parents_max=3,
noise=noise).generate()
assert type(agg) == nx.DiGraph
assert nx.is_directed_acyclic_graph(agg)
def test_causes():
for cause in [gmm_cause, gaussian_cause]:
agg, data = AcyclicGraphGenerator(
"linear",
points=200,
nodes=10,
parents_max=3,
initial_variable_generator=cause).generate()
assert type(agg) == nx.DiGraph
assert nx.is_directed_acyclic_graph(agg)
for nb_points in [200,500,1000]:
for nb_nodes in [20,50,100]:
for mecanism in mechanisms:
print(nb_points)
print(nb_nodes)
print(mecanism)
if(mecanism == "linear"):
cause = gaussian_cause
else:
cause = gmm_cause
agg, data = AcyclicGraphGenerator(mecanism, points=nb_points, nodes=nb_nodes, parents_max=5, initial_variable_generator=cause).generate()
Adj = nx.to_numpy_matrix(agg)
cpt = 0
for i in range(Adj.shape[0]):
for j in range(Adj.shape[1]):
if(Adj[i,j]==1):
cpt+=1
cause = data["V" + str(i)].values
effect = data["V" + str(j)].values
from cdt.generators import AcyclicGraphGenerator
for mechanism in ['polynomial','NN','sigmoid_add', 'sigmoid_mix',
'gp_add', 'gp_mix', 'linear']:# 'linear', 'sigmoid_add', 'sigmoid_mix',
a = AcyclicGraphGenerator(mechanism, nodes=20)
a.generate()
a.to_csv("-".join([str(j) for j in [mechanism,'train', str(20)]]), index=False)