def test_DCER_score(self):
p_mat = self.p_mat
graph = self.graph
estimator = DCEREstimator()
_test_score(estimator, p_mat, graph)
with pytest.raises(ValueError):
estimator.score_samples(graph=graph[1:500, 1:500])
def test_DCER_inputs(self):
with pytest.raises(TypeError):
DCEREstimator(directed="hey")
with pytest.raises(TypeError):
DCEREstimator(loops=6)
graph = er_np(100, 0.5)
dcere = DCEREstimator()
with pytest.raises(ValueError):
dcere.fit(graph[:, :99])
with pytest.raises(ValueError):
dcere.fit(graph[..., np.newaxis])
def test_DCER_nparams(self):
n_verts = 1000
graph = self.graph
e = DCEREstimator(directed=True)
e.fit(graph)
assert e._n_parameters() == (n_verts + 1)
def test_DCER_sample(self):
np.random.seed(8888)
estimator = DCEREstimator(directed=True, loops=False)
g = self.graph
p_mat = self.p_mat
with pytest.raises(NotFittedError):
estimator.sample()
estimator.fit(g)
with pytest.raises(ValueError):
estimator.sample(n_samples=-1)
with pytest.raises(TypeError):
estimator.sample(n_samples="nope")
B = 0.5
dc = np.random.uniform(0.25, 0.75, size=100)
p_mat = np.outer(dc, dc) * B
p_mat -= np.diag(np.diag(p_mat))
g = sample_edges(p_mat, directed=True)
estimator.fit(g)
estimator.p_mat_ = p_mat
_test_sample(estimator, p_mat, n_samples=1000, atol=0.2)
fig.suptitle('Erdos-Reyni (ER) model', fontsize=16)
print(f"ER \"p\" parameter: {er.p_}")
plt.savefig('figure.png')
# ---------------------------------------
# Degree-corrected Erdos-Reyni (DCER) model
# ---------------------------------------
'''A slightly more complicated variant of the ER model is the degree-corrected
Erdos-Reyni model (DCER). Here, there is still a global parameter 𝑝 to specify
relative connection probability between all edges. However, we add a promiscuity
parameter 𝜃𝑖 for each node 𝑖 which specifies its expected degree relative to other
nodes: 𝑃𝑖𝑗=𝜃𝑖𝜃𝑗𝑝, so the probility of an edge from 𝑖 to 𝑗 is a function of the two nodes'
degree-correction parameters, and the overall probability of an edge in the graph'''
from graspy.models import DCEREstimator
dcer = DCEREstimator(directed=True, loops=False)
dcer.fit(adj) # Fit Degree-corrected Erdos-Reyni model
promiscuities = dcer.degree_corrections_
# Show results
fig, ax = plt.subplots(1, 2)
plotHeatmap(dcer.p_mat_,
"DCER probability matrix",
params={
'vmin': 0,
'vmax': 1,
'ax': ax[0]
})
plotHeatmap(dcer.sample()[0], "DCER sample", params={'ax': ax[1]})
fig.suptitle('Degree-corrected Erdos-Reyni (ER) model', fontsize=16)
print(f"DCER \"p\" parameter: {dcer.p_}")
outer_hier_labels=lcc_hemisphere,
hier_label_fontsize=10,
sort_nodes=True,
)
gridplot(
[embed_graph],
height=15,
inner_hier_labels=lcc_simple_classes,
outer_hier_labels=lcc_hemisphere,
hier_label_fontsize=10,
sizes=(1, 10),
sort_nodes=True,
)
#%%
dcer = DCEREstimator()
dcer.fit(embed_graph)
gridplot(
[binarize(embed_graph)],
height=15,
inner_hier_labels=lcc_simple_classes,
outer_hier_labels=lcc_hemisphere,
hier_label_fontsize=10,
sizes=(5, 5),
sort_nodes=True,
)
gridplot(
[dcer.sample()[0]],
height=15,
inner_hier_labels=lcc_simple_classes,
outer_hier_labels=lcc_hemisphere,
font_scale=1.5,
title="ER probability matrix",
vmin=0, vmax=1,
sort_nodes=True)
plt.savefig("ERProbabilityMatrix", bbox_inches='tight')
heatmap(er.sample()[0],
inner_hier_labels=labels,
font_scale=1.5,
title="ER sample",
sort_nodes=True);
plt.savefig("ERSample", bbox_inches='tight')
dcer = DCEREstimator(directed=True,loops=False)
dcer.fit(adj)
print(f"DCER \"p\" parameter: {dcer.p_}")
heatmap(dcer.p_mat_,
inner_hier_labels=labels,
vmin=0,
vmax=1,
font_scale=1.5,
title="DCER probability matrix",
sort_nodes=True)
plt.savefig("DCERProbabilityMatrix", bbox_inches='tight')
heatmap(dcer.sample()[0],
inner_hier_labels=labels,
font_scale=1.5,
def test_DCER_score(self):
p_mat = self.p_mat
graph = self.graph
estimator = DCEREstimator()
_test_score(estimator, p_mat, graph)