Skip to content
GitLab
Commit 17cfac1f authored by Tiago Peixoto's avatar Tiago Peixoto
Browse files

inference.rst: Update doctests

parent 187b6c04
Pipeline #234 failed with stage
in 289 minutes and 20 seconds
Hide whitespace changes
Inline Side-by-side
doc/demos/inference/inference.rst
View file @ 17cfac1f
......@@ -456,7 +456,7 @@ case of the `C. elegans` network we have
:options: +NORMALIZE_WHITESPACE
Non-degree-corrected DL: 8507.97432099
Degree-corrected DL: 8228.11609772
Degree-corrected DL: 8249.76252971
Since it yields the smallest description length, the degree-corrected
fit should be preferred. The statistical significance of the choice can
......@@ -482,12 +482,12 @@ fits. In our particular case, we have
.. testoutput:: model-selection
:options: +NORMALIZE_WHITESPACE
ln Λ: -279.858223272
ln Λ: -258.211791282
The precise threshold that should be used to decide when to `reject a
hypothesis <https://en.wikipedia.org/wiki/Hypothesis_testing>`_ is
subjective and context-dependent, but the value above implies that the
particular degree-corrected fit is around :math:`e^{280} \sim 10^{121}`
particular degree-corrected fit is around :math:`e^{258} \sim 10^{112}`
times more likely than the non-degree corrected one, and hence it can be
safely concluded that it provides a substantially better fit.
......@@ -509,12 +509,12 @@ example, for the American football network above, we have:
.. testoutput:: model-selection
:options: +NORMALIZE_WHITESPACE
Non-degree-corrected DL: 1749.51938237
Degree-corrected DL: 1780.57671694
ln Λ: -31.0573345685
Non-degree-corrected DL: 1738.71605436
Degree-corrected DL: 1795.27140372
ln Λ: -56.5553493622
Hence, with a posterior odds ratio of :math:`\Lambda \sim e^{-36} \sim
10^{-16}` in favor of the non-degree-corrected model, it seems like the
Hence, with a posterior odds ratio of :math:`\Lambda \sim e^{-56} \sim
10^{-24}` in favor of the non-degree-corrected model, it seems like the
degree-corrected variant is an unnecessarily complex description for
this network.
......@@ -1068,8 +1068,8 @@ evidence efficiently, as we show below, using
.. testoutput:: model-evidence
Model evidence for deg_corr = True: -575.864972067 (mean field), -802.39062289 (Bethe)
Model evidence for deg_corr = False: -584.307313493 (mean field), -707.827204203 (Bethe)
Model evidence for deg_corr = True: -575.889667037 (mean field), -794.733483508 (Bethe)
Model evidence for deg_corr = False: -618.120673252 (mean field), -729.104723167 (Bethe)
If we consider the more accurate approximation, the outcome shows a
preference for the non-degree-corrected model.
......@@ -1133,8 +1133,8 @@ approach for the same network, using the nested model.
.. testoutput:: model-evidence
Model evidence for deg_corr = True: -346.618790006 (mean field), -601.313781849 (Bethe)
Model evidence for deg_corr = False: -374.614350884 (mean field), -563.256840699 (Bethe)
Model evidence for deg_corr = True: -333.948404832 (mean field), -651.91054824 (Bethe)
Model evidence for deg_corr = False: -360.235692786 (mean field), -549.293909028 (Bethe)
The results are similar: If we consider the most accurate approximation,
the non-degree-corrected model possesses the largest evidence. Note also
......@@ -1363,11 +1363,11 @@ above).
.. testoutput:: missing-edges
likelihood-ratio for (101, 102): 0.372308
likelihood-ratio for (17, 56): 0.627692
likelihood-ratio for (101, 102): 0.436558
likelihood-ratio for (17, 56): 0.563442
From which we can conclude that edge :math:`(17, 56)` is around twice as
likely as :math:`(101, 102)` to be a missing edge.
From which we can conclude that edge :math:`(17, 56)` is more likely
than :math:`(101, 102)` to be a missing edge.
The prediction using the non-nested model can be performed in an
entirely analogous fashion.
......
doc/test_inference.py
View file @ 17cfac1f
......@@ -28,23 +28,26 @@ rec_p = g.new_ep("double", random(g.num_edges()))
rec_s = g.new_ep("double", normal(0, 10, g.num_edges()))
def gen_state(directed, deg_corr, layers, overlap, rec, allow_empty):
def gen_state(directed, deg_corr, layers, overlap, rec, rec_type, allow_empty):
u = GraphView(g, directed=directed)
if layers != False:
state = graph_tool.inference.LayeredBlockState(u, B=u.num_vertices(),
deg_corr=deg_corr,
ec=ec.copy(),
rec=rec,
recs=[rec] if rec is not None else [],
rec_types=[rec_type] if rec is not None else [],
overlap=overlap,
layers=layers == True,
allow_empty=allow_empty)
elif overlap:
state = graph_tool.inference.OverlapBlockState(u, B=2 * u.num_edges(),
rec=rec,
recs=[rec] if rec is not None else [],
rec_types=[rec_type] if rec is not None else [],
deg_corr=deg_corr)
else:
state = graph_tool.inference.BlockState(u, B=u.num_vertices(),
rec=rec,
recs=[rec] if rec is not None else [],
rec_types=[rec_type] if rec is not None else [],
deg_corr=deg_corr,
allow_empty=allow_empty)
return state
......@@ -53,7 +56,7 @@ def gen_state(directed, deg_corr, layers, overlap, rec, allow_empty):
pranges = [("directed", [False, True]),
("overlap", [False, True]),
("layered", [False, "covariates", True]),
("rec", [None, "positive", "signed"]),
("rec", [None, "real-exponential", "real-normal"]),
("deg_corr", [False, True]),
("dl", [False, True]),
("degree_dl_kind", ["uniform", "distributed", "entropy"]),
......@@ -84,13 +87,13 @@ for pvals in iter_ranges(pranges):
print(params, file=out)
rec_ = None
if rec == "positive":
if rec == "real-exponential":
rec_ = rec_p
elif rec == "signed":
elif rec == "real-normal":
rec_ = rec_s
print("\t mcmc (unweighted)", file=out)
state = gen_state(directed, deg_corr, layered, overlap, rec_, allow_empty)
state = gen_state(directed, deg_corr, layered, overlap, rec_, rec, allow_empty)
print("\t\t",
state.mcmc_sweep(beta=0,
......@@ -107,7 +110,7 @@ for pvals in iter_ranges(pranges):
exact=exact)),
state.get_nonempty_B(), file=out)
state = gen_state(directed, deg_corr, layered, overlap, rec_, allow_empty)
state = gen_state(directed, deg_corr, layered, overlap, rec_, rec, allow_empty)
if not overlap:
print("\t mcmc", file=out)
......@@ -136,7 +139,7 @@ for pvals in iter_ranges(pranges):
print("\t merge", file=out)
state = gen_state(directed, deg_corr, layered, overlap, rec_, allow_empty)
state = gen_state(directed, deg_corr, layered, overlap, rec_, rec, allow_empty)
if not overlap:
bstate = state.get_block_state(vweight=True, deg_corr=deg_corr)
......@@ -174,7 +177,7 @@ for pvals in iter_ranges(pranges):
print("\t shrink", file=out)
state = gen_state(directed, deg_corr, layered, overlap, rec_, allow_empty)
state = gen_state(directed, deg_corr, layered, overlap, rec_, rec, allow_empty)
state = state.shrink(B=5, entropy_args=dict(dl=dl,
degree_dl_kind=degree_dl_kind,
multigraph=False,
......@@ -185,7 +188,7 @@ for pvals in iter_ranges(pranges):
pranges = [("directed", [False, True]),
("overlap", [False, True]),
("layered", [False, "covariates", True]),
("rec", [None, "positive", "signed"]),
("rec", [None, "real-exponential", "real-normal"]),
("deg_corr", [False, True]),
("degree_dl_kind", ["uniform", "distributed", "entropy"]),
("exact", [True])]
......@@ -205,16 +208,17 @@ for pvals in iter_ranges(pranges):
print(params, file=out)
rec_ = None
if rec == "positive":
rec_ = []
if rec == "real-exponential":
rec_ = rec_p
elif rec == "signed":
elif rec == "real-normal":
rec_ = rec_s
if layered != False:
state_args = dict(ec=ec, layers=(layered == True), rec=rec_)
state_args = dict(ec=ec, layers=(layered == True), recs=[rec_],
rec_types=[rec])
else:
state_args = dict(rec=rec_)
state_args = dict(recs=[rec_], rec_types=[rec])
entropy_args = dict(exact=exact)
......
Supports Markdown
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment