def testGetPosterior(self):
self.assertEqual(gum.getPosterior(self.bn, {}, "A"),
self.joint.margSumIn(["A"]))
self.assertEqual(gum.getPosterior(self.bn, {}, 2),
self.joint.margSumIn(["C"]))
self.assertEqual(gum.getPosterior(self.bn, {}, 'D'),
self.joint.margSumIn(["D"]))
self.ie.eraseAllTargets()
self.ie.addTarget("A")
self.ie.addTarget("F")
self.ie.addEvidence("B", 2)
self.ie.addEvidence("D", [0.2, 0.6, 0.6])
self.ie.makeInference()
posterior_joint = self.joint * \
gum.Potential().add(self.bn.variable("B")).fillWith([0, 0, 1]) * \
gum.Potential().add(self.bn.variable("D")).fillWith([0.2, 0.6, 0.6])
self.assertEqual(
gum.getPosterior(self.bn, {
1: 2,
"D": [0.2, 0.6, 0.6]
}, "A"),
posterior_joint.margSumIn(["A"]).normalize())
self.assertEqual(
gum.getPosterior(self.bn, {
"B": 2,
3: [0.2, 0.6, 0.6]
}, "F"),
posterior_joint.margSumIn(["F"]).normalize())
def getOneSampleFromMN(mn):
"""
Get one sample from a markov network
Examples
--------
>>> sample=mnl.getOneSampleFromMN(mn)
Parameters
----------
mn : pyAgrum.MarkovNet
the markov network
Returns
-------
dict
the sample, as a dictionary, keys are the variables names and values are the sampled values
"""
variablesToSample = mn.names()
sampledVariables = {}
while variablesToSample != []:
sampledValue = gum.getPosterior(mn,
target=variablesToSample[0],
evs=sampledVariables).draw()
sampledVariables.update({variablesToSample[0]: sampledValue})
variablesToSample.pop(0)
return sampledVariables
def testEvidenceImpactWithNodeId(self):
bn = gum.loadBN(self.agrumSrcDir('asia.bif'), [],
verbose=False) # verbose=False : don't want to see the warnings
ie = gum.LazyPropagation(bn)
self.assertEqual(len(ie.BN().arcs()),8)
with self.assertRaises(gum.InvalidArgument):
res = ie.evidenceImpact(0, [0, 1, 2])
res = ie.evidenceImpact(0, [1, 2])
self.assertEqual(res.nbrDim(), 2) # 2 indep 0 given 1
self.assertEqual(res.extract({"tuberculosis?": 0}), gum.getPosterior(bn, target=0, evs={1: 0}))
self.assertEqual(res.extract({"tuberculosis?": 1}), gum.getPosterior(bn, target=0, evs={1: 1}))
def showPosterior(bn, evs, target):
"""
shortcut for showProba(gum.getPosterior(bn,evs,target))
:param bn: the BayesNet
:param evs: map of evidence
:param target: name of target variable
"""
showProba(gum.getPosterior(bn, evs=evs, target=target))
def testEvidenceImpactWithName(self):
bn = gum.loadBN(self.agrumSrcDir('asia.bif'), [],
verbose=False) # verbose=False : don't want to see the warnings
ie = gum.LazyPropagation(bn)
with self.assertRaises(gum.InvalidArgument):
res = ie.evidenceImpact("visit_to_Asia?", ["visit_to_Asia?", "tuberculosis?", "tuberculos_or_cancer?"])
with self.assertRaises(gum.NotFound):
res = ie.evidenceImpact("visit_to_Asia?", ["toto", "tuberculosis?", "tuberculos_or_cancer?"])
res = ie.evidenceImpact("visit_to_Asia?", ["tuberculosis?", "tuberculos_or_cancer?"])
self.assertEqual(res.nbrDim(), 2) # 2 indep 0 given 1
self.assertEqual(res.extract({"tuberculosis?": 0}),
gum.getPosterior(bn, target="visit_to_Asia?", evs={"tuberculosis?": 0}))
self.assertEqual(res.extract({"tuberculosis?": 1}),
gum.getPosterior(bn, target="visit_to_Asia?", evs={"tuberculosis?": 1}))
def getPosterior(bn, evs, target):
"""
shortcut for proba2histo(gum.getPosterior(bn,evs,target))
:param bn: the BayesNet
:type bn: gum.BayesNet
:param evs: map of evidence
:type evs: dict(str->int)
:param target: name of target variable
:type target: str
:return: the matplotlib graph
"""
fig = proba2histo(gum.getPosterior(bn, evs=evs, target=target))
plt.close()
return _getMatplotFig(fig)
def fastSampleFromMNRecursive(mn, samples, samplesNumber,
variablesToSampleOrder, variablesEvidences,
onlineComputedPotential, sampledVariables, prog):
nextSampleIndex = len(sampledVariables)
if nextSampleIndex == len(variablesToSampleOrder):
samples.extend(sampledVariables for _ in range(samplesNumber))
if prog != None:
prog.increment_amount(samplesNumber)
prog.display()
else:
variable = variablesToSampleOrder[nextSampleIndex]
variableSample = defaultdict(lambda: 0)
sampledVariablesTuple = tuple(
sampledVariables[key]
for key in variablesToSampleOrder[:nextSampleIndex]
if key in variablesEvidences[variable])
if sampledVariablesTuple in onlineComputedPotential[variable]:
posterior = onlineComputedPotential[variable][
sampledVariablesTuple]
else:
posterior = gum.getPosterior(mn,
target=variable,
evs=sampledVariables)
onlineComputedPotential[variable][
sampledVariablesTuple] = posterior
for _ in range(samplesNumber):
value = posterior.draw()
variableSample[value] += 1
for value in variableSample.keys():
fastSampleFromMNRecursive(mn, samples, variableSample[value],
variablesToSampleOrder,
variablesEvidences,
onlineComputedPotential, {
**sampledVariables,
**{
variable: value
}
}, prog)
# Exercice 7
alpha = 0.05
bn_struct = learn_BN_structure(data, dico, alpha)
#display_BN(names, bn_struct, "asia", style)
display_BN_graphviz(names, bn_struct, "asia", style)
# Exercice 7bis
# création du réseau bayésien à la aGrUM
bn = learn_parameters(bn_struct, "2015_tme5_asia.csv")
# affichage de sa taille
print(bn)
# récupération de la ''conditional probability table'' (CPT) et affichage de cette table
gnb.showPotential(bn.cpt(bn.idFromName('bronchitis?')))
# calcul de la marginale
proba = gum.getPosterior(bn, {}, 'bronchitis?')
# affichage de la marginale
gnb.showPotential(proba)
#calcul d'une distribution marginale a posteriori : P(bronchitis? | smoking? = true, turberculosis? = false )
gnb.showPotential(
gum.getPosterior(bn, {
'smoking?': 'true',
'tuberculosis?': 'false'
}, 'bronchitis?'))