def main3(self):
values = [[-0.1, None], [-0.5, None], [-0.3, True], [0.3, None],
[0.4, None], [0.1, None], [-0.4, None], [-0.2, None],
[-0.3, None], [0.1, None], [0.4, False], [0.2, None]]
# construct a BN with one boolean root and one continuous child
a = CPT("A", [])
b = GDT("B", ["A"])
#b.setTieVariances(True)
bn = BNet()
bn.add(a)
bn.add(b)
#bn.EM_PRINT_STATUS=True
# train it using EM
# bn.train(["B","A"], values, [], 5)
bn.train(["B"], values, ["A"], 5)
print a
print b
b.instance = 0.0
test = bn.infer_naive(["A"])
test.normalize()
print test
def main2(self):
values = [
# A G F E
[True, False, True, True],
[True, True, True, True],
[True, False, True, True],
[False, True, True, False],
[True, False, True, False],
[True, True, True, False],
[True, False, False, True],
[False, True, False, True],
[True, True, False, True],
[False, False, False, False],
[False, True, False, False],
[False, True, False, False],
[False, True, False, False]
]
# construct a BN with two root nodes
a = CPT("A", [])
b = CPT("B", [])
d = CPT("D", ["A"])
e = CPT("E", ["A", "B"])
f = CPT("F", ["D", "E"])
g = CPT("G", ["F"])
# put them all into the BN
bn = BNet()
bn.add(a)
bn.add(b)
bn.add(d)
bn.add(e)
bn.add(f)
bn.add(g)
# train it using EM
bn.train(["A", "G", "F", "E"], values, ["B", "D"], 9)
print a
print b
print d
print e
print f
print g
a.instance = True
test = bn.infer_naive(["E"])
test.normalize()
print test
def main(self):
values = [
# A G F E C
[True, False, True, True, 0.1],
[True, True, True, True, 0.2],
[True, False, True, True, 0.5],
[False, True, True, False, -0.3],
[True, False, True, False, -0.0],
[True, True, True, False, -0.2],
[True, False, False, True, 0.6],
[False, True, False, True, 0.8],
[True, True, False, True, 0.1],
[False, False, False, False, -0.8],
[False, True, False, False, -0.2],
[False, True, False, False, -0.3],
[False, True, False, False, -0.5]
]
# construct a BN with two root nodes
a = CPT("A", [])
b = CPT("B", [])
c = GDT("C", ["B"]) # Gaussian continuous
d = CPT("D", ["A"])
e = CPT("E", ["A", "B"])
f = CPT("F", ["D", "E"])
g = CPT("G", ["F"])
# put them all into the BN
bn = BNet()
bn.add(a)
bn.add(b)
bn.add(c)
bn.add(d)
bn.add(e)
bn.add(f)
bn.add(g)
# train it using EM
bn.train(["A", "G", "F", "E", "C"], values, ["B", "D"], 9)
print a
print b
print c
print d
print e
print f
print g
c.instance = 0.5
a.instance = True
test = bn.infer_naive(["E"])
test.normalize()
print test
def main4(self):
values = [
# A G F E C B D
[True, False, True, True, 0.1, False, True],
[True, True, True, True, 0.2, True, True],
[True, False, True, True, 0.5, False, False],
[False, True, True, False, -0.3, False, True],
[True, False, True, False, -0.0, True, True],
[True, True, True, False, -0.2, False, True],
[True, False, False, True, 0.6, False, True],
[False, True, False, True, 0.8, False, True],
[True, True, False, True, 0.1, True, True],
[False, False, False, False, -0.8, True, False],
[False, True, False, False, -0.2, False, True],
[False, True, False, False, -0.3, True, True],
[False, True, False, False, -0.5, False, True]
]
a = CPT("A", [])
b = CPT("B", [])
c = GDT("C", ["B"]) # Gaussian continuous
d = CPT("D", ["A"])
e = CPT("E", ["A", "B"])
f = CPT("F", ["D", "E"])
g = CPT("G", ["F"])
# put them all into the BN
bn = BNet()
bn.add(a)
bn.add(b)
bn.add(c)
bn.add(d)
bn.add(e)
bn.add(f)
bn.add(g)
# train it using EM
bn.train(["A", "G", "F", "E", "C", "B", "D"], values, [], 9)
# bn.train( ["A", "G", "F", "E"], values, ["B", "D"], 9)
print a
print b
print c
print d
print e
print f
print g
c.instance = 0.5
a.instance = True
test = bn.infer_naive(["E"])
test.normalize()
print test
def testtrain1(self):
"""
\n****Testing training the large train network,
two unknown, one GDT****
"""
values=[
# A G F E C
[True, False, True, True, 0.1],
[True, True, True, True, 0.2],
[True, False, True, True, 0.5],
[False, True, True, False, -0.3],
[True, False, True, False, -0.0],
[True, True, True, False, -0.2],
[True, False, False, True, 0.6],
[False, True, False, True, 0.8],
[True, True, False, True, 0.1],
[False, False, False, False, -0.8],
[False, True, False, False, -0.2],
[False, True, False, False, -0.3],
[False, True, False, False, -0.5]]
a = CPT("A", [])
b = CPT("B", [])
c = GDT("C", ["B"]) # Gaussian continuous
d = CPT("D", ["A"])
e = CPT("E", ["A", "B"])
f = CPT("F", ["D", "E"])
g = CPT("G", ["F"])
bn = BNet()
bn.add(a)
bn.add(b)
bn.add(c)
bn.add(d)
bn.add(e)
bn.add(f)
bn.add(g)
bn.train( ["A", "G", "F", "E", "C"], values, ["B", "D"], 9)
c.instance = 0.5
a.instance = True
test = bn.infer_naive(["E"])
test.normalize()
print test
def setUp(self):
self.burglary = CPT("B", [])
self.burglary.put([], 0.001)
self.earthquake = CPT("E", [])
self.earthquake.put([], 0.002)
self.alarm = CPT("A", ["B", "E"],
{
(True, True): 0.95,
(True, False): 0.94,
(False, True): 0.29,
(False, False): 0.001
}
)
self.johncalls = CPT("J", ["A"],
{
(True,): 0.9,
(False,): 0.05
}
)
self.marycalls = CPT("M", ["A"],
{
(True,) : 0.7,
(False,) : 0.01
}
)
self.bn = BNet()
self.bn.add(self.burglary)
self.bn.add(self.earthquake)
self.bn.add(self.alarm)
self.bn.add(self.johncalls)
self.bn.add(self.marycalls)
self.A = CPT("A", [])
self.A.put([], 0.36)
self.B = CPT("B", [])
self.B.put([], 0.21)
self.D = CPT("D", ["A"])
self.D.put([True], 0.12)
self.D.put([False], 0.84)
self.E = CPT("E", ["A", "B"])
self.E.put([True, True], 0.76)
self.E.put([False, True], 0.14)
self.E.put([True, False], 0.57)
self.E.put([False, False], 0.68)
self.F = CPT("F", ["D", "E"])
self.F.put([True, True], 0.06)
self.F.put([False, True], 0.007)
self.F.put([True, False], 0.61)
self.F.put([False, False], 0.061)
self.G = CPT("G", ["F"])
self.G.put([True], 0.481)
self.G.put([False], 0.085)
self.bn_2 = BNet()
self.bn_2.add(self.A)
self.bn_2.add(self.B)
self.bn_2.add(self.D)
self.bn_2.add(self.E)
self.bn_2.add(self.F)
self.bn_2.add(self.G)
class Test(unittest.TestCase):
def setUp(self):
self.burglary = CPT("B", [])
self.burglary.put([], 0.001)
self.earthquake = CPT("E", [])
self.earthquake.put([], 0.002)
self.alarm = CPT("A", ["B", "E"],
{
(True, True): 0.95,
(True, False): 0.94,
(False, True): 0.29,
(False, False): 0.001
}
)
self.johncalls = CPT("J", ["A"],
{
(True,): 0.9,
(False,): 0.05
}
)
self.marycalls = CPT("M", ["A"],
{
(True,) : 0.7,
(False,) : 0.01
}
)
self.bn = BNet()
self.bn.add(self.burglary)
self.bn.add(self.earthquake)
self.bn.add(self.alarm)
self.bn.add(self.johncalls)
self.bn.add(self.marycalls)
self.A = CPT("A", [])
self.A.put([], 0.36)
self.B = CPT("B", [])
self.B.put([], 0.21)
self.D = CPT("D", ["A"])
self.D.put([True], 0.12)
self.D.put([False], 0.84)
self.E = CPT("E", ["A", "B"])
self.E.put([True, True], 0.76)
self.E.put([False, True], 0.14)
self.E.put([True, False], 0.57)
self.E.put([False, False], 0.68)
self.F = CPT("F", ["D", "E"])
self.F.put([True, True], 0.06)
self.F.put([False, True], 0.007)
self.F.put([True, False], 0.61)
self.F.put([False, False], 0.061)
self.G = CPT("G", ["F"])
self.G.put([True], 0.481)
self.G.put([False], 0.085)
self.bn_2 = BNet()
self.bn_2.add(self.A)
self.bn_2.add(self.B)
self.bn_2.add(self.D)
self.bn_2.add(self.E)
self.bn_2.add(self.F)
self.bn_2.add(self.G)
def tearDown(self):
pass
def testsimple1(self):
print "\n****Setting JohnCalls = True and MaryCalls = True****"
self.johncalls.instance = True
self.marycalls.instance = True
test = self.bn.infer_naive(["B"])
test.normalize()
print test
self.assertAlmostEqual(test.get([True]), 0.284, 3)
def testsimple2(self):
print "\n****Setting Earthquake = False****"
self.burglary.instance = None
self.earthquake.instance = False
test = self.bn.infer_naive(["B", "E"])
test.normalize()
print test
def testsimple3(self):
print "\n****Setting burglary = True****"
self.earthquake.instance = None
self.burglary.instance = True
test = self.bn.infer_naive(["B", "E"])
test.normalize()
print test
def testancestors(self):
print "\n****Checking Ancestors****"
self.assertEqual(self.bn.get_ancestors("J"),
set(["B", "E", "A"]))
self.assertEqual(self.bn.get_ancestors("M"),
set(["B", "E", "A"]))
self.assertEqual(self.bn.get_ancestors("B"),
set([]))
self.assertEqual(self.bn.get_ancestors("E"),
set([]))
def testnodevalueget(self):
print "\n****Checking getting node values****"
print "Burglary True"
print self.burglary.get(True, [])
print "Earthquake False"
print self.earthquake.get(False, [])
print "Alarm True, parents True:True"
print self.alarm.get(True, [True, True])
print "Alarm True, parents False:True"
print self.alarm.get(True, [False, True])
def testroot(self):
print "\n****Checking node root****"
self.assertTrue(self.burglary.is_root())
self.assertFalse(self.alarm.is_root())
self.assertFalse(self.johncalls.is_root())
X = CPT("X", ["M"])
self.assertFalse(X.is_root())
def testadvanced(self):
print "\n****Testing inference larger network****"
print "Setting marycalls = False & burglary = True, inferring X"
X = CPT("X", ["M"])
X.put([True], .89)
X.put([False], .56)
Y = CPT("Y", ["A", "B"])
Y.put([True, False], .23)
Y.put([False, True], .74)
Y.put([False, False], .57)
Y.put([True, True], .12)
#
Z = CPT("Z", ["X", "Y"])
Z.put([True, True], 0.23)
Z.put([False, False], 0.41)
Z.put([True, False], 0.16)
Z.put([False, True], 0.75)
self.bn.add(X)
self.bn.add(Y)
self.bn.add(Z)
# print X
# print Y
# print Z
self.marycalls.instance = False
self.burglary.instance = True
# for node in self.bn.nodes.values():
# print node
# print node.name
# print node.instance
test = self.bn.infer_naive(["X"])
test.normalize()
print test
test2 = self.bn.infer_naive(["Z", "A", "E"])
def testadvanced2(self):
print "\n****Testing inference larger network****"
print "Setting alarm = False & burglary = True, inferring [Z,A,E]"
X = CPT("X", ["M"])
X.put([True], .89)
X.put([False], .56)
Y = CPT("Y", ["A", "B"])
Y.put([True, False], .23)
Y.put([False, True], .74)
Y.put([False, False], .57)
Y.put([True, True], .12)
#
Z = CPT("Z", ["X", "Y"])
Z.put([True, True], 0.23)
Z.put([False, False], 0.41)
Z.put([True, False], 0.16)
Z.put([False, True], 0.75)
self.bn.add(X)
self.bn.add(Y)
self.bn.add(Z)
self.alarm.instance = False
self.burglary.instance = True
test = self.bn.infer_naive(["Z", "A", "E"])
test.normalize()
print test
def testadvanced3(self):
print "\n****Testing inference larger network****"
print "Setting alarm = False & Y = True, inferring C"
X = CPT("X", ["M"])
X.put([True], .89)
X.put([False], .56)
Y = CPT("Y", ["A", "B"])
Y.put([True, False], .23)
Y.put([False, True], .74)
Y.put([False, False], .57)
Y.put([True, True], .12)
#
Z = CPT("Z", ["X", "Y"])
Z.put([True, True], 0.23)
Z.put([False, False], 0.41)
Z.put([True, False], 0.16)
Z.put([False, True], 0.75)
C = CPT("C", ["B"])
C.put([True], .36)
C.put([False], .67)
self.bn.add(X)
self.bn.add(Y)
self.bn.add(Z)
self.bn.add(C)
Y.instance = True
self.alarm.instance = False
test = self.bn.infer_naive(["C"])
test.normalize()
print test
def testadvanced4(self):
print "\n****Testing inference larger network (Train network)****"
print "Setting A = True, inferring F"
self.A.instance = True
test = self.bn_2.infer_naive(["F"])
test.normalize()
print test
def testadvanced5(self):
print "\n****Testing inference larger network (Train network)****"
print "Setting A = True, inferring extra C node (not GDT)"
C = CPT("C", ["B"])
C.put([True], 0.95)
C.put([False], 0.86)
self.bn_2.add(C)
self.A.instance = True
test = self.bn_2.infer_naive(["C"])
test.normalize()
print test
def testadvanced6(self):
print "\n****Testing inference larger network (Train network)****"
print "Setting A = True, inferring extra C node (not GDT) and D,E"
C = CPT("C", ["B"])
C.put([True], 0.95)
C.put([False], 0.86)
self.bn_2.add(C)
self.A.instance = True
test = self.bn_2.infer_naive(["C", "D", "E"])
test.normalize()
print test
def testrelevant(self):
print "\n****Testing getting relevant network****"
print "Getting relevant network for node J in expanded network"
X = CPT("X", ["M"])
X.put([True], .89)
X.put([False], .56)
Y = CPT("Y", ["A", "B"])
Y.put([True, False], .23)
Y.put([False, True], .74)
Y.put([False, False], .57)
Y.put([True, True], .12)
#
Z = CPT("Z", ["X", "Y"])
Z.put([True, True], 0.23)
Z.put([False, False], 0.41)
Z.put([True, False], 0.16)
Z.put([False, True], 0.75)
self.bn.add(X)
self.bn.add(Y)
self.bn.add(Z)
print self.bn.get_relevant("J").nodes.keys()
def test_GDT_advanced(self):
print "\n****Testing large network inferring GDT****"
X = CPT("X", ["M"])
X.put([True], .89)
X.put([False], .56)
Y = CPT("Y", ["A", "B"])
Y.put([True, False], .23)
Y.put([False, True], .74)
Y.put([False, False], .57)
Y.put([True, True], .12)
#
Z = CPT("Z", ["X", "Y"])
Z.put([True, True], 0.23)
Z.put([False, False], 0.41)
Z.put([True, False], 0.16)
Z.put([False, True], 0.75)
G = GDT("G", ["X"])
G.put(Gaussian(0.24, 0.5), [True])
G.put(Gaussian(0.62, .1), [False])
self.bn.add(X)
self.bn.add(Y)
self.bn.add(Z)
self.bn.add(G)
# print self.bn.get_relevant("G").nodes.keys()
# print G
test = self.bn.infer_naive("G")
print test
def testtrain1(self):
"""
\n****Testing training the large train network,
two unknown, one GDT****
"""
values=[
# A G F E C
[True, False, True, True, 0.1],
[True, True, True, True, 0.2],
[True, False, True, True, 0.5],
[False, True, True, False, -0.3],
[True, False, True, False, -0.0],
[True, True, True, False, -0.2],
[True, False, False, True, 0.6],
[False, True, False, True, 0.8],
[True, True, False, True, 0.1],
[False, False, False, False, -0.8],
[False, True, False, False, -0.2],
[False, True, False, False, -0.3],
[False, True, False, False, -0.5]]
a = CPT("A", [])
b = CPT("B", [])
c = GDT("C", ["B"]) # Gaussian continuous
d = CPT("D", ["A"])
e = CPT("E", ["A", "B"])
f = CPT("F", ["D", "E"])
g = CPT("G", ["F"])
bn = BNet()
bn.add(a)
bn.add(b)
bn.add(c)
bn.add(d)
bn.add(e)
bn.add(f)
bn.add(g)
bn.train( ["A", "G", "F", "E", "C"], values, ["B", "D"], 9)
c.instance = 0.5
a.instance = True
test = bn.infer_naive(["E"])
test.normalize()
print test
def testtrain2(self):
"""
\n****Testing training the small train network,
****
"""
values=[
[-0.1, None],
[-0.5, None],
[-0.3, True],
[0.3, None],
[0.4, None],
[0.1, None],
[-0.4, None],
[-0.2, None],
[-0.3, None],
[0.1, None],
[0.4, False],
[0.2, None]
]
# construct a BN with one boolean root and one continuous child
a= CPT("A", [])
b= GDT("B", ["A"])
#b.setTieVariances(True)
bn= BNet()
bn.add(a)
bn.add(b)
#bn.EM_PRINT_STATUS=True
# train it using EM
# bn.train(["B","A"], values, [], 5)
bn.train(["B"], values, ["A"], 5)
print a
print b
b.instance = 0.0
test = bn.infer_naive(["A"])
test.normalize()
print test