def load_bn_from_BIF(dataset_name='child', folder=DATA_FOLDER, verbose=False):
bif_file = os.path.join(folder, BIF_FOLDER_MAP[dataset_name],
dataset_name + '.bif', dataset_name + '.bif')
image_file = os.path.join(folder, BIF_FOLDER_MAP[dataset_name],
dataset_name + '.png')
if verbose:
print(f'Loading graph from {bif_file}')
reader = BIFReader(bif_file)
model = reader.get_model()
if os.path.exists(image_file):
if verbose:
print(f'Loading graph image from {image_file}')
model.image = PIL.Image.open(image_file)
# Take the leaves as features
__FEATURES = model.get_leaves()
__ROOTS = model.get_roots()
__NODES = model.nodes()
__NOT_FEATURES = list({
node
for node in __NODES if node not in __FEATURES and node not in __ROOTS
})
if verbose:
print(f'Nodes: {__NODES} ({len(__NODES)})')
print(f'Features/Leaves: {__FEATURES} ({len(__FEATURES)})')
print(f'Roots: {__ROOTS} ({len(__ROOTS)})')
print(
f'Intermediate (non-roots/non-leaves): {__NOT_FEATURES} ({len(__NOT_FEATURES)})'
)
return model
def __init__(self, evidence, bif_path=None, model=None):
"""
Args:
evidence A dictionary with observations, e.g., {node-name: node-value}.
model / bif_path PGMPY model specification or .bif file path.
"""
assert bif_path is not None or model is not None, \
"You must set either pgmpy model or bif_path!"
if model is None:
logger.debug("[BayesianNetworkVI] reading from %s" % bif_path)
reader = BIFReader(bif_path)
model = reader.get_model()
self._model = model
self._evidence = evidence
latent_variables = sorted( set(v for cpd in self._model.get_cpds()
for v in cpd.variables).difference(evidence) )
observed_variables = sorted(evidence)
self._no2var = dict(enumerate(latent_variables+observed_variables))
self._var2no = dict((var, no) for no, var in self._no2var.items())
logger.debug("[BayesianNetworkVI] mapping(id->variable)=%s" % self._no2var)
variable2values = cpds_to_variables(self._model.get_cpds())
self._variable2values = dict( (var, dict( (v, i) for i, v in enumerate(vals) ))
for var, vals in variable2values.items())
self._prob_arrays = [cpd_to_prob_array(cpd)+EPS for cpd in self._model.get_cpds()] #!TODO renormalization
logger.info("[BayesianNetworkVI] %i vars with max cardinality=%i => enumeration size=%i" %
(len(variable2values), self.cardinality, self.enumeration_size))
from pgmpy.sampling import BayesianModelSampling
import networkx as nx
import sys
from tqdm import tqdm
if __name__ == '__main__':
dataset = sys.argv[1]
sample_size = int(sys.argv[2])
n_runs = int(sys.argv[3])
score = sys.argv[4]
alpha = float(sys.argv[5])
palim = int(sys.argv[6])
# Read the BN model
reader = BIFReader(dataset + '.bif')
model = reader.get_model()
arities = dict(model.get_cardinality())
for key, value in arities.items():
arities[key] = [value]
nodes = list(nx.topological_sort(model))
for run in tqdm(range(n_runs)):
data = pd.read_csv(os.path.join(
dataset,
str(sample_size) + '_' + str(run) + '.csv'),
sep=' ')
data = data.iloc[1:] # Remove the arities from the table
if score == 'BDeu':
save_folder = os.path.join(
dataset,
str(alpha) + '_' + score + '_' + str(sample_size) + '_' +
class TestBIFReader(unittest.TestCase):
def setUp(self):
self.reader = BIFReader(string="""
// Bayesian Network in the Interchange Format
// Produced by BayesianNetworks package in JavaBayes
// Output created Sun Nov 02 17:49:49 GMT+00:00 1997
// Bayesian network
network "Dog-Problem" { //5 variables and 5 probability distributions
property "credal-set constant-density-bounded 1.1" ;
}
variable "light-on" { //2 values
type discrete[2] { "true" "false" };
property "position = (218, 195)" ;
}
variable "bowel-problem" { //2 values
type discrete[2] { "true" "false" };
property "position = (335, 99)" ;
}
variable "dog-out" { //2 values
type discrete[2] { "true" "false" };
property "position = (300, 195)" ;
}
variable "hear-bark" { //2 values
type discrete[2] { "true" "false" };
property "position = (296, 268)" ;
}
variable "family-out" { //2 values
type discrete[2] { "true" "false" };
property "position = (257, 99)" ;
}
probability ( "light-on" "family-out" ) { //2 variable(s) and 4 values
table 0.6 0.05 0.4 0.95 ;
}
probability ( "bowel-problem" ) { //1 variable(s) and 2 values
table 0.01 0.99 ;
}
probability ( "dog-out" "bowel-problem" "family-out" ) { //3 variable(s) and 8 values
table 0.99 0.97 0.9 0.3 0.01 0.03 0.1 0.7 ;
}
probability ( "hear-bark" "dog-out" ) { //2 variable(s) and 4 values
table 0.7 0.01 0.3 0.99 ;
}
probability ( "family-out" ) { //1 variable(s) and 2 values
table 0.15 0.85 ;
}
""")
def test_network_name(self):
name_expected = 'Dog-Problem'
self.assertEqual(self.reader.network_name, name_expected)
def test_get_variables(self):
var_expected = [
'light-on', 'bowel-problem', 'dog-out', 'hear-bark', 'family-out'
]
self.assertListEqual(self.reader.get_variables(), var_expected)
def test_states(self):
states_expected = {
'bowel-problem': ['true', 'false'],
'dog-out': ['true', 'false'],
'family-out': ['true', 'false'],
'hear-bark': ['true', 'false'],
'light-on': ['true', 'false']
}
states = self.reader.get_states()
for variable in states_expected:
self.assertListEqual(states_expected[variable], states[variable])
def test_get_property(self):
property_expected = {
'bowel-problem': ['position = (335, 99)'],
'dog-out': ['position = (300, 195)'],
'family-out': ['position = (257, 99)'],
'hear-bark': ['position = (296, 268)'],
'light-on': ['position = (218, 195)']
}
prop = self.reader.get_property()
for variable in property_expected:
self.assertListEqual(property_expected[variable], prop[variable])
def test_get_cpd(self):
cpd_expected = {
'bowel-problem': np.array([[0.01], [0.99]]),
'dog-out': np.array([[0.99, 0.97, 0.9, 0.3],
[0.01, 0.03, 0.1, 0.7]]),
'family-out': np.array([[0.15], [0.85]]),
'hear-bark': np.array([[0.7, 0.01], [0.3, 0.99]]),
'light-on': np.array([[0.6, 0.05], [0.4, 0.95]])
}
cpd = self.reader.variable_cpds
for variable in cpd_expected:
np_test.assert_array_equal(cpd_expected[variable], cpd[variable])
def test_get_parents(self):
parents_expected = {
'bowel-problem': [],
'dog-out': ['bowel-problem', 'family-out'],
'family-out': [],
'hear-bark': ['dog-out'],
'light-on': ['family-out']
}
parents = self.reader.get_parents()
for variable in parents_expected:
self.assertListEqual(parents_expected[variable], parents[variable])
def test_get_edges(self):
edges_expected = [['family-out', 'dog-out'],
['bowel-problem', 'dog-out'],
['family-out', 'light-on'], ['dog-out', 'hear-bark']]
self.assertListEqual(sorted(self.reader.variable_edges),
sorted(edges_expected))
def test_get_model(self):
edges_expected = [('family-out', 'dog-out'),
('bowel-problem', 'dog-out'),
('family-out', 'light-on'), ('dog-out', 'hear-bark')]
nodes_expected = [
'bowel-problem', 'hear-bark', 'light-on', 'dog-out', 'family-out'
]
edge_expected = {
'bowel-problem': {
'dog-out': {}
},
'dog-out': {
'hear-bark': {}
},
'family-out': {
'dog-out': {},
'light-on': {}
},
'hear-bark': {},
'light-on': {}
}
node_expected = {
'bowel-problem': {
'position': '(335, 99)'
},
'dog-out': {
'position': '(300, 195)'
},
'family-out': {
'position': '(257, 99)'
},
'hear-bark': {
'position': '(296, 268)'
},
'light-on': {
'position': '(218, 195)'
}
}
cpds_expected = [
np.array([[0.01], [0.99]]),
np.array([[0.99, 0.97, 0.9, 0.3], [0.01, 0.03, 0.1, 0.7]]),
np.array([[0.15], [0.85]]),
np.array([[0.7, 0.01], [0.3, 0.99]]),
np.array([[0.6, 0.05], [0.4, 0.95]])
]
model = self.reader.get_model()
for cpd_index in range(0, len(cpds_expected)):
np_test.assert_array_equal(model.get_cpds()[cpd_index].get_cpd(),
cpds_expected[cpd_index])
self.assertDictEqual(model.node, node_expected)
self.assertDictEqual(model.edge, edge_expected)
self.assertListEqual(sorted(model.nodes()), sorted(nodes_expected))
self.assertListEqual(sorted(model.edges()), sorted(edges_expected))
def tearDown(self):
del self.reader
def get_dataset(dataset):
"""
Fetches the `dataset` and returns a pgmpy.model instance.
Parameter
---------
dataset: str
Any dataset from bnlearn repository (http://www.bnlearn.com/bnrepository).
Discrete Bayesian Network Options:
Small Networks:
1. asia
2. cancer
3. earthquake
4. sachs
5. survey
Medium Networks:
1. alarm
2. barley
3. child
4. insurance
5. mildew
6. water
Large Networks:
1. hailfinder
2. hepar2
3. win95pts
Very Large Networks:
1. andes
2. diabetes
3. link
4. munin1
5. munin2
6. munin3
7. munin4
8. pathfinder
9. pigs
10. munin
Gaussian Bayesian Network Options:
1. ecoli70
2. magic-niab
3. magic-irri
4. arth150
Conditional Linear Gaussian Bayesian Network Options:
1. sangiovese
2. mehra
Example
-------
>>> from pgmpy.data import get_dataset
>>> model = get_dataset(dataset='asia')
>>> model
Returns
-------
pgmpy.models instance: An instance of one of the model classes in pgmpy.models
depending on the type of dataset.
"""
dataset_links = {
"asia": "http://www.bnlearn.com/bnrepository/asia/asia.bif.gz",
"cancer": "http://www.bnlearn.com/bnrepository/cancer/cancer.bif.gz",
"earthquake":
"http://www.bnlearn.com/bnrepository/earthquake/earthquake.bif.gz",
"sachs": "http://www.bnlearn.com/bnrepository/sachs/sachs.bif.gz",
"survey": "http://www.bnlearn.com/bnrepository/survey/survey.bif.gz",
"alarm": "http://www.bnlearn.com/bnrepository/alarm/alarm.bif.gz",
"barley": "http://www.bnlearn.com/bnrepository/barley/barley.bif.gz",
"child": "http://www.bnlearn.com/bnrepository/child/child.bif.gz",
"insurance":
"http://www.bnlearn.com/bnrepository/insurance/insurance.bif.gz",
"mildew": "http://www.bnlearn.com/bnrepository/mildew/mildew.bif.gz",
"water": "http://www.bnlearn.com/bnrepository/water/water.bif.gz",
"hailfinder":
"http://www.bnlearn.com/bnrepository/hailfinder/hailfinder.bif.gz",
"hepar2": "http://www.bnlearn.com/bnrepository/hepar2/hepar2.bif.gz",
"win95pts":
"http://www.bnlearn.com/bnrepository/win95pts/win95pts.bif.gz",
"andes": "http://www.bnlearn.com/bnrepository/andes/andes.bif.gz",
"diabetes":
"http://www.bnlearn.com/bnrepository/diabetes/diabetes.bif.gz",
"link": "http://www.bnlearn.com/bnrepository/link/link.bif.gz",
"munin1": "http://www.bnlearn.com/bnrepository/munin4/munin1.bif.gz",
"munin2": "http://www.bnlearn.com/bnrepository/munin4/munin2.bif.gz",
"munin3": "http://www.bnlearn.com/bnrepository/munin4/munin3.bif.gz",
"munin4": "http://www.bnlearn.com/bnrepository/munin4/munin4.bif.gz",
"pathfinder":
"http://www.bnlearn.com/bnrepository/pathfinder/pathfinder.bif.gz",
"pigs": "http://www.bnlearn.com/bnrepository/pigs/pigs.bif.gz",
"munin": "http://www.bnlearn.com/bnrepository/munin/munin.bif.gz",
"ecoli70": "",
"magic-niab": "",
"magic-irri": "",
"arth150": "",
"sangiovese": "",
"mehra": "",
}
if dataset not in dataset_links.keys():
raise ValueError("dataset should be one of the options")
if dataset_links[dataset] == "":
raise NotImplementedError("The specified dataset isn't supported")
filename, _ = urlretrieve(dataset_links[dataset])
with gzip.open(filename, "rb") as f:
content = f.read()
reader = BIFReader(content)
return reader.get_model()
class TestBIFReader(unittest.TestCase):
def setUp(self):
self.reader = BIFReader(string="""
// Bayesian Network in the Interchange Format
// Produced by BayesianNetworks package in JavaBayes
// Output created Sun Nov 02 17:49:49 GMT+00:00 1997
// Bayesian network
network "Dog-Problem" { //5 variables and 5 probability distributions
property "credal-set constant-density-bounded 1.1" ;
}
variable "light-on" { //2 values
type discrete[2] { "true" "false" };
property "position = (218, 195)" ;
}
variable "bowel-problem" { //2 values
type discrete[2] { "true" "false" };
property "position = (335, 99)" ;
}
variable "dog-out" { //2 values
type discrete[2] { "true" "false" };
property "position = (300, 195)" ;
}
variable "hear-bark" { //2 values
type discrete[2] { "true" "false" };
property "position = (296, 268)" ;
}
variable "family-out" { //2 values
type discrete[2] { "true" "false" };
property "position = (257, 99)" ;
}
probability ( "light-on" "family-out" ) { //2 variable(s) and 4 values
(true) 0.6 0.4 ;
(false) 0.05 0.95 ;
}
probability ( "bowel-problem" ) { //1 variable(s) and 2 values
table 0.01 0.99 ;
}
probability ( "dog-out" "bowel-problem" "family-out" ) { //3 variable(s) and 8 values
table 0.99 0.97 0.9 0.3 0.01 0.03 0.1 0.7 ;
}
probability ( "hear-bark" "dog-out" ) { //2 variable(s) and 4 values
table 0.7 0.01 0.3 0.99 ;
}
probability ( "family-out" ) { //1 variable(s) and 2 values
table 0.15 0.85 ;
}
""")
def test_network_name(self):
name_expected = 'Dog-Problem'
self.assertEqual(self.reader.network_name, name_expected)
def test_get_variables(self):
var_expected = ['light-on', 'bowel-problem', 'dog-out',
'hear-bark', 'family-out']
self.assertListEqual(self.reader.get_variables(), var_expected)
def test_states(self):
states_expected = {'bowel-problem': ['true', 'false'],
'dog-out': ['true', 'false'],
'family-out': ['true', 'false'],
'hear-bark': ['true', 'false'],
'light-on': ['true', 'false']}
states = self.reader.get_states()
for variable in states_expected:
self.assertListEqual(states_expected[variable], states[variable])
def test_get_property(self):
property_expected = {'bowel-problem': ['position = (335, 99)'],
'dog-out': ['position = (300, 195)'],
'family-out': ['position = (257, 99)'],
'hear-bark': ['position = (296, 268)'],
'light-on': ['position = (218, 195)']}
prop = self.reader.get_property()
for variable in property_expected:
self.assertListEqual(property_expected[variable],
prop[variable])
def test_get_values(self):
cpd_expected = {'bowel-problem': np.array([[0.01],
[0.99]]),
'dog-out': np.array([[0.99, 0.97, 0.9, 0.3],
[0.01, 0.03, 0.1, 0.7]]),
'family-out': np.array([[0.15],
[0.85]]),
'hear-bark': np.array([[0.7, 0.01],
[0.3, 0.99]]),
'light-on': np.array([[0.6, 0.05],
[0.4, 0.95]])}
cpd = self.reader.variable_cpds
for variable in cpd_expected:
np_test.assert_array_equal(cpd_expected[variable],
cpd[variable])
def test_get_parents(self):
parents_expected = {'bowel-problem': [],
'dog-out': ['bowel-problem', 'family-out'],
'family-out': [],
'hear-bark': ['dog-out'],
'light-on': ['family-out']}
parents = self.reader.get_parents()
for variable in parents_expected:
self.assertListEqual(parents_expected[variable],
parents[variable])
def test_get_edges(self):
edges_expected = [['family-out', 'dog-out'],
['bowel-problem', 'dog-out'],
['family-out', 'light-on'],
['dog-out', 'hear-bark']]
self.assertListEqual(sorted(self.reader.variable_edges),
sorted(edges_expected))
def test_get_model(self):
edges_expected = [('family-out', 'dog-out'),
('bowel-problem', 'dog-out'),
('family-out', 'light-on'),
('dog-out', 'hear-bark')]
nodes_expected = ['bowel-problem', 'hear-bark', 'light-on',
'dog-out', 'family-out']
edge_expected = {'bowel-problem': {'dog-out': {'weight': None}},
'dog-out': {'hear-bark': {'weight': None}},
'family-out': {'dog-out': {'weight': None},
'light-on': {'weight': None}},
'hear-bark': {},
'light-on': {}}
node_expected = {'bowel-problem': {'weight': None,
'position': '(335, 99)'},
'dog-out': {'weight': None,
'position': '(300, 195)'},
'family-out': {'weight': None,
'position': '(257, 99)'},
'hear-bark': {'weight': None,
'position': '(296, 268)'},
'light-on': {'weight': None,
'position': '(218, 195)'}}
cpds_expected = [np.array([[0.01],
[0.99]]),
np.array([[0.99, 0.97, 0.9, 0.3],
[0.01, 0.03, 0.1, 0.7]]),
np.array([[0.15],
[0.85]]),
np.array([[0.7, 0.01],
[0.3, 0.99]]),
np.array([[0.6, 0.05],
[0.4, 0.95]])]
model = self.reader.get_model()
for cpd_index in range(0, len(cpds_expected)):
np_test.assert_array_equal(model.get_cpds()[cpd_index].get_values(),
cpds_expected[cpd_index])
self.assertDictEqual(model.node, node_expected)
self.assertDictEqual(model.edge, edge_expected)
self.assertListEqual(sorted(model.nodes()), sorted(nodes_expected))
self.assertListEqual(sorted(model.edges()), sorted(edges_expected))
def tearDown(self):
del self.reader
state, c, states, costs = annealing(maxsteps=3000, debug=True)
nodes = ['asia', 'tub', 'smoke', 'lung', 'bronc', 'either', 'xray', 'dysp']
G = vetor_Rede(state, nodes)
nx.draw(G, with_labels=True)
print(state)
print(c)
with open('Asia.csv') as csv_file:
csv_reader = csv.reader(csv_file, delimiter=',')
aux = 0
data = []
data1 = [[] for i in range(8)]
for row in csv_reader:
data.append(row)
for i in range(len(row)):
data1[i].append(row[i])
aux = aux + 1
if aux == 50001:
break
#22376.39851240954
data = {}
for i in range(len(data1)):
data[data1[i][0]] = [data1[i][j] for j in range(1, len(data1[i]))]
data = pd.DataFrame(data)
print("Data: ")
print(data) #Dados Retirandos do arquivo
reader = BIFReader('asia.bif') # melhor rede do asia, como esta no bnlearn.com
asia_model = reader.get_model() # lendo esse modelo
print("Score BIC")
print(abs(BicScore(data).score(asia_model)))
#see_annealing(states, costs)
