class TestBayesianModelMethods(unittest.TestCase):
def setUp(self):
self.G = BayesianModel([('a', 'd'), ('b', 'd'),
('d', 'e'), ('b', 'c')])
def test_moral_graph(self):
moral_graph = self.G.moralize()
self.assertListEqual(sorted(moral_graph.nodes()), ['a', 'b', 'c', 'd', 'e'])
for edge in moral_graph.edges():
self.assertTrue(edge in [('a', 'b'), ('a', 'd'), ('b', 'c'), ('d', 'b'), ('e', 'd')] or
(edge[1], edge[0]) in [('a', 'b'), ('a', 'd'), ('b', 'c'), ('d', 'b'), ('e', 'd')])
def test_moral_graph_with_edge_present_over_parents(self):
G = BayesianModel([('a', 'd'), ('d', 'e'), ('b', 'd'), ('b', 'c'), ('a', 'b')])
moral_graph = G.moralize()
self.assertListEqual(sorted(moral_graph.nodes()), ['a', 'b', 'c', 'd', 'e'])
for edge in moral_graph.edges():
self.assertTrue(edge in [('a', 'b'), ('c', 'b'), ('d', 'a'), ('d', 'b'), ('d', 'e')] or
(edge[1], edge[0]) in [('a', 'b'), ('c', 'b'), ('d', 'a'), ('d', 'b'), ('d', 'e')])
def test_local_independencies(self):
self.assertEqual(self.G.local_independencies('a'), Independencies(['a', ['b', 'c']]))
self.assertEqual(self.G.local_independencies('c'), Independencies(['c',['a','d','e'],'b']))
self.assertEqual(self.G.local_independencies('d'), Independencies(['d','c',['b','a']]))
self.assertEqual(self.G.local_independencies('e'), Independencies(['e',['c','b','a'],'d']))
self.assertEqual(self.G.local_independencies('b'), Independencies(['b','a']))
def tearDown(self):
del self.G
class TestBayesianModelMethods(unittest.TestCase):
def setUp(self):
self.G = BayesianModel([('a', 'd'), ('b', 'd'), ('d', 'e'),
('b', 'c')])
def test_moral_graph(self):
moral_graph = self.G.moralize()
self.assertListEqual(sorted(moral_graph.nodes()),
['a', 'b', 'c', 'd', 'e'])
for edge in moral_graph.edges():
self.assertTrue(edge in [('a', 'b'), ('a', 'd'), ('b', 'c'),
('d', 'b'), ('e', 'd')]
or (edge[1], edge[0]) in [('a', 'b'), ('a', 'd'),
('b', 'c'), ('d', 'b'),
('e', 'd')])
def test_moral_graph_with_edge_present_over_parents(self):
G = BayesianModel([('a', 'd'), ('d', 'e'), ('b', 'd'), ('b', 'c'),
('a', 'b')])
moral_graph = G.moralize()
self.assertListEqual(sorted(moral_graph.nodes()),
['a', 'b', 'c', 'd', 'e'])
for edge in moral_graph.edges():
self.assertTrue(edge in [('a', 'b'), ('c', 'b'), ('d', 'a'),
('d', 'b'), ('d', 'e')]
or (edge[1], edge[0]) in [('a', 'b'), ('c', 'b'),
('d', 'a'), ('d', 'b'),
('d', 'e')])
def test_local_independencies(self):
self.assertEqual(self.G.local_independencies('a'),
Independencies(['a', ['b', 'c']]))
self.assertEqual(self.G.local_independencies('c'),
Independencies(['c', ['a', 'd', 'e'], 'b']))
self.assertEqual(self.G.local_independencies('d'),
Independencies(['d', 'c', ['b', 'a']]))
self.assertEqual(self.G.local_independencies('e'),
Independencies(['e', ['c', 'b', 'a'], 'd']))
self.assertEqual(self.G.local_independencies('b'),
Independencies(['b', 'a']))
def tearDown(self):
del self.G
class TestBayesianModelMethods(unittest.TestCase):
def setUp(self):
self.G = BayesianModel([('a', 'd'), ('b', 'd'), ('d', 'e'),
('b', 'c')])
self.G1 = BayesianModel([('diff', 'grade'), ('intel', 'grade')])
diff_cpd = TabularCPD('diff', 2, values=[[0.2], [0.8]])
intel_cpd = TabularCPD('intel', 3, values=[[0.5], [0.3], [0.2]])
grade_cpd = TabularCPD('grade',
3,
values=[[0.1, 0.1, 0.1, 0.1, 0.1, 0.1],
[0.1, 0.1, 0.1, 0.1, 0.1, 0.1],
[0.8, 0.8, 0.8, 0.8, 0.8, 0.8]],
evidence=['diff', 'intel'],
evidence_card=[2, 3])
self.G1.add_cpds(diff_cpd, intel_cpd, grade_cpd)
def test_moral_graph(self):
moral_graph = self.G.moralize()
self.assertListEqual(sorted(moral_graph.nodes()),
['a', 'b', 'c', 'd', 'e'])
for edge in moral_graph.edges():
self.assertTrue(edge in [('a', 'b'), ('a', 'd'), ('b', 'c'),
('d', 'b'), ('e', 'd')]
or (edge[1], edge[0]) in [('a', 'b'), ('a', 'd'),
('b', 'c'), ('d', 'b'),
('e', 'd')])
def test_moral_graph_with_edge_present_over_parents(self):
G = BayesianModel([('a', 'd'), ('d', 'e'), ('b', 'd'), ('b', 'c'),
('a', 'b')])
moral_graph = G.moralize()
self.assertListEqual(sorted(moral_graph.nodes()),
['a', 'b', 'c', 'd', 'e'])
for edge in moral_graph.edges():
self.assertTrue(edge in [('a', 'b'), ('c', 'b'), ('d', 'a'),
('d', 'b'), ('d', 'e')]
or (edge[1], edge[0]) in [('a', 'b'), ('c', 'b'),
('d', 'a'), ('d', 'b'),
('d', 'e')])
def test_local_independencies(self):
self.assertEqual(self.G.local_independencies('a'),
Independencies(['a', ['b', 'c']]))
self.assertEqual(self.G.local_independencies('c'),
Independencies(['c', ['a', 'd', 'e'], 'b']))
self.assertEqual(self.G.local_independencies('d'),
Independencies(['d', 'c', ['b', 'a']]))
self.assertEqual(self.G.local_independencies('e'),
Independencies(['e', ['c', 'b', 'a'], 'd']))
self.assertEqual(self.G.local_independencies('b'),
Independencies(['b', 'a']))
self.assertEqual(self.G1.local_independencies('grade'),
Independencies())
def test_get_independencies(self):
chain = BayesianModel([('X', 'Y'), ('Y', 'Z')])
self.assertEqual(chain.get_independencies(),
Independencies(('X', 'Z', 'Y'), ('Z', 'X', 'Y')))
fork = BayesianModel([('Y', 'X'), ('Y', 'Z')])
self.assertEqual(fork.get_independencies(),
Independencies(('X', 'Z', 'Y'), ('Z', 'X', 'Y')))
collider = BayesianModel([('X', 'Y'), ('Z', 'Y')])
self.assertEqual(collider.get_independencies(),
Independencies(('X', 'Z'), ('Z', 'X')))
def test_is_imap(self):
val = [
0.01, 0.01, 0.08, 0.006, 0.006, 0.048, 0.004, 0.004, 0.032, 0.04,
0.04, 0.32, 0.024, 0.024, 0.192, 0.016, 0.016, 0.128
]
JPD = JointProbabilityDistribution(['diff', 'intel', 'grade'],
[2, 3, 3], val)
fac = Factor(['diff', 'intel', 'grade'], [2, 3, 3], val)
self.assertTrue(self.G1.is_imap(JPD))
self.assertRaises(TypeError, self.G1.is_imap, fac)
def test_get_immoralities(self):
G = BayesianModel([('x', 'y'), ('z', 'y'), ('x', 'z'), ('w', 'y')])
self.assertEqual(G.get_immoralities(), {('w', 'x'), ('w', 'z')})
G1 = BayesianModel([('x', 'y'), ('z', 'y'), ('z', 'x'), ('w', 'y')])
self.assertEqual(G1.get_immoralities(), {('w', 'x'), ('w', 'z')})
G2 = BayesianModel([('x', 'y'), ('z', 'y'), ('x', 'z'), ('w', 'y'),
('w', 'x')])
self.assertEqual(G2.get_immoralities(), {('w', 'z')})
def test_is_iequivalent(self):
G = BayesianModel([('x', 'y'), ('z', 'y'), ('x', 'z'), ('w', 'y')])
self.assertRaises(TypeError, G.is_iequivalent, MarkovModel())
G1 = BayesianModel([('V', 'W'), ('W', 'X'), ('X', 'Y'), ('Z', 'Y')])
G2 = BayesianModel([('W', 'V'), ('X', 'W'), ('X', 'Y'), ('Z', 'Y')])
self.assertTrue(G1.is_iequivalent(G2))
G3 = BayesianModel([('W', 'V'), ('W', 'X'), ('Y', 'X'), ('Z', 'Y')])
self.assertFalse(G3.is_iequivalent(G2))
def tearDown(self):
del self.G
del self.G1
values=[[0.998], [0.002]])
cpd_alarm = TabularCPD(variable='Alarm', variable_card=2,
values=[[0.999, 0.71, 0.06, 0.05],
[0.001, 0.29, 0.94, 0.95]],
evidence=['Burglary', 'Earthquake'],
evidence_card=[2, 2])
cpd_johncalls = TabularCPD(variable='JohnCalls', variable_card=2,
values=[[0.95, 0.1], [0.05, 0.9]],
evidence=['Alarm'], evidence_card=[2])
cpd_marycalls = TabularCPD(variable='MaryCalls', variable_card=2,
values=[[0.1, 0.7], [0.9, 0.3]],
evidence=['Alarm'], evidence_card=[2])
# Associating the parameters with the model structure
alarm_model.add_cpds(cpd_burglary, cpd_earthquake, cpd_alarm, cpd_johncalls, cpd_marycalls)
#new cell
alarm_model.check_model()
#new cell
alarm_model.nodes()
#new cell
alarm_model.edges()
#new cell
alarm_model.local_independencies('Burglary')
#new cell
alarm_model.local_independencies('JohnCalls')
cpd_s = TabularCPD(variable='S',
variable_card=2,
values=[[0.95, 0.2], [0.05, 0.8]],
evidence=['I'],
evidence_card=[2])
# 将有向无环图与条件概率分布表关联
model.add_cpds(cpd_d, cpd_i, cpd_g, cpd_l, cpd_s)
# 验证模型：检查网络结构和CPD，并验证CPD是否正确定义和总和为1
model.check_model()
# 获取概率图模型
model.get_cpds()
# 获取节点G的概率表
#print(model.get_cpds('G'))
# 获取节点G的基数
model.get_cardinality('G')
# 获取整个贝叶斯网络的局部依赖
model.local_independencies(['D', 'I', 'S', 'G', 'L'])
from pgmpy.inference import VariableElimination
infer = VariableElimination(model)
# 边缘化其他变量，求某一变量的概率
print(infer.query(['G'])['G'])
# 计算条件概率分布
print(infer.query(['G'], evidence={'D': 1, 'I': 1})['G'])
print(111, infer.query(['G'], evidence={'I': 1, 'L': 1, 'D': 1})['G'])
# 对于给定条件的变量状态进行预测
print(infer.map_query('G'))
print(infer.map_query('G', evidence={'D': 0, 'I': 1}))
print(infer.map_query('G', evidence={'D': 0, 'I': 1, 'L': 1, 'S': 1}))
def localIndependencySynonyms(model: BayesianModel,
query: RandomVariable,
useNotation=False) -> List[Name]:
'''
Generates all possible equivalent independencies, given a query node and separator nodes.
For example, for the independency (G _|_ S, L | I, D), all possible equivalent independencies are made by permuting the letters S, L and I, D in their positions. An resulting equivalent independency would then be (G _|_ L, S | I, D) or (G _|_ L, S | D, I) etc.
Arguments:
query: the node from which local independencies are to be calculated.
condNodes: either List[str] or List[List[str]].
---> When it is List[str], it contains a list of nodes that are only after the conditional | sign. For instance, for (D _|_ G,S,L,I), the otherNodes = ['D','S','L','I'].
---> when it is List[List[str]], otherNodes contains usually two elements, the list of nodes BEFORE and AFTER the conditional | sign. For instance, for (G _|_ L, S | I, D), otherNodes = [ ['L','S'], ['I','D'] ], where the nodes before the conditional sign are L,S and the nodes after the conditional sign are I, D.
Returns:
List of generated string independency combinations.
'''
# First check that the query node has local independencies!
# TODO check how to match up with the otherNodes argument
if model.local_independencies(query.var) == Independencies():
return
locIndeps = model.local_independencies(query.var)
_, condExpr = str(locIndeps).split('_|_')
condNodes: List[List[Name]] = []
if "|" in condExpr:
beforeCond, afterCond = condExpr.split("|")
# Removing the paranthesis after the last letter:
afterCond = afterCond[0:len(afterCond) - 1]
beforeCondList: List[Name] = list(
map(lambda letter: letter.strip(), beforeCond.split(",")))
afterCondList: List[Name] = list(
map(lambda letter: letter.strip(), afterCond.split(",")))
condNodes: List[List[Name]] = [beforeCondList] + [afterCondList]
else: # just have an expr like "leters" that are only before cond
beforeCond = condExpr[0:len(condExpr) - 1]
beforeCondList: List[Name] = list(
map(lambda letter: letter.strip(), beforeCond.split(",")))
condNodes: List[List[Name]] = [beforeCondList]
otherComboStrList = []
for letterSet in condNodes:
# NOTE: could use comma here instead of the '∩' (and) symbol
if useNotation: # use 'set and' symbol and brackets (set notation, clearer than simple notation)
comboStrs: List[str] = list(
map(
lambda letterCombo: "{" + ' ∩ '.join(letterCombo) + "}"
if len(letterCombo) > 1 else ' ∩ '.join(letterCombo),
itertools.permutations(letterSet)))
else: # use commas and no brackets (simple notation)
comboStrs: List[str] = list(
map(lambda letterCombo: ', '.join(letterCombo),
itertools.permutations(letterSet)))
# Add this particular combination of letters (variables) to the list.
otherComboStrList.append(comboStrs)
# Do product of the after-before variable string combinations.
# (For instance, given the list [['S,L', 'L,S'], ['D,I', 'I,D']], this operation returns the product list: [('S,L', 'D,I'), ('S,L', 'I,D'), ('L,S', 'D,I'), ('L,S', 'I,D')]
condComboStr: List[Tuple[Name]] = list(
itertools.product(*otherComboStrList))
# Joining the individual strings in the tuples (above) with conditional sign '|'
condComboStr: List[str] = list(
map(lambda condPair: ' | '.join(condPair), condComboStr))
independencyCombos: List[str] = list(
map(lambda letterComboStr: f"({query.var} _|_ {letterComboStr})",
condComboStr))
return independencyCombos
def bayesnet():
"""
References:
https://class.coursera.org/pgm-003/lecture/17
http://www.cs.ubc.ca/~murphyk/Bayes/bnintro.html
http://www3.cs.stonybrook.edu/~sael/teaching/cse537/Slides/chapter14d_BP.pdf
http://www.cse.unsw.edu.au/~cs9417ml/Bayes/Pages/PearlPropagation.html
https://github.com/pgmpy/pgmpy.git
http://pgmpy.readthedocs.org/en/latest/
http://nipy.bic.berkeley.edu:5000/download/11
"""
# import operator as op
# # Enumerate all possible events
# varcard_list = list(map(op.attrgetter('variable_card'), cpd_list))
# _esdat = list(ut.iprod(*map(range, varcard_list)))
# _escol = list(map(op.attrgetter('variable'), cpd_list))
# event_space = pd.DataFrame(_esdat, columns=_escol)
# # Custom compression of event space to inspect a specific graph
# def compress_space_flags(event_space, var1, var2, var3, cmp12_):
# """
# var1, var2, cmp_ = 'Lj', 'Lk', op.eq
# """
# import vtool as vt
# data = event_space
# other_cols = ut.setdiff_ordered(data.columns.tolist(), [var1, var2, var3])
# case_flags12 = cmp12_(data[var1], data[var2]).values
# # case_flags23 = cmp23_(data[var2], data[var3]).values
# # case_flags = np.logical_and(case_flags12, case_flags23)
# case_flags = case_flags12
# case_flags = case_flags.astype(np.int64)
# subspace = np.hstack((case_flags[:, None], data[other_cols].values))
# sel_ = vt.unique_row_indexes(subspace)
# flags = np.logical_and(mask, case_flags)
# return flags
# # Build special cases
# case_same = event_space.loc[compress_space_flags(event_space, 'Li', 'Lj', 'Lk', op.eq)]
# case_diff = event_space.loc[compress_space_flags(event_space, 'Li', 'Lj', 'Lk', op.ne)]
# special_cases = [
# case_same,
# case_diff,
# ]
from pgmpy.factors import TabularCPD
from pgmpy.models import BayesianModel
import pandas as pd
from pgmpy.inference import BeliefPropagation # NOQA
from pgmpy.inference import VariableElimination # NOQA
name_nice = ['n1', 'n2', 'n3']
score_nice = ['low', 'high']
match_nice = ['diff', 'same']
num_names = len(name_nice)
num_scores = len(score_nice)
nid_basis = list(range(num_names))
score_basis = list(range(num_scores))
semtype2_nice = {
'score': score_nice,
'name': name_nice,
'match': match_nice,
}
var2_cpd = {
}
globals()['semtype2_nice'] = semtype2_nice
globals()['var2_cpd'] = var2_cpd
name_combo = np.array(list(ut.iprod(nid_basis, nid_basis)))
combo_is_same = name_combo.T[0] == name_combo.T[1]
def get_expected_scores_prob(level1, level2):
part1 = combo_is_same * level1
part2 = (1 - combo_is_same) * (1 - (level2))
expected_scores_level = part1 + part2
return expected_scores_level
# def make_cpd():
def name_cpd(aid):
from pgmpy.factors import TabularCPD
cpd = TabularCPD(
variable='N' + aid,
variable_card=num_names,
values=[[1.0 / num_names] * num_names])
cpd.semtype = 'name'
return cpd
name_cpds = [name_cpd('i'), name_cpd('j'), name_cpd('k')]
var2_cpd.update(dict(zip([cpd.variable for cpd in name_cpds], name_cpds)))
if True:
num_same_diff = 2
samediff_measure = np.array([
# get_expected_scores_prob(.12, .2),
# get_expected_scores_prob(.88, .8),
get_expected_scores_prob(0, 0),
get_expected_scores_prob(1, 1),
])
samediff_vals = (samediff_measure / samediff_measure.sum(axis=0)).tolist()
def samediff_cpd(aid1, aid2):
cpd = TabularCPD(
variable='A' + aid1 + aid2,
variable_card=num_same_diff,
values=samediff_vals,
evidence=['N' + aid1, 'N' + aid2], # [::-1],
evidence_card=[num_names, num_names]) # [::-1])
cpd.semtype = 'match'
return cpd
samediff_cpds = [samediff_cpd('i', 'j'), samediff_cpd('j', 'k'), samediff_cpd('k', 'i')]
var2_cpd.update(dict(zip([cpd.variable for cpd in samediff_cpds], samediff_cpds)))
if True:
def score_cpd(aid1, aid2):
semtype = 'score'
evidence = ['A' + aid1 + aid2, 'N' + aid1, 'N' + aid2]
evidence_cpds = [var2_cpd[key] for key in evidence]
evidence_nice = [semtype2_nice[cpd.semtype] for cpd in evidence_cpds]
evidence_card = list(map(len, evidence_nice))
evidence_states = list(ut.iprod(*evidence_nice))
variable_basis = semtype2_nice[semtype]
variable_values = []
for mystate in variable_basis:
row = []
for state in evidence_states:
if state[0] == state[1]:
if state[2] == 'same':
val = .2 if mystate == 'low' else .8
else:
val = 1
# val = .5 if mystate == 'low' else .5
elif state[0] != state[1]:
if state[2] == 'same':
val = .5 if mystate == 'low' else .5
else:
val = 1
# val = .9 if mystate == 'low' else .1
row.append(val)
variable_values.append(row)
cpd = TabularCPD(
variable='S' + aid1 + aid2,
variable_card=len(variable_basis),
values=variable_values,
evidence=evidence, # [::-1],
evidence_card=evidence_card) # [::-1])
cpd.semtype = semtype
return cpd
else:
score_values = [
[.8, .1],
[.2, .9],
]
def score_cpd(aid1, aid2):
cpd = TabularCPD(
variable='S' + aid1 + aid2,
variable_card=num_scores,
values=score_values,
evidence=['A' + aid1 + aid2], # [::-1],
evidence_card=[num_same_diff]) # [::-1])
cpd.semtype = 'score'
return cpd
score_cpds = [score_cpd('i', 'j'), score_cpd('j', 'k')]
cpd_list = name_cpds + score_cpds + samediff_cpds
else:
score_measure = np.array([get_expected_scores_prob(level1, level2)
for level1, level2 in
zip(np.linspace(.1, .9, num_scores),
np.linspace(.2, .8, num_scores))])
score_values = (score_measure / score_measure.sum(axis=0)).tolist()
def score_cpd(aid1, aid2):
cpd = TabularCPD(
variable='S' + aid1 + aid2,
variable_card=num_scores,
values=score_values,
evidence=['N' + aid1, 'N' + aid2],
evidence_card=[num_names, num_names])
cpd.semtype = 'score'
return cpd
score_cpds = [score_cpd('i', 'j'), score_cpd('j', 'k')]
cpd_list = name_cpds + score_cpds
pass
input_graph = []
for cpd in cpd_list:
if cpd.evidence is not None:
for evar in cpd.evidence:
input_graph.append((evar, cpd.variable))
name_model = BayesianModel(input_graph)
name_model.add_cpds(*cpd_list)
var2_cpd.update(dict(zip([cpd.variable for cpd in cpd_list], cpd_list)))
globals()['var2_cpd'] = var2_cpd
varnames = [cpd.variable for cpd in cpd_list]
# --- PRINT CPDS ---
cpd = score_cpds[0]
def print_cpd(cpd):
print('CPT: %r' % (cpd,))
index = semtype2_nice[cpd.semtype]
if cpd.evidence is None:
columns = ['None']
else:
basis_lists = [semtype2_nice[var2_cpd[ename].semtype] for ename in cpd.evidence]
columns = [','.join(x) for x in ut.iprod(*basis_lists)]
data = cpd.get_cpd()
print(pd.DataFrame(data, index=index, columns=columns))
for cpd in name_model.get_cpds():
print('----')
print(cpd._str('phi'))
print_cpd(cpd)
# --- INFERENCE ---
Ni = name_cpds[0]
event_space_combos = {}
event_space_combos[Ni.variable] = 0 # Set ni to always be Fred
for cpd in cpd_list:
if cpd.semtype == 'score':
event_space_combos[cpd.variable] = list(range(cpd.variable_card))
evidence_dict = ut.all_dict_combinations(event_space_combos)
# Query about name of annotation k given different event space params
def pretty_evidence(evidence):
return [key + '=' + str(semtype2_nice[var2_cpd[key].semtype][val])
for key, val in evidence.items()]
def print_factor(factor):
row_cards = factor.cardinality
row_vars = factor.variables
values = factor.values.reshape(np.prod(row_cards), 1).flatten()
# col_cards = 1
# col_vars = ['']
basis_lists = list(zip(*list(ut.iprod(*[range(c) for c in row_cards]))))
nice_basis_lists = []
for varname, basis in zip(row_vars, basis_lists):
cpd = var2_cpd[varname]
_nice_basis = ut.take(semtype2_nice[cpd.semtype], basis)
nice_basis = ['%s=%s' % (varname, val) for val in _nice_basis]
nice_basis_lists.append(nice_basis)
row_lbls = [', '.join(sorted(x)) for x in zip(*nice_basis_lists)]
print(ut.repr3(dict(zip(row_lbls, values)), precision=3, align=True, key_order_metric='-val'))
# name_belief = BeliefPropagation(name_model)
name_belief = VariableElimination(name_model)
import pgmpy
import six # NOQA
def try_query(evidence):
print('--------')
query_vars = ut.setdiff_ordered(varnames, list(evidence.keys()))
evidence_str = ', '.join(pretty_evidence(evidence))
probs = name_belief.query(query_vars, evidence)
factor_list = probs.values()
joint_factor = pgmpy.factors.factor_product(*factor_list)
print('P(' + ', '.join(query_vars) + ' | ' + evidence_str + ')')
# print(six.text_type(joint_factor))
factor = joint_factor # NOQA
# print_factor(factor)
# import utool as ut
print(ut.hz_str([(f._str(phi_or_p='phi')) for f in factor_list]))
for evidence in evidence_dict:
try_query(evidence)
evidence = {'Aij': 1, 'Ajk': 1, 'Aki': 1, 'Ni': 0}
try_query(evidence)
evidence = {'Aij': 0, 'Ajk': 0, 'Aki': 0, 'Ni': 0}
try_query(evidence)
globals()['score_nice'] = score_nice
globals()['name_nice'] = name_nice
globals()['score_basis'] = score_basis
globals()['nid_basis'] = nid_basis
print('Independencies')
print(name_model.get_independencies())
print(name_model.local_independencies([Ni.variable]))
# name_belief = BeliefPropagation(name_model)
# # name_belief = VariableElimination(name_model)
# for case in special_cases:
# test_data = case.drop('Lk', axis=1)
# test_data = test_data.reset_index(drop=True)
# print('----')
# for i in range(test_data.shape[0]):
# evidence = test_data.loc[i].to_dict()
# probs = name_belief.query(['Lk'], evidence)
# factor = probs['Lk']
# probs = factor.values
# evidence_ = evidence.copy()
# evidence_['Li'] = name_nice[evidence['Li']]
# evidence_['Lj'] = name_nice[evidence['Lj']]
# evidence_['Sij'] = score_nice[evidence['Sij']]
# evidence_['Sjk'] = score_nice[evidence['Sjk']]
# nice2_prob = ut.odict(zip(name_nice, probs.tolist()))
# ut.print_python_code('P(Lk | {evidence}) = {cpt}'.format(
# evidence=(ut.repr2(evidence_, explicit=True, nobraces=True, strvals=True)),
# cpt=ut.repr3(nice2_prob, precision=3, align=True, key_order_metric='-val')
# ))
# for case in special_cases:
# test_data = case.drop('Lk', axis=1)
# test_data = test_data.drop('Lj', axis=1)
# test_data = test_data.reset_index(drop=True)
# print('----')
# for i in range(test_data.shape[0]):
# evidence = test_data.loc[i].to_dict()
# query_vars = ['Lk', 'Lj']
# probs = name_belief.query(query_vars, evidence)
# for queryvar in query_vars:
# factor = probs[queryvar]
# print(factor._str('phi'))
# probs = factor.values
# evidence_ = evidence.copy()
# evidence_['Li'] = name_nice[evidence['Li']]
# evidence_['Sij'] = score_nice[evidence['Sij']]
# evidence_['Sjk'] = score_nice[evidence['Sjk']]
# nice2_prob = ut.odict(zip([queryvar + '=' + x for x in name_nice], probs.tolist()))
# ut.print_python_code('P({queryvar} | {evidence}) = {cpt}'.format(
# query_var=query_var,
# evidence=(ut.repr2(evidence_, explicit=True, nobraces=True, strvals=True)),
# cpt=ut.repr3(nice2_prob, precision=3, align=True, key_order_metric='-val')
# ))
# _ draw model
import plottool as pt
import networkx as netx
fig = pt.figure() # NOQA
fig.clf()
ax = pt.gca()
netx_nodes = [(node, {}) for node in name_model.nodes()]
netx_edges = [(etup[0], etup[1], {}) for etup in name_model.edges()]
netx_graph = netx.DiGraph()
netx_graph.add_nodes_from(netx_nodes)
netx_graph.add_edges_from(netx_edges)
# pos = netx.graphviz_layout(netx_graph)
pos = netx.pydot_layout(netx_graph, prog='dot')
netx.draw(netx_graph, pos=pos, ax=ax, with_labels=True)
pt.plt.savefig('foo.png')
ut.startfile('foo.png')
class TestBayesianModelMethods(unittest.TestCase):
def setUp(self):
self.G = BayesianModel([('a', 'd'), ('b', 'd'),
('d', 'e'), ('b', 'c')])
self.G1 = BayesianModel([('diff', 'grade'), ('intel', 'grade')])
diff_cpd = TabularCPD('diff', 2, values=[[0.2], [0.8]])
intel_cpd = TabularCPD('intel', 3, values=[[0.5], [0.3], [0.2]])
grade_cpd = TabularCPD('grade', 3, values=[[0.1, 0.1, 0.1, 0.1, 0.1, 0.1],
[0.1, 0.1, 0.1, 0.1, 0.1, 0.1],
[0.8, 0.8, 0.8, 0.8, 0.8, 0.8]],
evidence=['diff', 'intel'], evidence_card=[2, 3])
self.G1.add_cpds(diff_cpd, intel_cpd, grade_cpd)
self.G2 = BayesianModel([('d', 'g'), ('g', 'l'), ('i', 'g'), ('i', 'l')])
def test_moral_graph(self):
moral_graph = self.G.moralize()
self.assertListEqual(sorted(moral_graph.nodes()), ['a', 'b', 'c', 'd', 'e'])
for edge in moral_graph.edges():
self.assertTrue(edge in [('a', 'b'), ('a', 'd'), ('b', 'c'), ('d', 'b'), ('e', 'd')] or
(edge[1], edge[0]) in [('a', 'b'), ('a', 'd'), ('b', 'c'), ('d', 'b'), ('e', 'd')])
def test_moral_graph_with_edge_present_over_parents(self):
G = BayesianModel([('a', 'd'), ('d', 'e'), ('b', 'd'), ('b', 'c'), ('a', 'b')])
moral_graph = G.moralize()
self.assertListEqual(sorted(moral_graph.nodes()), ['a', 'b', 'c', 'd', 'e'])
for edge in moral_graph.edges():
self.assertTrue(edge in [('a', 'b'), ('c', 'b'), ('d', 'a'), ('d', 'b'), ('d', 'e')] or
(edge[1], edge[0]) in [('a', 'b'), ('c', 'b'), ('d', 'a'), ('d', 'b'), ('d', 'e')])
def test_get_ancestors_of_success(self):
ancenstors1 = self.G2._get_ancestors_of('g')
ancenstors2 = self.G2._get_ancestors_of('d')
ancenstors3 = self.G2._get_ancestors_of(['i', 'l'])
self.assertEqual(ancenstors1, {'d', 'i', 'g'})
self.assertEqual(ancenstors2, {'d'})
self.assertEqual(ancenstors3, {'g', 'i', 'l', 'd'})
def test_get_ancestors_of_failure(self):
self.assertRaises(ValueError, self.G2._get_ancestors_of, 'h')
def test_get_cardinality(self):
self.assertDictEqual(self.G1.get_cardinality(), {'diff': 2, 'intel': 3, 'grade': 3})
def test_get_cardinality_with_node(self):
self.assertEqual(self.G1.get_cardinality('diff'), 2)
self.assertEqual(self.G1.get_cardinality('intel'), 3)
self.assertEqual(self.G1.get_cardinality('grade'), 3)
def test_local_independencies(self):
self.assertEqual(self.G.local_independencies('a'), Independencies(['a', ['b', 'c']]))
self.assertEqual(self.G.local_independencies('c'), Independencies(['c', ['a', 'd', 'e'], 'b']))
self.assertEqual(self.G.local_independencies('d'), Independencies(['d', 'c', ['b', 'a']]))
self.assertEqual(self.G.local_independencies('e'), Independencies(['e', ['c', 'b', 'a'], 'd']))
self.assertEqual(self.G.local_independencies('b'), Independencies(['b', 'a']))
self.assertEqual(self.G1.local_independencies('grade'), Independencies())
def test_get_independencies(self):
chain = BayesianModel([('X', 'Y'), ('Y', 'Z')])
self.assertEqual(chain.get_independencies(), Independencies(('X', 'Z', 'Y'), ('Z', 'X', 'Y')))
fork = BayesianModel([('Y', 'X'), ('Y', 'Z')])
self.assertEqual(fork.get_independencies(), Independencies(('X', 'Z', 'Y'), ('Z', 'X', 'Y')))
collider = BayesianModel([('X', 'Y'), ('Z', 'Y')])
self.assertEqual(collider.get_independencies(), Independencies(('X', 'Z'), ('Z', 'X')))
def test_is_imap(self):
val = [0.01, 0.01, 0.08, 0.006, 0.006, 0.048, 0.004, 0.004, 0.032,
0.04, 0.04, 0.32, 0.024, 0.024, 0.192, 0.016, 0.016, 0.128]
JPD = JointProbabilityDistribution(['diff', 'intel', 'grade'], [2, 3, 3], val)
fac = DiscreteFactor(['diff', 'intel', 'grade'], [2, 3, 3], val)
self.assertTrue(self.G1.is_imap(JPD))
self.assertRaises(TypeError, self.G1.is_imap, fac)
def test_markov_blanet(self):
G = BayesianModel([('x', 'y'), ('z', 'y'), ('y', 'w'), ('y', 'v'), ('u', 'w'),
('s', 'v'), ('w', 't'), ('w', 'm'), ('v', 'n'), ('v', 'q')])
self.assertEqual(set(G.get_markov_blanket('y')), set(['s', 'w', 'x', 'u', 'z', 'v']))
def test_get_immoralities(self):
G = BayesianModel([('x', 'y'), ('z', 'y'), ('x', 'z'), ('w', 'y')])
self.assertEqual(G.get_immoralities(), {('w', 'x'), ('w', 'z')})
G1 = BayesianModel([('x', 'y'), ('z', 'y'), ('z', 'x'), ('w', 'y')])
self.assertEqual(G1.get_immoralities(), {('w', 'x'), ('w', 'z')})
G2 = BayesianModel([('x', 'y'), ('z', 'y'), ('x', 'z'), ('w', 'y'), ('w', 'x')])
self.assertEqual(G2.get_immoralities(), {('w', 'z')})
def test_is_iequivalent(self):
G = BayesianModel([('x', 'y'), ('z', 'y'), ('x', 'z'), ('w', 'y')])
self.assertRaises(TypeError, G.is_iequivalent, MarkovModel())
G1 = BayesianModel([('V', 'W'), ('W', 'X'), ('X', 'Y'), ('Z', 'Y')])
G2 = BayesianModel([('W', 'V'), ('X', 'W'), ('X', 'Y'), ('Z', 'Y')])
self.assertTrue(G1.is_iequivalent(G2))
G3 = BayesianModel([('W', 'V'), ('W', 'X'), ('Y', 'X'), ('Z', 'Y')])
self.assertFalse(G3.is_iequivalent(G2))
def test_copy(self):
model_copy = self.G1.copy()
self.assertEqual(sorted(self.G1.nodes()), sorted(model_copy.nodes()))
self.assertEqual(sorted(self.G1.edges()), sorted(model_copy.edges()))
self.assertNotEqual(id(self.G1.get_cpds('diff')),
id(model_copy.get_cpds('diff')))
self.G1.remove_cpds('diff')
diff_cpd = TabularCPD('diff', 2, values=[[0.3], [0.7]])
self.G1.add_cpds(diff_cpd)
self.assertNotEqual(self.G1.get_cpds('diff'),
model_copy.get_cpds('diff'))
self.G1.remove_node('intel')
self.assertNotEqual(sorted(self.G1.nodes()), sorted(model_copy.nodes()))
self.assertNotEqual(sorted(self.G1.edges()), sorted(model_copy.edges()))
def test_remove_node(self):
self.G1.remove_node('diff')
self.assertEqual(sorted(self.G1.nodes()), sorted(['grade', 'intel']))
self.assertRaises(ValueError, self.G1.get_cpds, 'diff')
def test_remove_nodes_from(self):
self.G1.remove_nodes_from(['diff', 'grade'])
self.assertEqual(sorted(self.G1.nodes()), sorted(['intel']))
self.assertRaises(ValueError, self.G1.get_cpds, 'diff')
self.assertRaises(ValueError, self.G1.get_cpds, 'grade')
def tearDown(self):
del self.G
del self.G1
values=[[0.9], [0.1]])
cpd_smoke = TabularCPD(variable='Smoker',
variable_card=2,
values=[[0.3], [0.7]])
cpd_cancer = TabularCPD(variable='Cancer',
variable_card=2,
values=[[0.03, 0.05, 0.001, 0.02],
[0.97, 0.95, 0.999, 0.98]],
evidence=['Smoker', 'Pollution'],
evidence_card=[2, 2])
cpd_xray = TabularCPD(variable='Xray',
variable_card=2,
values=[[0.9, 0.2], [0.1, 0.8]],
evidence=['Cancer'],
evidence_card=[2])
cpd_dysp = TabularCPD(variable='Dyspnoea',
variable_card=2,
values=[[0.65, 0.3], [0.35, 0.7]],
evidence=['Cancer'],
evidence_card=[2])
# Associating the parameters with the model structure.
cancer_model.add_cpds(cpd_poll, cpd_smoke, cpd_cancer, cpd_xray, cpd_dysp)
# Checking if the cpds are valid for the model.
print(cancer_model.check_model())
# Doing some simple queries on the network
cancer_model.is_active_trail('Pollution', 'Smoker')
cancer_model.is_active_trail('Pollution', 'Smoker', observed=['Cancer'])
cancer_model.local_independencies('Xray')
class TestBayesianModelMethods(unittest.TestCase):
def setUp(self):
self.G = BayesianModel([("a", "d"), ("b", "d"), ("d", "e"),
("b", "c")])
self.G1 = BayesianModel([("diff", "grade"), ("intel", "grade")])
diff_cpd = TabularCPD("diff", 2, values=[[0.2], [0.8]])
intel_cpd = TabularCPD("intel", 3, values=[[0.5], [0.3], [0.2]])
grade_cpd = TabularCPD(
"grade",
3,
values=[
[0.1, 0.1, 0.1, 0.1, 0.1, 0.1],
[0.1, 0.1, 0.1, 0.1, 0.1, 0.1],
[0.8, 0.8, 0.8, 0.8, 0.8, 0.8],
],
evidence=["diff", "intel"],
evidence_card=[2, 3],
)
self.G1.add_cpds(diff_cpd, intel_cpd, grade_cpd)
self.G2 = BayesianModel([("d", "g"), ("g", "l"), ("i", "g"),
("i", "l")])
def test_moral_graph(self):
moral_graph = self.G.moralize()
self.assertListEqual(sorted(moral_graph.nodes()),
["a", "b", "c", "d", "e"])
for edge in moral_graph.edges():
self.assertTrue(edge in [("a", "b"), ("a", "d"), ("b", "c"),
("d", "b"), ("e", "d")]
or (edge[1], edge[0]) in [("a", "b"), ("a", "d"),
("b", "c"), ("d", "b"),
("e", "d")])
def test_moral_graph_with_edge_present_over_parents(self):
G = BayesianModel([("a", "d"), ("d", "e"), ("b", "d"), ("b", "c"),
("a", "b")])
moral_graph = G.moralize()
self.assertListEqual(sorted(moral_graph.nodes()),
["a", "b", "c", "d", "e"])
for edge in moral_graph.edges():
self.assertTrue(edge in [("a", "b"), ("c", "b"), ("d", "a"),
("d", "b"), ("d", "e")]
or (edge[1], edge[0]) in [("a", "b"), ("c", "b"),
("d", "a"), ("d", "b"),
("d", "e")])
def test_get_ancestors_of_success(self):
ancenstors1 = self.G2._get_ancestors_of("g")
ancenstors2 = self.G2._get_ancestors_of("d")
ancenstors3 = self.G2._get_ancestors_of(["i", "l"])
self.assertEqual(ancenstors1, {"d", "i", "g"})
self.assertEqual(ancenstors2, {"d"})
self.assertEqual(ancenstors3, {"g", "i", "l", "d"})
def test_get_ancestors_of_failure(self):
self.assertRaises(ValueError, self.G2._get_ancestors_of, "h")
def test_get_cardinality(self):
self.assertDictEqual(self.G1.get_cardinality(), {
"diff": 2,
"intel": 3,
"grade": 3
})
def test_get_cardinality_with_node(self):
self.assertEqual(self.G1.get_cardinality("diff"), 2)
self.assertEqual(self.G1.get_cardinality("intel"), 3)
self.assertEqual(self.G1.get_cardinality("grade"), 3)
def test_local_independencies(self):
self.assertEqual(self.G.local_independencies("a"),
Independencies(["a", ["b", "c"]]))
self.assertEqual(
self.G.local_independencies("c"),
Independencies(["c", ["a", "d", "e"], "b"]),
)
self.assertEqual(self.G.local_independencies("d"),
Independencies(["d", "c", ["b", "a"]]))
self.assertEqual(
self.G.local_independencies("e"),
Independencies(["e", ["c", "b", "a"], "d"]),
)
self.assertEqual(self.G.local_independencies("b"),
Independencies(["b", "a"]))
self.assertEqual(self.G1.local_independencies("grade"),
Independencies())
def test_get_independencies(self):
chain = BayesianModel([("X", "Y"), ("Y", "Z")])
self.assertEqual(chain.get_independencies(),
Independencies(("X", "Z", "Y"), ("Z", "X", "Y")))
fork = BayesianModel([("Y", "X"), ("Y", "Z")])
self.assertEqual(fork.get_independencies(),
Independencies(("X", "Z", "Y"), ("Z", "X", "Y")))
collider = BayesianModel([("X", "Y"), ("Z", "Y")])
self.assertEqual(collider.get_independencies(),
Independencies(("X", "Z"), ("Z", "X")))
def test_is_imap(self):
val = [
0.01,
0.01,
0.08,
0.006,
0.006,
0.048,
0.004,
0.004,
0.032,
0.04,
0.04,
0.32,
0.024,
0.024,
0.192,
0.016,
0.016,
0.128,
]
JPD = JointProbabilityDistribution(["diff", "intel", "grade"],
[2, 3, 3], val)
fac = DiscreteFactor(["diff", "intel", "grade"], [2, 3, 3], val)
self.assertTrue(self.G1.is_imap(JPD))
self.assertRaises(TypeError, self.G1.is_imap, fac)
def test_markov_blanet(self):
G = DAG([
("x", "y"),
("z", "y"),
("y", "w"),
("y", "v"),
("u", "w"),
("s", "v"),
("w", "t"),
("w", "m"),
("v", "n"),
("v", "q"),
])
self.assertEqual(set(G.get_markov_blanket("y")),
set(["s", "w", "x", "u", "z", "v"]))
def test_get_immoralities(self):
G = BayesianModel([("x", "y"), ("z", "y"), ("x", "z"), ("w", "y")])
self.assertEqual(G.get_immoralities(), {("w", "x"), ("w", "z")})
G1 = BayesianModel([("x", "y"), ("z", "y"), ("z", "x"), ("w", "y")])
self.assertEqual(G1.get_immoralities(), {("w", "x"), ("w", "z")})
G2 = BayesianModel([("x", "y"), ("z", "y"), ("x", "z"), ("w", "y"),
("w", "x")])
self.assertEqual(G2.get_immoralities(), {("w", "z")})
def test_is_iequivalent(self):
G = BayesianModel([("x", "y"), ("z", "y"), ("x", "z"), ("w", "y")])
self.assertRaises(TypeError, G.is_iequivalent, MarkovModel())
G1 = BayesianModel([("V", "W"), ("W", "X"), ("X", "Y"), ("Z", "Y")])
G2 = BayesianModel([("W", "V"), ("X", "W"), ("X", "Y"), ("Z", "Y")])
self.assertTrue(G1.is_iequivalent(G2))
G3 = BayesianModel([("W", "V"), ("W", "X"), ("Y", "X"), ("Z", "Y")])
self.assertFalse(G3.is_iequivalent(G2))
def test_copy(self):
model_copy = self.G1.copy()
self.assertEqual(sorted(self.G1.nodes()), sorted(model_copy.nodes()))
self.assertEqual(sorted(self.G1.edges()), sorted(model_copy.edges()))
self.assertNotEqual(id(self.G1.get_cpds("diff")),
id(model_copy.get_cpds("diff")))
self.G1.remove_cpds("diff")
diff_cpd = TabularCPD("diff", 2, values=[[0.3], [0.7]])
self.G1.add_cpds(diff_cpd)
self.assertNotEqual(self.G1.get_cpds("diff"),
model_copy.get_cpds("diff"))
self.G1.remove_node("intel")
self.assertNotEqual(sorted(self.G1.nodes()),
sorted(model_copy.nodes()))
self.assertNotEqual(sorted(self.G1.edges()),
sorted(model_copy.edges()))
def test_remove_node(self):
self.G1.remove_node("diff")
self.assertEqual(sorted(self.G1.nodes()), sorted(["grade", "intel"]))
self.assertRaises(ValueError, self.G1.get_cpds, "diff")
def test_remove_nodes_from(self):
self.G1.remove_nodes_from(["diff", "grade"])
self.assertEqual(sorted(self.G1.nodes()), sorted(["intel"]))
self.assertRaises(ValueError, self.G1.get_cpds, "diff")
self.assertRaises(ValueError, self.G1.get_cpds, "grade")
def tearDown(self):
del self.G
del self.G1
class TestBayesianModelMethods(unittest.TestCase):
def setUp(self):
self.G = BayesianModel([('a', 'd'), ('b', 'd'),
('d', 'e'), ('b', 'c')])
self.G1 = BayesianModel([('diff', 'grade'), ('intel', 'grade')])
diff_cpd = TabularCPD('diff', 2, values=[[0.2], [0.8]])
intel_cpd = TabularCPD('intel', 3, values=[[0.5], [0.3], [0.2]])
grade_cpd = TabularCPD('grade', 3, values=[[0.1, 0.1, 0.1, 0.1, 0.1, 0.1],
[0.1, 0.1, 0.1, 0.1, 0.1, 0.1],
[0.8, 0.8, 0.8, 0.8, 0.8, 0.8]],
evidence=['diff', 'intel'], evidence_card=[2, 3])
self.G1.add_cpds(diff_cpd, intel_cpd, grade_cpd)
self.G2 = BayesianModel([('d', 'g'), ('g', 'l'), ('i', 'g'), ('i', 'l')])
def test_moral_graph(self):
moral_graph = self.G.moralize()
self.assertListEqual(sorted(moral_graph.nodes()), ['a', 'b', 'c', 'd', 'e'])
for edge in moral_graph.edges():
self.assertTrue(edge in [('a', 'b'), ('a', 'd'), ('b', 'c'), ('d', 'b'), ('e', 'd')] or
(edge[1], edge[0]) in [('a', 'b'), ('a', 'd'), ('b', 'c'), ('d', 'b'), ('e', 'd')])
def test_moral_graph_with_edge_present_over_parents(self):
G = BayesianModel([('a', 'd'), ('d', 'e'), ('b', 'd'), ('b', 'c'), ('a', 'b')])
moral_graph = G.moralize()
self.assertListEqual(sorted(moral_graph.nodes()), ['a', 'b', 'c', 'd', 'e'])
for edge in moral_graph.edges():
self.assertTrue(edge in [('a', 'b'), ('c', 'b'), ('d', 'a'), ('d', 'b'), ('d', 'e')] or
(edge[1], edge[0]) in [('a', 'b'), ('c', 'b'), ('d', 'a'), ('d', 'b'), ('d', 'e')])
def test_get_ancestors_of_success(self):
ancenstors1 = self.G2._get_ancestors_of('g')
ancenstors2 = self.G2._get_ancestors_of('d')
ancenstors3 = self.G2._get_ancestors_of(['i', 'l'])
self.assertEqual(ancenstors1, {'d', 'i', 'g'})
self.assertEqual(ancenstors2, {'d'})
self.assertEqual(ancenstors3, {'g', 'i', 'l', 'd'})
def test_get_ancestors_of_failure(self):
self.assertRaises(ValueError, self.G2._get_ancestors_of, 'h')
def test_local_independencies(self):
self.assertEqual(self.G.local_independencies('a'), Independencies(['a', ['b', 'c']]))
self.assertEqual(self.G.local_independencies('c'), Independencies(['c', ['a', 'd', 'e'], 'b']))
self.assertEqual(self.G.local_independencies('d'), Independencies(['d', 'c', ['b', 'a']]))
self.assertEqual(self.G.local_independencies('e'), Independencies(['e', ['c', 'b', 'a'], 'd']))
self.assertEqual(self.G.local_independencies('b'), Independencies(['b', 'a']))
self.assertEqual(self.G1.local_independencies('grade'), Independencies())
def test_get_independencies(self):
chain = BayesianModel([('X', 'Y'), ('Y', 'Z')])
self.assertEqual(chain.get_independencies(), Independencies(('X', 'Z', 'Y'), ('Z', 'X', 'Y')))
fork = BayesianModel([('Y', 'X'), ('Y', 'Z')])
self.assertEqual(fork.get_independencies(), Independencies(('X', 'Z', 'Y'), ('Z', 'X', 'Y')))
collider = BayesianModel([('X', 'Y'), ('Z', 'Y')])
self.assertEqual(collider.get_independencies(), Independencies(('X', 'Z'), ('Z', 'X')))
def test_is_imap(self):
val = [0.01, 0.01, 0.08, 0.006, 0.006, 0.048, 0.004, 0.004, 0.032,
0.04, 0.04, 0.32, 0.024, 0.024, 0.192, 0.016, 0.016, 0.128]
JPD = JointProbabilityDistribution(['diff', 'intel', 'grade'], [2, 3, 3], val)
fac = DiscreteFactor(['diff', 'intel', 'grade'], [2, 3, 3], val)
self.assertTrue(self.G1.is_imap(JPD))
self.assertRaises(TypeError, self.G1.is_imap, fac)
def test_get_immoralities(self):
G = BayesianModel([('x', 'y'), ('z', 'y'), ('x', 'z'), ('w', 'y')])
self.assertEqual(G.get_immoralities(), {('w', 'x'), ('w', 'z')})
G1 = BayesianModel([('x', 'y'), ('z', 'y'), ('z', 'x'), ('w', 'y')])
self.assertEqual(G1.get_immoralities(), {('w', 'x'), ('w', 'z')})
G2 = BayesianModel([('x', 'y'), ('z', 'y'), ('x', 'z'), ('w', 'y'), ('w', 'x')])
self.assertEqual(G2.get_immoralities(), {('w', 'z')})
def test_is_iequivalent(self):
G = BayesianModel([('x', 'y'), ('z', 'y'), ('x', 'z'), ('w', 'y')])
self.assertRaises(TypeError, G.is_iequivalent, MarkovModel())
G1 = BayesianModel([('V', 'W'), ('W', 'X'), ('X', 'Y'), ('Z', 'Y')])
G2 = BayesianModel([('W', 'V'), ('X', 'W'), ('X', 'Y'), ('Z', 'Y')])
self.assertTrue(G1.is_iequivalent(G2))
G3 = BayesianModel([('W', 'V'), ('W', 'X'), ('Y', 'X'), ('Z', 'Y')])
self.assertFalse(G3.is_iequivalent(G2))
def test_copy(self):
model_copy = self.G1.copy()
self.assertEqual(sorted(self.G1.nodes()), sorted(model_copy.nodes()))
self.assertEqual(sorted(self.G1.edges()), sorted(model_copy.edges()))
self.assertNotEqual(id(self.G1.get_cpds('diff')),
id(model_copy.get_cpds('diff')))
self.G1.remove_cpds('diff')
diff_cpd = TabularCPD('diff', 2, values=[[0.3], [0.7]])
self.G1.add_cpds(diff_cpd)
self.assertNotEqual(self.G1.get_cpds('diff'),
model_copy.get_cpds('diff'))
self.G1.remove_node('intel')
self.assertNotEqual(sorted(self.G1.nodes()), sorted(model_copy.nodes()))
self.assertNotEqual(sorted(self.G1.edges()), sorted(model_copy.edges()))
def test_remove_node(self):
self.G1.remove_node('diff')
self.assertEqual(sorted(self.G1.nodes()), sorted(['grade', 'intel']))
self.assertRaises(ValueError, self.G1.get_cpds, 'diff')
def test_remove_nodes_from(self):
self.G1.remove_nodes_from(['diff', 'grade'])
self.assertEqual(sorted(self.G1.nodes()), sorted(['intel']))
self.assertRaises(ValueError, self.G1.get_cpds, 'diff')
self.assertRaises(ValueError, self.G1.get_cpds, 'grade')
def tearDown(self):
del self.G
del self.G1
def bayesnet():
"""
References:
https://class.coursera.org/pgm-003/lecture/17
http://www.cs.ubc.ca/~murphyk/Bayes/bnintro.html
http://www3.cs.stonybrook.edu/~sael/teaching/cse537/Slides/chapter14d_BP.pdf
http://www.cse.unsw.edu.au/~cs9417ml/Bayes/Pages/PearlPropagation.html
https://github.com/pgmpy/pgmpy.git
http://pgmpy.readthedocs.org/en/latest/
http://nipy.bic.berkeley.edu:5000/download/11
"""
# import operator as op
# # Enumerate all possible events
# varcard_list = list(map(op.attrgetter('variable_card'), cpd_list))
# _esdat = list(ut.iprod(*map(range, varcard_list)))
# _escol = list(map(op.attrgetter('variable'), cpd_list))
# event_space = pd.DataFrame(_esdat, columns=_escol)
# # Custom compression of event space to inspect a specific graph
# def compress_space_flags(event_space, var1, var2, var3, cmp12_):
# """
# var1, var2, cmp_ = 'Lj', 'Lk', op.eq
# """
# import vtool as vt
# data = event_space
# other_cols = ut.setdiff_ordered(data.columns.tolist(), [var1, var2, var3])
# case_flags12 = cmp12_(data[var1], data[var2]).values
# # case_flags23 = cmp23_(data[var2], data[var3]).values
# # case_flags = np.logical_and(case_flags12, case_flags23)
# case_flags = case_flags12
# case_flags = case_flags.astype(np.int64)
# subspace = np.hstack((case_flags[:, None], data[other_cols].values))
# sel_ = vt.unique_row_indexes(subspace)
# flags = np.logical_and(mask, case_flags)
# return flags
# # Build special cases
# case_same = event_space.loc[compress_space_flags(event_space, 'Li', 'Lj', 'Lk', op.eq)]
# case_diff = event_space.loc[compress_space_flags(event_space, 'Li', 'Lj', 'Lk', op.ne)]
# special_cases = [
# case_same,
# case_diff,
# ]
from pgmpy.factors import TabularCPD
from pgmpy.models import BayesianModel
import pandas as pd
from pgmpy.inference import BeliefPropagation # NOQA
from pgmpy.inference import VariableElimination # NOQA
name_nice = ['n1', 'n2', 'n3']
score_nice = ['low', 'high']
match_nice = ['diff', 'same']
num_names = len(name_nice)
num_scores = len(score_nice)
nid_basis = list(range(num_names))
score_basis = list(range(num_scores))
semtype2_nice = {
'score': score_nice,
'name': name_nice,
'match': match_nice,
}
var2_cpd = {
}
globals()['semtype2_nice'] = semtype2_nice
globals()['var2_cpd'] = var2_cpd
name_combo = np.array(list(ut.iprod(nid_basis, nid_basis)))
combo_is_same = name_combo.T[0] == name_combo.T[1]
def get_expected_scores_prob(level1, level2):
part1 = combo_is_same * level1
part2 = (1 - combo_is_same) * (1 - (level2))
expected_scores_level = part1 + part2
return expected_scores_level
# def make_cpd():
def name_cpd(aid):
from pgmpy.factors import TabularCPD
cpd = TabularCPD(
variable='N' + aid,
variable_card=num_names,
values=[[1.0 / num_names] * num_names])
cpd.semtype = 'name'
return cpd
name_cpds = [name_cpd('i'), name_cpd('j'), name_cpd('k')]
var2_cpd.update(dict(zip([cpd.variable for cpd in name_cpds], name_cpds)))
if True:
num_same_diff = 2
samediff_measure = np.array([
# get_expected_scores_prob(.12, .2),
# get_expected_scores_prob(.88, .8),
get_expected_scores_prob(0, 0),
get_expected_scores_prob(1, 1),
])
samediff_vals = (samediff_measure / samediff_measure.sum(axis=0)).tolist()
def samediff_cpd(aid1, aid2):
cpd = TabularCPD(
variable='A' + aid1 + aid2,
variable_card=num_same_diff,
values=samediff_vals,
evidence=['N' + aid1, 'N' + aid2], # [::-1],
evidence_card=[num_names, num_names]) # [::-1])
cpd.semtype = 'match'
return cpd
samediff_cpds = [samediff_cpd('i', 'j'), samediff_cpd('j', 'k'), samediff_cpd('k', 'i')]
var2_cpd.update(dict(zip([cpd.variable for cpd in samediff_cpds], samediff_cpds)))
if True:
def score_cpd(aid1, aid2):
semtype = 'score'
evidence = ['A' + aid1 + aid2, 'N' + aid1, 'N' + aid2]
evidence_cpds = [var2_cpd[key] for key in evidence]
evidence_nice = [semtype2_nice[cpd.semtype] for cpd in evidence_cpds]
evidence_card = list(map(len, evidence_nice))
evidence_states = list(ut.iprod(*evidence_nice))
variable_basis = semtype2_nice[semtype]
variable_values = []
for mystate in variable_basis:
row = []
for state in evidence_states:
if state[0] == state[1]:
if state[2] == 'same':
val = .2 if mystate == 'low' else .8
else:
val = 1
# val = .5 if mystate == 'low' else .5
elif state[0] != state[1]:
if state[2] == 'same':
val = .5 if mystate == 'low' else .5
else:
val = 1
# val = .9 if mystate == 'low' else .1
row.append(val)
variable_values.append(row)
cpd = TabularCPD(
variable='S' + aid1 + aid2,
variable_card=len(variable_basis),
values=variable_values,
evidence=evidence, # [::-1],
evidence_card=evidence_card) # [::-1])
cpd.semtype = semtype
return cpd
else:
score_values = [
[.8, .1],
[.2, .9],
]
def score_cpd(aid1, aid2):
cpd = TabularCPD(
variable='S' + aid1 + aid2,
variable_card=num_scores,
values=score_values,
evidence=['A' + aid1 + aid2], # [::-1],
evidence_card=[num_same_diff]) # [::-1])
cpd.semtype = 'score'
return cpd
score_cpds = [score_cpd('i', 'j'), score_cpd('j', 'k')]
cpd_list = name_cpds + score_cpds + samediff_cpds
else:
score_measure = np.array([get_expected_scores_prob(level1, level2)
for level1, level2 in
zip(np.linspace(.1, .9, num_scores),
np.linspace(.2, .8, num_scores))])
score_values = (score_measure / score_measure.sum(axis=0)).tolist()
def score_cpd(aid1, aid2):
cpd = TabularCPD(
variable='S' + aid1 + aid2,
variable_card=num_scores,
values=score_values,
evidence=['N' + aid1, 'N' + aid2],
evidence_card=[num_names, num_names])
cpd.semtype = 'score'
return cpd
score_cpds = [score_cpd('i', 'j'), score_cpd('j', 'k')]
cpd_list = name_cpds + score_cpds
pass
input_graph = []
for cpd in cpd_list:
if cpd.evidence is not None:
for evar in cpd.evidence:
input_graph.append((evar, cpd.variable))
name_model = BayesianModel(input_graph)
name_model.add_cpds(*cpd_list)
var2_cpd.update(dict(zip([cpd.variable for cpd in cpd_list], cpd_list)))
globals()['var2_cpd'] = var2_cpd
varnames = [cpd.variable for cpd in cpd_list]
# --- PRINT CPDS ---
cpd = score_cpds[0]
def print_cpd(cpd):
print('CPT: %r' % (cpd,))
index = semtype2_nice[cpd.semtype]
if cpd.evidence is None:
columns = ['None']
else:
basis_lists = [semtype2_nice[var2_cpd[ename].semtype] for ename in cpd.evidence]
columns = [','.join(x) for x in ut.iprod(*basis_lists)]
data = cpd.get_cpd()
print(pd.DataFrame(data, index=index, columns=columns))
for cpd in name_model.get_cpds():
print('----')
print(cpd._str('phi'))
print_cpd(cpd)
# --- INFERENCE ---
Ni = name_cpds[0]
event_space_combos = {}
event_space_combos[Ni.variable] = 0 # Set ni to always be Fred
for cpd in cpd_list:
if cpd.semtype == 'score':
event_space_combos[cpd.variable] = list(range(cpd.variable_card))
evidence_dict = ut.all_dict_combinations(event_space_combos)
# Query about name of annotation k given different event space params
def pretty_evidence(evidence):
return [key + '=' + str(semtype2_nice[var2_cpd[key].semtype][val])
for key, val in evidence.items()]
def print_factor(factor):
row_cards = factor.cardinality
row_vars = factor.variables
values = factor.values.reshape(np.prod(row_cards), 1).flatten()
# col_cards = 1
# col_vars = ['']
basis_lists = list(zip(*list(ut.iprod(*[range(c) for c in row_cards]))))
nice_basis_lists = []
for varname, basis in zip(row_vars, basis_lists):
cpd = var2_cpd[varname]
_nice_basis = ut.take(semtype2_nice[cpd.semtype], basis)
nice_basis = ['%s=%s' % (varname, val) for val in _nice_basis]
nice_basis_lists.append(nice_basis)
row_lbls = [', '.join(sorted(x)) for x in zip(*nice_basis_lists)]
print(ut.repr3(dict(zip(row_lbls, values)), precision=3, align=True, key_order_metric='-val'))
# name_belief = BeliefPropagation(name_model)
name_belief = VariableElimination(name_model)
import pgmpy
import six # NOQA
def try_query(evidence):
print('--------')
query_vars = ut.setdiff_ordered(varnames, list(evidence.keys()))
evidence_str = ', '.join(pretty_evidence(evidence))
probs = name_belief.query(query_vars, evidence)
factor_list = probs.values()
joint_factor = pgmpy.factors.factor_product(*factor_list)
print('P(' + ', '.join(query_vars) + ' | ' + evidence_str + ')')
# print(six.text_type(joint_factor))
factor = joint_factor # NOQA
# print_factor(factor)
# import utool as ut
print(ut.hz_str([(f._str(phi_or_p='phi')) for f in factor_list]))
for evidence in evidence_dict:
try_query(evidence)
evidence = {'Aij': 1, 'Ajk': 1, 'Aki': 1, 'Ni': 0}
try_query(evidence)
evidence = {'Aij': 0, 'Ajk': 0, 'Aki': 0, 'Ni': 0}
try_query(evidence)
globals()['score_nice'] = score_nice
globals()['name_nice'] = name_nice
globals()['score_basis'] = score_basis
globals()['nid_basis'] = nid_basis
print('Independencies')
print(name_model.get_independencies())
print(name_model.local_independencies([Ni.variable]))
# name_belief = BeliefPropagation(name_model)
# # name_belief = VariableElimination(name_model)
# for case in special_cases:
# test_data = case.drop('Lk', axis=1)
# test_data = test_data.reset_index(drop=True)
# print('----')
# for i in range(test_data.shape[0]):
# evidence = test_data.loc[i].to_dict()
# probs = name_belief.query(['Lk'], evidence)
# factor = probs['Lk']
# probs = factor.values
# evidence_ = evidence.copy()
# evidence_['Li'] = name_nice[evidence['Li']]
# evidence_['Lj'] = name_nice[evidence['Lj']]
# evidence_['Sij'] = score_nice[evidence['Sij']]
# evidence_['Sjk'] = score_nice[evidence['Sjk']]
# nice2_prob = ut.odict(zip(name_nice, probs.tolist()))
# ut.print_python_code('P(Lk | {evidence}) = {cpt}'.format(
# evidence=(ut.repr2(evidence_, explicit=True, nobraces=True, strvals=True)),
# cpt=ut.repr3(nice2_prob, precision=3, align=True, key_order_metric='-val')
# ))
# for case in special_cases:
# test_data = case.drop('Lk', axis=1)
# test_data = test_data.drop('Lj', axis=1)
# test_data = test_data.reset_index(drop=True)
# print('----')
# for i in range(test_data.shape[0]):
# evidence = test_data.loc[i].to_dict()
# query_vars = ['Lk', 'Lj']
# probs = name_belief.query(query_vars, evidence)
# for queryvar in query_vars:
# factor = probs[queryvar]
# print(factor._str('phi'))
# probs = factor.values
# evidence_ = evidence.copy()
# evidence_['Li'] = name_nice[evidence['Li']]
# evidence_['Sij'] = score_nice[evidence['Sij']]
# evidence_['Sjk'] = score_nice[evidence['Sjk']]
# nice2_prob = ut.odict(zip([queryvar + '=' + x for x in name_nice], probs.tolist()))
# ut.print_python_code('P({queryvar} | {evidence}) = {cpt}'.format(
# query_var=query_var,
# evidence=(ut.repr2(evidence_, explicit=True, nobraces=True, strvals=True)),
# cpt=ut.repr3(nice2_prob, precision=3, align=True, key_order_metric='-val')
# ))
# _ draw model
import plottool as pt
import networkx as netx
fig = pt.figure() # NOQA
fig.clf()
ax = pt.gca()
netx_nodes = [(node, {}) for node in name_model.nodes()]
netx_edges = [(etup[0], etup[1], {}) for etup in name_model.edges()]
netx_graph = netx.DiGraph()
netx_graph.add_nodes_from(netx_nodes)
netx_graph.add_edges_from(netx_edges)
# pos = netx.graphviz_layout(netx_graph)
pos = netx.pydot_layout(netx_graph, prog='dot')
netx.draw(netx_graph, pos=pos, ax=ax, with_labels=True)
pt.plt.savefig('foo.png')
ut.startfile('foo.png')
print("test data:", len(data_test))
#################################################################################
##### Defining the model
#################################################################################
model = BayesianModel([('age_bin', 'class'), ('sex', 'class')])
#model = NaiveBayes([('class', 'age_bin'), ('class', 'sex')])
#model = BayesianModel([('sex', 'class')])
#model = BayesianModel([('class', 'sex')])
# Learing CPDs using Maximum Likelihood Estimators
model.fit(data_train, estimator=MaximumLikelihoodEstimator)
### independencies of network
print("independencies")
print(model.local_independencies('sex'))
#print(model.get_independencies())
#print(model.get_cpds('class'))
#################################################################################
##### using the model
#################################################################################
# Doing exact inference using Variable Elimination
model_infer = VariableElimination(model)
# Computing the probability of class given sex
q1 = model_infer.query(variables=['class'], evidence={'sex': 0})
print(q1['class'])
q2 = model_infer.query(variables=['class'], evidence={'sex': 1})
print(q2['class'])
