-
Notifications
You must be signed in to change notification settings - Fork 0
/
dag_generate.py
200 lines (153 loc) · 6.03 KB
/
dag_generate.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
import sys
import random
import logging
import numpy as np
import networkx as nx
from pgmpy.models.BayesianModel import BayesianModel
from pgmpy.factors.discrete import TabularCPD
from pgmpy.sampling import BayesianModelSampling
from asciinet import graph_to_ascii
import pcalg
from gsq import ci_tests
logging.basicConfig(level=logging.DEBUG, format="%(asctime)s %(levelname)s %(message)s")
# def graph_complete(nodes, undirected=True):
# if (undirected):
# g = nx.Graph()
# else:
# g = nx.DiGraph()
# for node in nodes:
# g.add_node(node)
# for s in range(0, len(nodes)):
# for t in range(s+1, len(nodes)):
# g.add_edge(nodes[s], nodes[t])
# if not (undirected):
# g.add_edge(nodes[t], nodes[s])
# return g
def get_node_name(n):
return 'ABCDEFGHIJKLMNOPQRSTUVWXYZ'[n]
def create_random_skeleton(node_count, edge_count):
nodes = [ get_node_name(i) for i in range(node_count) ]
g = nx.DiGraph()
g.add_nodes_from(nodes)
edge_pool = [ (s, t) for s in nodes for t in nodes if (s != t) ]
random.shuffle(edge_pool)
while g.number_of_edges() < edge_count:
s, t = edge_pool.pop(0)
g.add_edge(s, t)
if not nx.algorithms.dag.is_directed_acyclic_graph(g):
logging.debug("rejected edge {} -> {}".format(s, t))
g.remove_edge(s, t)
else:
logging.debug("added edge {} -> {}".format(s, t))
return g
def create_random_cpds(g, card):
node_pool = g.nodes()
node_count = len(node_pool)
random.shuffle(node_pool)
cpds = {}
dones = set()
while len(dones) < node_count:
node = node_pool.pop(0)
parents = set(g.predecessors(node))
logging.debug("creating cpd for {} (parents: {})".format(node, parents))
if len(parents) == 0:
cpds[node] = [ [p] for p in random_distrib(card) ]
dones.add(node)
elif parents.issubset(dones):
# construct a cpd whose size depends on the number of parents and the cardinality
distribs = [random_distrib(card) for i in range(card ** len(parents))]
cpds[node] = zip(*distribs)
dones.add(node)
else:
node_pool.append(node)
continue
if node in dones:
logging.debug("cpd for {}: {})".format(node, cpds[node]))
return cpds
def random_distrib(sz):
vals = []
slack = 1.0
for i in range(sz - 1):
val = random.uniform(0.0, slack)
slack -= val
vals.append(val)
vals.append(slack)
random.shuffle(vals)
return vals
# >>> student = BayesianModel([('diff', 'grade'), ('intel', 'grade')])
# >>> cpd_d = TabularCPD('diff', 2, [[0.6], [0.4]])
# >>> cpd_i = TabularCPD('intel', 2, [[0.7], [0.3]])
# >>> cpd_g = TabularCPD('grade', 3, [[0.3, 0.05, 0.9, 0.5], [0.4, 0.25,
# ... 0.08, 0.3], [0.3, 0.7, 0.02, 0.2]],
# ... ['intel', 'diff'], [2, 2])
# >>> student.add_cpds(cpd_d, cpd_i, cpd_g)
# >>> inference = BayesianModelSampling(student)
# >>> inference.forward_sample(size=2, return_type='recarray')
# rec.array([(0, 0, 1), (1, 0, 2)], dtype=
# [('diff', '<i8'), ('intel', '<i8'), ('grade', '<i8')])
def create_random_dag(node_count, edge_count, card):
logging.debug("creating skeleton")
dag = create_random_skeleton(node_count, edge_count)
logging.debug("creating cpds")
cpds = create_random_cpds(dag, card)
for node, cpd in cpds.items():
dag.node[node]['cpd'] = cpd
return dag
def sample_dag(dag, num):
#zzz this loses disconnected nodes!!!
# bayesmod = BayesianModel(dag.edges())
# bayesmod = BayesianModel(dag)
bayesmod = BayesianModel()
bayesmod.add_nodes_from(dag.nodes())
bayesmod.add_edges_from(dag.edges())
tab_cpds = []
cards = { node: len(dag.node[node]['cpd']) for node in dag.nodes() }
for node in dag.nodes():
parents = dag.predecessors(node)
cpd = dag.node[node]['cpd']
if parents:
parent_cards = [ cards[par] for par in parents ]
logging.debug("TablularCPD({}, {}, {}, {}, {})".format(node, cards[node], cpd,
parents, parent_cards))
tab_cpds.append(TabularCPD(node, cards[node], cpd, parents, parent_cards))
else:
logging.debug("TablularCPD({}, {}, {})".format(node, cards[node], cpd))
tab_cpds.append(TabularCPD(node, cards[node], cpd))
logging.debug("cpds add: {}".format(tab_cpds))
print "model variables:", bayesmod.nodes()
for tab_cpd in tab_cpds:
print "cpd variables:", tab_cpd.variables
bayesmod.add_cpds(*tab_cpds)
logging.debug("cpds get: {}".format(bayesmod.get_cpds()))
inference = BayesianModelSampling(bayesmod)
logging.debug("generating data")
recs = inference.forward_sample(size=num, return_type='recarray')
return recs
def run_pc(data_orig, col_names=None):
data = np.array([ list(r) for r in data_orig ])
(skel_graph, sep_set) = pcalg.estimate_skeleton(indep_test_func=ci_tests.ci_test_dis,
data_matrix=data,
alpha=0.01)
# gdir = nx.DiGraph()
# gdir.add_nodes_from(g.nodes())
# gdir.add_edges_from(g.edges())
dag = pcalg.estimate_cpdag(skel_graph, sep_set)
if col_names:
name_map = { i: col_names[i] for i in range(len(dag.nodes())) }
nx.relabel.relabel_nodes(dag, name_map, copy=False)
return dag
#####################################
if __name__ == '__main__':
num_nodes = int(sys.argv[1])
num_edges = int(sys.argv[2])
attr_card = int(sys.argv[3])
dag = create_random_dag(num_nodes, num_edges, attr_card)
print graph_to_ascii(dag)
recs = sample_dag(dag, 10000)
print dag.nodes()
print recs[:10]
for node in dag.nodes():
print "col", node, recs[node][:10]
gdir = run_pc(recs)
print graph_to_ascii(gdir)
print "graphs are isomorphic: ", nx.algorithms.isomorphism.is_isomorphic(dag, gdir)