-
Notifications
You must be signed in to change notification settings - Fork 60
/
causal_model.py
286 lines (231 loc) · 7.31 KB
/
causal_model.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
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
from typing import List
import networkx as nx
from causalgraphicalmodels import CausalGraphicalModel, StructuralCausalModel
from carla.data.causal_model.synthethic_data import ScmDataset
from carla.data.load_scm import load_scm_equations
class CausalModel:
"""
Class with topological methods given a structural causal model. Uses the StructuralCausalModel and
CausalGraphicalModel from https://github.com/ijmbarr/causalgraphicalmodels
Parameters
----------
scm_class: str
Name of the structural causal model
Attributes
----------
scm: StructuralCausalModel
StructuralCausalModel from assignment of the form { variable: Function(parents) }.
cgm: CausalGraphicalModel
scm_class: str
Name of the structural causal model
structural_equations_np: dict
Contains the equations for the features in Numpy format.
structural_equations_ts: dict
Contains the equations for the features in Tensorflow format.
noise_distributions: dict
Defines the noise variables.
"""
def __init__(
self,
scm_class: str,
):
self._scm_class = scm_class
(
self._structural_equations_np,
self._structural_equations_ts,
self._noise_distributions,
self._continuous,
self._categorical,
self._immutables,
) = load_scm_equations(scm_class)
self._scm = StructuralCausalModel(self._structural_equations_np)
self._cgm = self._scm.cgm
self._endogenous = list(self._structural_equations_np.keys())
self._exogenous = list(self._noise_distributions.keys())
self._continuous_noise = list(set(self._continuous) - set(self.endogenous))
self._categorical_noise = list(set(self._categorical) - set(self.endogenous))
self._continuous = list(set(self._continuous) - set(self._exogenous))
self._categorical = list(set(self._categorical) - set(self._exogenous))
def get_topological_ordering(self, node_type="endogenous"):
"""Returns a generator of nodes in topologically sorted order.
A topological sort is a non-unique permutation of the nodes such that an
edge from u to v implies that u appears before v in the topological sort
order.
Parameters
----------
node_type: str
"endogenous" or "exogenous", i.e. nodes with "x" or "u" prefix respectively
Returns
-------
iterable
An iterable of node names in topological sorted order.
"""
tmp = nx.topological_sort(self._cgm.dag)
if node_type == "endogenous":
return tmp
elif node_type == "exogenous":
return ["u" + node[1:] for node in tmp]
else:
raise Exception(f"{node_type} not recognized.")
def get_children(self, node: str) -> set:
"""Returns an iterator over successor nodes of n.
A successor of n is a node m such that there exists a directed
edge from n to m.
Parameters
----------
node: str
A node in the graph
"""
return set(self._cgm.dag.successors(node))
def get_parents(self, node: str, return_sorted: bool = True):
"""Returns an set over predecessor nodes of n.
A predecessor of n is a node m such that there exists a directed
edge from m to n.
Parameters
----------
node : str
A node in the graph
return_sorted : bool
Return the set sorted
"""
tmp = set(self._cgm.dag.predecessors(node))
return sorted(tmp) if return_sorted else tmp
def get_ancestors(self, node: str) -> set:
"""Returns all nodes having a path to `node`.
Parameters
----------
node : str
A node in the graph
Returns
-------
set()
The ancestors of node
"""
return nx.ancestors(self._cgm.dag, node)
def get_descendents(self, node: str) -> set:
"""Returns all nodes reachable from `node`.
Parameters
----------
node : str
A node in the graph
Returns
-------
set()
The descendants of `node`
"""
return nx.descendants(self._cgm.dag, node)
def get_non_descendents(self, node: str) -> set:
"""Returns all nodes not reachable from `node`.
Parameters
----------
node : str
A node in the graph
Returns
-------
set()
The non-descendants of `node`
"""
return (
set(nx.topological_sort(self._cgm.dag))
.difference(self.get_descendents(node))
.symmetric_difference(set([node]))
)
def generate_dataset(self, size: int) -> ScmDataset:
"""Generates a Data object using the structural causal equations
Parameters
----------
size: int
Number of samples in the dataset
Returns
-------
ScmDataset
a Data object filled with samples
"""
return ScmDataset(self, size)
def visualize_graph(self, experiment_folder_name=None):
"""
Visualize the causal graph.
Parameters
----------
experiment_folder_name: str
Where to save figure.
Returns
-------
"""
if experiment_folder_name:
save_path = f"{experiment_folder_name}/_causal_graph"
view_flag = False
else:
save_path = "_tmp/_causal_graph"
view_flag = True
self._cgm.draw().render(save_path, view=view_flag)
@property
def scm(self) -> StructuralCausalModel:
"""
Returns
-------
StructuralCausalModel
"""
return self._scm
@property
def cgm(self) -> CausalGraphicalModel:
"""
Returns
-------
CausalGraphicalModel
"""
return self._cgm
@property
def scm_class(self) -> str:
"""
Name of the structural causal model used to define the CausalModel
Returns
-------
str
"""
return self._scm_class
@property
def structural_equations_np(self) -> dict:
"""
Contains the equations for the features in Numpy format.
Returns
-------
dict
"""
return self._structural_equations_np
@property
def structural_equations_ts(self) -> dict:
"""
Contains the equations for the features in Tensorflow format.
Returns
-------
dict
"""
return self._structural_equations_ts
@property
def noise_distributions(self) -> dict:
"""
Defines the noise variables.
Returns
-------
dict
"""
return self._noise_distributions
@property
def exogenous(self) -> List[str]:
"""
Get the exogenous nodes, i.e. the noise nodes.
Returns
-------
List[str]
"""
return self._exogenous
@property
def endogenous(self) -> List[str]:
"""
Get the endogenous nodes, i.e. the signal nodes.
Returns
-------
List[str]
"""
return self._endogenous