pyAgrum-nightly 2.3.0.9.dev202512061764412981__cp310-abi3-macosx_11_0_arm64.whl

pyagrum/causal/_doorCriteria.py

@@ -0,0 +1,374 @@
+############################################################################
+#   This file is part of the aGrUM/pyAgrum library.                        #
+#                                                                          #
+#   Copyright (c) 2005-2025 by                                             #
+#       - Pierre-Henri WUILLEMIN(_at_LIP6)                                 #
+#       - Christophe GONZALES(_at_AMU)                                     #
+#                                                                          #
+#   The aGrUM/pyAgrum library is free software; you can redistribute it    #
+#   and/or modify it under the terms of either :                           #
+#                                                                          #
+#    - the GNU Lesser General Public License as published by               #
+#      the Free Software Foundation, either version 3 of the License,      #
+#      or (at your option) any later version,                              #
+#    - the MIT license (MIT),                                              #
+#    - or both in dual license, as here.                                   #
+#                                                                          #
+#   (see https://agrum.gitlab.io/articles/dual-licenses-lgplv3mit.html)    #
+#                                                                          #
+#   This aGrUM/pyAgrum library is distributed in the hope that it will be  #
+#   useful, but WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED,          #
+#   INCLUDING BUT NOT LIMITED TO THE WARRANTIES MERCHANTABILITY or FITNESS #
+#   FOR A PARTICULAR PURPOSE  AND NONINFRINGEMENT. IN NO EVENT SHALL THE   #
+#   AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER #
+#   LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,        #
+#   ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR  #
+#   OTHER DEALINGS IN THE SOFTWARE.                                        #
+#                                                                          #
+#   See LICENCES for more details.                                         #
+#                                                                          #
+#   SPDX-FileCopyrightText: Copyright 2005-2025                            #
+#       - Pierre-Henri WUILLEMIN(_at_LIP6)                                 #
+#       - Christophe GONZALES(_at_AMU)                                     #
+#   SPDX-License-Identifier: LGPL-3.0-or-later OR MIT                      #
+#                                                                          #
+#   Contact  : info_at_agrum_dot_org                                       #
+#   homepage : http://agrum.gitlab.io                                      #
+#   gitlab   : https://gitlab.com/agrumery/agrum                           #
+#                                                                          #
+############################################################################
+
+"""
+This files define some routines for front and back door
+"""
+
+from typing import Optional
+import itertools as it
+
+import pyagrum
+
+from pyagrum.causal._types import NameSet, NodeId, NodeSet
+from pyagrum.causal._dSeparation import isParent, dSep_reduce, descendants, isDSep_parents
+
+# pylint: disable=unused-import
+import pyagrum.causal  # for annotations
+
+
+def backdoor_path(bn: "pyagrum.BayesNet", x: str, y: str, zset: NameSet = None) -> bool:
+  """
+  Predicate on the existence of an open back door path from ``x`` to ``y``,
+  conditioning on the set of variables ``zset``
+
+  Parameters
+  ----------
+  bn: pyagrum.BayesNez
+      the DAG model
+  x: str
+      name of source node
+  y: str
+      name of destination node
+  zset: Set[str]
+      names of conditioning nodes
+
+  Returns
+  -------
+  bool
+      True if such backdoor exists
+  """
+  if zset is None:
+    zset = set()
+  return not isDSep_parents(bn, x, y, zset)
+
+
+def exists_unblocked_directed_path(bn: "pyagrum.BayesNet", x: str, y: str, zset: NameSet = None) -> bool:
+  """
+  Predicate on the existence of a directed path from ``x`` to ``y`` in the Bayesian network ``bn``
+  not blocked by nodes of ``zset``
+
+  Parameters
+  ----------
+  bn: pyagrum.BayesNez
+      the DAG model
+  x: str
+      name of source node
+  y: str
+      name of destination node
+  zset: Set[str]
+      names of conditioning nodes
+
+  Returns
+  -------
+  bool
+  """
+  if zset is None:
+    zset = set()
+  if y in bn.children(x):
+    return True
+
+  for c in bn.children(x):
+    if (c not in zset) and exists_unblocked_directed_path(bn, c, y, zset):
+      return True
+  return False
+
+
+def is_frontdoor(bn: "pyagrum.BayesNet", x: str, y: str, zset: NameSet) -> bool:
+  """
+  Tests whether or not ``zset`` satisifies the front door criterion for ``x`` and ``y``, in the Bayesian network ``bn``
+
+  Parameters
+  ----------
+  bn: pyagrum.BayesNez
+      the DAG model
+  x: str
+      name of source node
+  y: str
+      name of destination node
+  zset: Set[str]
+      names of conditioning nodes
+
+  Returns
+  -------
+  bool
+  """
+  if exists_unblocked_directed_path(bn, x, y, zset):
+    return False
+
+  if len(zset & backdoor_reach(bn, x)) != 0:
+    return False
+
+  bn_reduit = dSep_reduce(bn, zset | {x, y})
+  for i in zset:
+    if backdoor_path(bn_reduit, i, y, {x}):
+      return False
+
+  return True
+
+
+def is_backdoor(bn: "pyagrum.BayesNet", x: str, y: str, zset: NameSet) -> bool:
+  """
+  Tests whether or not ``zset`` satisifies the back door criterion for ``x`` and ``y``, in the Bayesian network ``bn``
+
+  Parameters
+  ----------
+  bn: pyagrum.BayesNez
+      the DAG model
+  x: str
+      name of source node
+  y: str
+      name of destination node
+  zset: Set[str]
+      names of conditioning nodes
+
+  Returns
+  -------
+  bool
+  """
+
+  desc_x = descendants(bn, x)
+  if len(desc_x & zset) != 0:
+    return False
+
+  return isDSep_parents(bn, x, y, zset)
+
+
+def backdoor_reach(bn: "pyagrum.BayesNet", a: NodeId) -> NodeSet:
+  """
+  Returns the set of nodes that can be reached through a backdoor path from ``a`` in the graph ``bn``
+
+  Parameters
+  ----------
+  bn: pyagrum.BayesNet
+    the model
+  a: int
+    the backdoor-ed node
+
+  Returns
+  -------
+  Set[int]
+    set of nodes
+  """
+
+  def inner_br(bn: "pyagrum.BayesNet", x: NodeId, pht: bool, reach0: NodeSet, reach1: NodeSet):
+    for c in bn.children(x):
+      if c not in reach0 and c not in reach1:
+        reach1.add(c)
+        inner_br(bn, c, True, reach0, reach1)
+
+    if not pht:
+      for p in bn.parents(x):
+        if p not in reach0:
+          reach0.add(p)
+          inner_br(bn, p, False, reach0, reach1)
+
+  r = {a}
+  r.union(bn.parents(a))
+  l = {a}
+  for pa in bn.parents(a):
+    inner_br(bn, pa, False, r, l)
+  s = r | l
+  if a in s:
+    s.remove(a)
+  return s
+
+
+def nodes_on_dipath(bn: "pyagrum.BayesNet", x: NodeId, y: NodeId) -> Optional[NodeSet]:
+  """
+  Returns the set of nodes through which there is a directed path from `x` to `y` in the graph `bn`
+
+  Parameters
+  ----------
+  bn: pyagrum.BayesNet
+  x  int
+  y: int
+
+  Returns
+  -------
+  Set[int] (maybe None)
+  """
+
+  def inner_nod(g: "pyagrum.BayesNet", a: NodeId, b: NodeId) -> Optional[NodeSet]:
+    if b == a:
+      return set()
+
+    inners = {a}
+    children = g.children(a)
+    if len(children) == 0:
+      return None
+
+    found = False
+    for c in children:
+      s = inner_nod(g, c, b)
+      if s is not None:
+        found = True
+        inners |= s
+    if found:
+      return inners
+    return None
+
+  r = inner_nod(bn, x, y)
+  if r:
+    r.remove(x)
+
+  return r
+
+
+def backdoor_generator(bn: "pyagrum.BayesNet", cause: NodeId, effect: NodeId, not_bd: NodeSet = None):
+  """
+  Generates backdoor sets for the pair of nodes `(cause, effect)` in the graph `bn` excluding the nodes in the set
+  `not_bd` (optional)
+
+  Parameters
+  ----------
+  bn: pyagrum.BayesNet
+  cause: int
+  effect: int
+  not_bd: Set[int] default=None
+
+  Yields
+  -------
+  List[int]
+    the different backdoors
+  """
+  if len(bn.parents(cause)) == 0:  # no parent of cause, no backdoor
+    return
+  if isParent(effect, cause, bn):  # causalDagFromBN(bn)):
+    return
+
+  if not_bd is None:
+    not_bd = set()
+
+  # simplifying the graph
+  interest = {cause, effect}
+  G = dSep_reduce(bn, interest)
+
+  # removing the non connected in G without descendants
+  # GG is a trash graph just to find the disjointed nodes in G
+  GG = pyagrum.DiGraph(G)
+  for i in descendants(bn, cause, set()):
+    GG.eraseNode(i)
+
+  # we only keep interesting connex components
+  for nodes in GG.connectedComponents().values():
+    if nodes.isdisjoint(interest):
+      for n in nodes:
+        G.eraseNode(n)
+
+  possible = set(G.nodes()) - (descendants(bn, cause, set()) | interest | not_bd)
+  if len(possible) == 0:
+    return
+
+  backdoors = set()
+
+  for i in range(len(possible)):
+    for subset in it.combinations(possible, i + 1):
+      sub = frozenset(subset)
+      worth_testing = True
+      for s in backdoors:
+        if s <= sub:
+          worth_testing = False
+      if worth_testing and isDSep_parents(G, {cause}, {effect}, sub):
+        backdoors.add(sub)
+        yield list(subset)
+
+
+def frontdoor_generator(bn: "pyagrum.BayesNet", x: NodeId, y: NodeId, not_fd: NodeSet = None):
+  """
+  Generates frontdoor sets for the pair of nodes `(x, y)` in the graph `bn` excluding the nodes in the set
+  `not_fd` (optional)
+
+  Parameters
+  ----------
+  bn: pyagrum.BayesNet
+  x: int
+  y: int
+  not_fd: Set[int] default=None
+
+  Yields
+  -------
+  List[int]
+    the different frontdoors
+  """
+  if isParent(x, y, bn):
+    return
+
+  if not_fd is None:
+    not_fd = set()
+
+  possible = nodes_on_dipath(bn, x, y)
+
+  nodiPath = False
+  if possible is None:
+    nodiPath = True
+    possible = None
+    for _, cc in bn.connectedComponents().items():
+      if x in cc and y in cc:
+        possible = cc
+    if possible is None:
+      return
+    possible -= {x, y}
+
+  possible -= backdoor_reach(bn, x)
+  possible -= not_fd
+  impossible = set()
+  g = dSep_reduce(bn, {x, y} | possible)
+  for z in possible:
+    if backdoor_path(g, {z}, {y}, {x}):
+      impossible.add(z)
+  possible -= impossible
+  frontdoors = set()
+
+  if nodiPath:
+    for s in possible:
+      yield [s]
+    return
+
+  for i in range(len(possible)):
+    for subset in it.combinations(possible, i + 1):
+      sub = frozenset(subset)
+      worth_testing = True
+      for s in frontdoors:
+        if s <= sub:
+          worth_testing = False
+      if worth_testing and not exists_unblocked_directed_path(bn, x, y, sub):
+        frontdoors.add(sub)
+        yield list(subset)

pyagrum/causal/_exceptions.py

@@ -0,0 +1,95 @@
+############################################################################
+#   This file is part of the aGrUM/pyAgrum library.                        #
+#                                                                          #
+#   Copyright (c) 2005-2025 by                                             #
+#       - Pierre-Henri WUILLEMIN(_at_LIP6)                                 #
+#       - Christophe GONZALES(_at_AMU)                                     #
+#                                                                          #
+#   The aGrUM/pyAgrum library is free software; you can redistribute it    #
+#   and/or modify it under the terms of either :                           #
+#                                                                          #
+#    - the GNU Lesser General Public License as published by               #
+#      the Free Software Foundation, either version 3 of the License,      #
+#      or (at your option) any later version,                              #
+#    - the MIT license (MIT),                                              #
+#    - or both in dual license, as here.                                   #
+#                                                                          #
+#   (see https://agrum.gitlab.io/articles/dual-licenses-lgplv3mit.html)    #
+#                                                                          #
+#   This aGrUM/pyAgrum library is distributed in the hope that it will be  #
+#   useful, but WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED,          #
+#   INCLUDING BUT NOT LIMITED TO THE WARRANTIES MERCHANTABILITY or FITNESS #
+#   FOR A PARTICULAR PURPOSE  AND NONINFRINGEMENT. IN NO EVENT SHALL THE   #
+#   AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER #
+#   LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,        #
+#   ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR  #
+#   OTHER DEALINGS IN THE SOFTWARE.                                        #
+#                                                                          #
+#   See LICENCES for more details.                                         #
+#                                                                          #
+#   SPDX-FileCopyrightText: Copyright 2005-2025                            #
+#       - Pierre-Henri WUILLEMIN(_at_LIP6)                                 #
+#       - Christophe GONZALES(_at_AMU)                                     #
+#   SPDX-License-Identifier: LGPL-3.0-or-later OR MIT                      #
+#                                                                          #
+#   Contact  : info_at_agrum_dot_org                                       #
+#   homepage : http://agrum.gitlab.io                                      #
+#   gitlab   : https://gitlab.com/agrumery/agrum                           #
+#                                                                          #
+############################################################################
+
+"""
+This file defines the specific exception for causal package
+"""
+
+from ._types import NameSet
+
+
+class HedgeException(Exception):
+  """
+  Represents an hedge exception for a causal query
+
+  Parameters
+  ----------
+  msg: str
+  observables: NameSet
+  gs: ???
+  """
+
+  def __init__(self, msg: str, observables: NameSet, gs):
+    """
+    Represents an hedge exception for a causal query
+
+    Parameters
+    ----------
+    msg: str
+    observables: NameSet
+    gs: ???
+    """
+    self.message = msg
+    self.type = "HedgeException"
+    self.observables = observables
+    self.gs = gs
+    super().__init__(self.message)
+
+
+class UnidentifiableException(Exception):
+  """
+  Represents an unidentifiability for a causal query
+
+  Parameters
+  ----------
+  msg: str
+  """
+
+  def __init__(self, msg):
+    """
+    Represents an unidentifiability for a causal query
+
+    Parameters
+    ----------
+    msg: str
+    """
+    self.message = msg
+    self.type = "Unidentifiable"
+    super().__init__(self.message)

pyagrum/causal/_types.py

@@ -0,0 +1,61 @@
+############################################################################
+#   This file is part of the aGrUM/pyAgrum library.                        #
+#                                                                          #
+#   Copyright (c) 2005-2025 by                                             #
+#       - Pierre-Henri WUILLEMIN(_at_LIP6)                                 #
+#       - Christophe GONZALES(_at_AMU)                                     #
+#                                                                          #
+#   The aGrUM/pyAgrum library is free software; you can redistribute it    #
+#   and/or modify it under the terms of either :                           #
+#                                                                          #
+#    - the GNU Lesser General Public License as published by               #
+#      the Free Software Foundation, either version 3 of the License,      #
+#      or (at your option) any later version,                              #
+#    - the MIT license (MIT),                                              #
+#    - or both in dual license, as here.                                   #
+#                                                                          #
+#   (see https://agrum.gitlab.io/articles/dual-licenses-lgplv3mit.html)    #
+#                                                                          #
+#   This aGrUM/pyAgrum library is distributed in the hope that it will be  #
+#   useful, but WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED,          #
+#   INCLUDING BUT NOT LIMITED TO THE WARRANTIES MERCHANTABILITY or FITNESS #
+#   FOR A PARTICULAR PURPOSE  AND NONINFRINGEMENT. IN NO EVENT SHALL THE   #
+#   AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER #
+#   LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,        #
+#   ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR  #
+#   OTHER DEALINGS IN THE SOFTWARE.                                        #
+#                                                                          #
+#   See LICENCES for more details.                                         #
+#                                                                          #
+#   SPDX-FileCopyrightText: Copyright 2005-2025                            #
+#       - Pierre-Henri WUILLEMIN(_at_LIP6)                                 #
+#       - Christophe GONZALES(_at_AMU)                                     #
+#   SPDX-License-Identifier: LGPL-3.0-or-later OR MIT                      #
+#                                                                          #
+#   Contact  : info_at_agrum_dot_org                                       #
+#   homepage : http://agrum.gitlab.io                                      #
+#   gitlab   : https://gitlab.com/agrumery/agrum                           #
+#                                                                          #
+############################################################################
+
+"""
+This file defines the types for the causal package
+"""
+
+from typing import NewType
+from typing import Set, List, Tuple, Union
+
+# pylint: disable=unused-import
+import pyagrum.causal  # for annotations
+
+DirectedModel = Union["pyagrum.BayesNet", "pyagrum.DAG", "pyagrum.causal.CausalModel"]
+
+NodeId = NewType("NodeId", int)
+NodeList = List[NodeId]
+NodeSet = Set[NodeId]
+
+ArcSet = Set[Tuple[NodeId, NodeId]]
+NameSet = Set[str]
+
+LatentDescriptor = Tuple[str, Tuple[str, str]]
+LatentDescriptorList = List[LatentDescriptor]