def __project_observations(self):
for evidence_var in self.query.evidence_variables:
appearing_factors = utils.appearing_factors(evidence_var.observed_variable, self.relevant_factors)
while(appearing_factors):
observation.project_observation(appearing_factors, evidence_var, self.relevant_factors, self.eliminated_factors, self.initial_factors, self.query.evidence_variables)
def __min_fill(variables, relevant_factors):
"""
Min. Fill: Devuelve el nodo que al eliminarlo introducira el menor numero de conexiones nuevas.
@type variables: set
@param variables: Conjunto de variables ocultas que quedan en la red bayesiana.
@type relevant_factors: list
@param relevant_factors: Lista de factores relevantes de la red bayesiana.
@rtype: Node
@return: Nodo a eliminar segun la heuristica.
"""
minimum = float("inf")
elimination_variable = None
for var in variables:
new_edges_count = 0
appearing_factors = utils.appearing_factors(var, relevant_factors)
for f1 in appearing_factors:
f1_position = appearing_factors.index(f1)
appearing_factors_size = len(appearing_factors)
for f2 in range(f1_position + 1, appearing_factors_size):
f2_factors = appearing_factors[f2].factors
intersection_size = len(set(f1.factors) & set(f2_factors))
new_edges_count += (len(f1.factors) - intersection_size) * (len(f2_factors) - intersection_size)
if new_edges_count < minimum:
minimum = new_edges_count
elimination_variable = var
return elimination_variable
def __min_factor(variables, relevant_factors):
"""
Min. Factor: Devuelve el nodo de menor peso, siendo el peso de cada nodo el producto del cardinal del dominio de sus vecinos.
@type variables: set
@param variables: Conjunto de variables ocultas que quedan en la red bayesiana.
@type relevant_factors: list
@param relevant_factors: Lista de factores relevantes de la red bayesiana.
@rtype: Node
@return: Nodo a eliminar segun la heuristica.
"""
minimum = float('inf')
elimination_variable = None
for var in variables:
neighbors = set(list(itertools.chain.from_iterable([f.factors for f in utils.appearing_factors(var, relevant_factors)])))
neighbors.remove(var)
weight = 1
for n in neighbors:
weight *= len(n.domain)
if weight < minimum:
minimum = weight
elimination_variable = var
return elimination_variable
def __min_degree(variables, relevant_factors):
"""
Min. Degree: Devuelve el nodo con el menor numero de conexiones.
@type variables: set
@param variables: Conjunto de variables ocultas que quedan en la red bayesiana.
@type relevant_factors: list
@param relevant_factors: Lista de factores relevantes de la red bayesiana.
@rtype: Node
@return: Nodo a eliminar segun la heuristica.
"""
new_factor_length = lambda var: len(set(itertools.chain.from_iterable([f.factors for f in utils.appearing_factors(var, relevant_factors)])))
return min(variables, key=new_factor_length)
