def initialize_relation(self):
variable_upper_to_ = {
var: None
for var in self.renormalized_model.variables
}
variables_lower_to_ = {
var: []
for var in self.renormalized_model.variables
}
factor_upper_to_ = {
var: Factor.scalar(1.0)
for var in self.renormalized_model.variables
}
upper_candidate_for_ = {
var: set()
for var in self.renormalized_model.variables
}
for fac in self.renormalized_model.factors:
lower_var = min(fac.variables,
key=self.renormalized_elimination_order.index)
factor_upper_to_[lower_var] = fac
upper_candidate_for_[lower_var] = upper_candidate_for_[
lower_var].union(set(fac.variables))
for var in self.renormalized_elimination_order:
m_vars = sorted(upper_candidate_for_[var],
key=self.renormalized_elimination_order.index)
upper_candidate_for_[var] = copy(m_vars[m_vars.index(var) + 1:])
if m_vars.index(var) + 1 < len(m_vars):
upper_var = m_vars[m_vars.index(var) + 1]
variable_upper_to_[var] = upper_var
variables_lower_to_[upper_var].append(var)
upper_candidate_for_[upper_var] = upper_candidate_for_[
upper_var].union(m_vars[m_vars.index(var) + 1:])
self.variable_upper_to_ = variable_upper_to_
self.variables_lower_to_ = variables_lower_to_
self.factors_adj_to_ = {
var: self.renormalized_model.get_adj_factors(var)
for var in self.renormalized_model.variables
}
self.factor_upper_to_ = factor_upper_to_
def run(self, elimination_order_method="random", **kwargs):
if elimination_order_method == "random":
elimination_order = copy(self.model.variables)
random.shuffle(elimination_order)
elif elimination_order_method == "not_random":
elimination_order = copy(self.model.variables)
elif elimination_order_method == "given":
elimination_order = kwargs["elimination_order"]
eliminated_model = self.model.copy()
max_ibound = 0
for var in elimination_order:
max_ibound = max(
max_ibound,
get_bucket_size(
[fac for fac in eliminated_model.get_adj_factors(var)]))
eliminated_model.contract_variable(var)
Z = Factor.scalar(1.0)
for fac in eliminated_model.factors:
Z = Z * fac
return Z.log_values
