def get_output_message(self, variable):
""" Returns output messages from this factor to the given variable node.
"""
om = defaultdict(list)
# FIXME: make this faster by using a dictionary
variable_index = [v.name for v in self.variables].index(variable.name)
selected_variables = [(i, v) for i, v in enumerate(self.variables)
if v.name != variable.name]
list_of_lists_of_values = [v.get_values() for v in self.variables]
for x in itertools.product(*list_of_lists_of_values):
log_prob = self.prob_table(*x)
for i, v in selected_variables:
log_prob += self.input_messages[v.name][v.value_2_index[x[i]]]
om[x[variable_index]].append(log_prob)
omlp = np.zeros_like(variable.log_probs)
for v in om:
omlp[variable.value_2_index[v]] = la.sum(om[v])
return omlp
def explain(self, full=False, linear_prob=False):
"""This function prints the values and their probabilities for this node.
"""
if full:
for value in sorted(self.value_2_index.keys()):
p = self.log_probs[self.value_2_index[value]]
if linear_prob:
p = np.exp(p)
print(
"Name: %-20s desc:%-10s Value: %-15s Probability: % .17f" %
(self.name, self.desc, value, p))
else:
for v in self.get_two_most_probable_values():
if linear_prob:
p = np.exp(v[1])
print(
"Name: %-20s desc:%-10s Value: %-15s Probability: % .17f" %
(self.name, self.desc, v[0], v[1]))
p = la.sum(self.log_probs)
if linear_prob:
p = np.exp(p)
print(('Cardinality: %.4d' + ' ' * 54 + ' Total: % .17f') %
(self.cardinality, p))
print
def get_output_message(self, variable):
""" Returns output messages from this factor to the given variable node.
"""
om = defaultdict(list)
# FIXME: make this faster by using a dictionary
variable_index = [v.name for v in self.variables].index(variable.name)
selected_variables = [(i, v) for i, v in enumerate(
self.variables) if v.name != variable.name]
list_of_lists_of_values = [v.get_values() for v in self.variables]
for x in itertools.product(*list_of_lists_of_values):
log_prob = self.prob_table(*x)
for i, v in selected_variables:
log_prob += self.input_messages[v.name][v.value_2_index[x[i]]]
om[x[variable_index]].append(log_prob)
omlp = np.zeros_like(variable.log_probs)
for v in om:
omlp[variable.value_2_index[v]] = la.sum(om[v])
return omlp
def explain(self, full=False, linear_prob=False):
"""This function prints the values and their probabilities for this node.
"""
if full:
for value in sorted(self.value_2_index.keys()):
p = self.log_probs[self.value_2_index[value]]
if linear_prob:
p = np.exp(p)
print("Name: %-20s desc:%-10s Value: %-15s Probability: % .17f"
% (self.name, self.desc, value, p))
else:
for v in self.get_two_most_probable_values():
if linear_prob:
p = np.exp(v[1])
print("Name: %-20s desc:%-10s Value: %-15s Probability: % .17f"
% (self.name, self.desc, v[0], v[1]))
p = la.sum(self.log_probs)
if linear_prob:
p = np.exp(p)
print(('Cardinality: %.4d' + ' ' * 54 + ' Total: % .17f')
% (self.cardinality, p))
print
def normalise(self):
"""This function normalise the sum of all probabilities to 1.0"""
self.log_probs -= la.sum(self.log_probs)
def normalise(self):
"""This function normalise the sum of all probabilities to 1.0"""
self.log_probs -= la.sum(self.log_probs)
