def get_probabilities_from_accessor(node, states: Dict[str, str],
table_accessor) -> List[float]:
variable_order = get_node_order(node)
JavaIntArray = jp.JArray(jp.JInt)
indexes = JavaIntArray([[
state.getName()
for state in bayesianpy.network.get_variable_from_node(n).getStates()
].index(states[n.getName()]) for n in variable_order])
return table_accessor.get(indexes)
def set_soft(self, variable_name, states: Dict[str, float]):
v = self._variables.get(variable_name)
v_states = v.getStates()
evidence = np.zeros(len(v_states))
for i, state in enumerate(v_states):
if state.getName() in states:
evidence[i] = states[state.getName()]
javaDblArray = jp.JArray(jp.JDouble)
j_evidence = javaDblArray(evidence.astype(float).tolist())
self._evidence.setStates(v, j_evidence)
return self._evidence
def get_state_indexes_from_iterator(node, table_iterator) -> List[int]:
JavaIntArray = jp.JArray(jp.JInt)
state_order = JavaIntArray([0] * len(get_node_order(node)))
table_iterator.getStates(state_order)
return list(state_order)
def results(self, inference_engine, query_output):
results = {}
def state_generator(variables):
for v in variables:
tv = bayesianpy.network.get_variable(self._network, v)
yield [state for state in tv.getStates()]
if self._is_discrete_head:
rows = []
query_variables = self._head_variables + self._tail_variables
# creates dataframe of combinations + probability.
for state_combinations in itertools.product(
*state_generator(query_variables)):
row = {}
state_array = jp.JArray(state_combinations[0].getClass())(
len(state_combinations))
for i, state in enumerate(state_combinations):
state_array[i] = state
for i, v in enumerate(state_combinations):
row.update({query_variables[i]: state_array[i].getName()})
row.update(
{'probability': self._distribution.get(state_array)})
rows.append(row)
return rows
else:
if len(self._head_variables) == 2 and len(
self._tail_variables) == 0:
h0 = bayesianpy.network.get_variable(self._network,
self._head_variables[0])
h1 = bayesianpy.network.get_variable(self._network,
self._head_variables[1])
results.update({
"{}_{}_covariance".format(h0.getName(), h1.getName()):
self._distribution.getCovariance(h0, h1),
"{}_mean".format(h0.getName()):
self._distribution.getMean(h0),
"{}_mean".format(h1.getName()):
self._distribution.getMean(h1)
})
else:
for state_combinations in itertools.product(
*state_generator(self._discrete_variables)):
state_array = jp.JArray(state_combinations[0].getClass())(
len(state_combinations))
for i, state in enumerate(state_combinations):
state_array[i] = state
dist = Distribution(
self._head_variables, self._tail_variables,
[state.getName() for state in state_combinations])
for i, h in enumerate(self._head_variables):
v = bayesianpy.network.get_variable(
self._network, h)
mean = self._distribution.getMean(v, state_array)
if dist.is_covariant():
dist.append_mean(mean)
for j, h1 in enumerate(self._head_variables):
v1 = bayesianpy.network.get_variable(
self._network, h1)
cov = self._distribution.getCovariance(
v, v1, state_array)
dist.set_covariance_value(i, j, cov)
else:
dist.set_mean_variance(
mean,
self._distribution.getVariance(
v, state_array))
results.update({dist.key(): dist})
return results