def to_conditional_random_variable(self):
query, evidences = self._remove_conditional()
if ... in evidences:
variables = self._probability_distribution.variables
evidences = Parser.lazy_notation(variables, subset=evidences, ignore=query)
return Conditional(SetOfRandomVariable(query), SetOfRandomVariable(evidences))
def reduction(self, *variables: RandomVariable) -> ProbabilityDistribution:
"""
:param list/tuple variables:
"""
variables = SetOfRandomVariable(variables)
P = self
events = Utils.extract_events(self, variables)
marginalized = P(*variables.only_random_variables())
# Fixme - Any event not presents in joint distribution (index)
try:
reduction = marginalized.series.loc[events]
# Fixme - All possibilities
#new_index = pd.MultiIndex.from_product(events, names=variables.names)
#reduction = reduction.reindex(new_index, fill_value=0, copy=False)
except:
reduction = 0
# FIXME Extract
if not isinstance(reduction, pd.Series):
names = list(variable.name for variable in variables)
if len(names) > 1:
index = pd.MultiIndex.from_arrays(events, names=names)
else:
index = pd.Index(events, name=names[0])
reduction = pd.Series([reduction], index=index)
reduction.name = 'P({})'.format(variables) # FIXME
# FIXME Returns a ProbabilityDistribution, not a Joint distribuion
return JointDistribution(reduction)
def from_series(series: Series) -> 'JointDistribution':
series = series.copy()
variables = tuple(RandomVariable(variable_name) for variable_name in series.index.names)
series.name = Builder._make_name(SetOfRandomVariable(variables))
return JointDistribution(series).normalize() # FIXME
def lazy_notation(
variables: SetOfRandomVariable,
subset: Tuple[Union[RandomVariable, type(...)], ...],
ignore: Union[SetOfRandomVariable,
None] = None) -> SetOfRandomVariable:
"""
Replace ellipsis (...) in subset for variables - subset - ignore
"""
if Ellipsis not in subset:
return SetOfRandomVariable(subset)
ignore = SetOfRandomVariable(tuple()) if ignore is None else ignore
index = subset.index(...)
head = subset[:index]
middle = tuple(set(variables) - set(subset) - set(ignore))
tail = subset[index + 1:]
return SetOfRandomVariable(head + middle + tail)
def joint_distribution(self, *variables: RandomVariable):
variables = Parser.lazy_notation(self.variables, subset=variables)
series = self.series
series = series.groupby(level=variables.names).sum()
new_probability = JointDistribution(series)
new_probability.series.rename('P({})'.format(SetOfRandomVariable(tuple(variables))), inplace=True)
return new_probability
def variables(self) -> SetOfRandomVariable:
index = self.series.index
names = [index.name] if type(index) == pd.Index else index.names
return SetOfRandomVariable(
tuple(RandomVariable(column) for column in names))
def _make_name(variables: SetOfRandomVariable) -> str:
return 'P({})'.format(variables.__repr__())