def s2c_sym_predict_staged_estimator(estimator):
'''
sklearn2code adapter
'''
try:
return comp(s2c_sym_predict(estimator.final_stage_), s2c_sym_transform(estimator))
except:
return comp(s2c_sym_predict(estimator.final_stage_), s2c_sym_transform(estimator))
def s2c_sym_transform_staged_estimator(estimator):
'''
sklearn2code adapter
'''
try:
return comp(*map(s2c_sym_transform, estimator.intermediate_stages_))
except:
return comp(*map(s2c_sym_transform, estimator.intermediate_stages_))
def sym_predict_ada_boost_regressor(estimator):
estimators = estimator.estimators_
weights = tuple(map(RealNumber, estimator.estimator_weights_))
predictors = [sym_predict(est) for est in estimators]
predictors = cart(*predictors)
Var = VariableFactory()
predictions = tuple(Var() for _ in predictors.outputs)
return comp(
Function(inputs=predictions,
calls=tuple(),
outputs=sym_weighted_median(predictions, weights)),
predictors)
def sym_predict_voting_classifier(estimator):
inputs = syms(estimator)
voters = cart(*map(sym_predict, estimator.estimators_))
if estimator.weights:
weights = tuple(map(RealNumber, estimator.weights))
else:
weights = tuple(repeat(RealNumber(1), len(voters.outputs)))
Var = VariableFactory(existing=inputs)
votes = tuple(Var() for _ in voters.outputs)
return comp(
sym_inverse_transform(estimator.le_),
Function(inputs=votes,
calls=tuple(),
outputs=sym_weighted_vote(votes, weights)), (voters))
def sym_predict_proba__calibrated_classifier(estimator):
if hasattr(estimator.base_estimator, 'decision_function'):
inner_pred = sym_decision_function(estimator.base_estimator)
elif hasattr(estimator.base_estimator, 'predict_proba'):
inner_pred = sym_predict_proba(estimator.base_estimator)
# inner_pred = fallback(sym_decision_function, sym_predict_proba)(estimator.base_estimator)
pre_result = reduce(__add__, map(
sym_predict, estimator.calibrators_)).compose(inner_pred) / RealNumber(
len(estimator.calibrators_))
Var = VariableFactory()
n_classes = len(estimator.classes_)
if n_classes == 2:
x = Var()
normalize = Function((x, ), tuple(), (RealNumber(1) - x, x))
else:
vars_ = tuple(Var() for _ in range(len(pre_result.outputs)))
total = Function(vars_, tuple(), (reduce(__add__, vars_), ))
t = Var()
normalize = Function(vars_, (((t, ), (total, vars_)), ),
tuple(v / t for v in vars_))
return comp(normalize, pre_result)
def sym_transform_intermediate_stages(estimator):
return comp(*map(compose(sym_transform, tupget(1)), estimator.steps[:-1]))