def test_neuroharmony_behaviour(resources):
"""Test Neuroharmony."""
x_train, x_test = resources.X_train_split, resources.X_test_split
neuroharmony = Neuroharmony(
resources.features,
resources.regression_features,
resources.covariates,
resources.eliminate_variance,
param_distributions=dict(
RandomForestRegressor__n_estimators=[5, 10, 15, 20],
RandomForestRegressor__random_state=[42, 78],
RandomForestRegressor__warm_start=[False, True],
),
estimator_args=dict(n_jobs=1, random_state=42),
randomized_search_args=dict(cv=5, n_jobs=27),
)
x_train_harmonized = neuroharmony.fit_transform(x_train)
x_test_harmonized = neuroharmony.predict(x_test)
data_harmonized = concat([x_train_harmonized, x_test_harmonized],
sort=False)
KS_original = ks_test_grid(resources.original_data, resources.features,
"scanner")
KS_harmonized = ks_test_grid(data_harmonized, resources.features,
"scanner")
assert KS_original[resources.features[0]].shape == (resources.n_scanners,
resources.n_scanners)
assert KS_harmonized[resources.features[0]].shape == (resources.n_scanners,
resources.n_scanners)
assert isinstance(x_test, NDFrame)
assert isinstance(neuroharmony, BaseEstimator)
assert not neuroharmony.prediction_is_covered_.all(
), "No subjects out of the range."
def model(resources):
"""Define mock model.
Parameters
----------
resources : namedtuple
Basic information needed to build a neuroharmony model.
Returns
-------
Neuroharmony class
Neuroharmony model.
"""
neuroharmony = Neuroharmony(
resources.features,
resources.regression_features,
resources.covariates,
resources.eliminate_variance,
param_distributions=dict(
RandomForestRegressor__n_estimators=[5, 10, 15, 20],
RandomForestRegressor__random_state=[42, 78],
RandomForestRegressor__warm_start=[False, True],
),
estimator_args=dict(n_jobs=1, random_state=42),
randomized_search_args=dict(cv=5, n_jobs=27),
)
return neuroharmony
def test_model_strategy(resources, model_strategy):
"""Test the model_strategy is implemented."""
x_train, x_test = resources.X_train_split, resources.X_test_split
neuroharmony = Neuroharmony(
resources.features,
resources.regression_features,
resources.covariates,
resources.eliminate_variance,
param_distributions=dict(
RandomForestRegressor__n_estimators=[5, 10, 15, 20],
RandomForestRegressor__random_state=[42, 78],
RandomForestRegressor__warm_start=[False, True],
),
model_strategy=model_strategy,
estimator_args=dict(n_jobs=1, random_state=42),
randomized_search_args=dict(cv=5, n_jobs=27),
)
x_train_harmonized = neuroharmony.fit_transform(x_train)
x_test_harmonized = neuroharmony.predict(x_test)
assert isinstance(x_train_harmonized, NDFrame)
assert isinstance(x_test_harmonized, NDFrame)
def model(resources):
neuroharmony = Neuroharmony(
resources.features,
resources.regression_features,
resources.covariates,
resources.eliminate_variance,
param_distributions=dict(
RandomForestRegressor__n_estimators=[5, 10, 15, 20],
RandomForestRegressor__random_state=[42, 78],
RandomForestRegressor__warm_start=[False, True],
),
estimator_args=dict(n_jobs=1, random_state=42),
randomized_search_args=dict(cv=5, n_jobs=27),
)
return neuroharmony
def test_model_invalid_strategy(resources):
"""Test the model_strategy is implemented."""
with pytest.raises(ValueError):
Neuroharmony(
resources.features,
resources.regression_features,
resources.covariates,
resources.eliminate_variance,
param_distributions=dict(
RandomForestRegressor__n_estimators=[5, 10, 15, 20],
RandomForestRegressor__random_state=[42, 78],
RandomForestRegressor__warm_start=[False, True],
),
model_strategy=None,
estimator_args=dict(n_jobs=1, random_state=42),
randomized_search_args=dict(cv=5, n_jobs=27),
)