def fit(self, X, y,
X_test=None,
y_test=None,
metric=None,
feat_type=None,
dataset_name=None):
# TODO
# regularly check https://github.com/scikit-learn/scikit-learn/issues/15336 whether
# histogram gradient boosting in scikit-learn finally support sparse data
is_sparse = scipy.sparse.issparse(X)
if is_sparse:
include_estimators = [
'extra_trees', 'passive_aggressive', 'random_forest', 'sgd', 'mlp',
]
else:
include_estimators = [
'extra_trees',
'passive_aggressive',
'random_forest',
'sgd',
'gradient_boosting',
'mlp',
]
self.include['classifier'] = include_estimators
if self.metric is None:
if len(y.shape) == 1 or y.shape[1] == 1:
self.metric = accuracy
else:
self.metric = log_loss
if self.metric in metrics:
metric_name = self.metric.name
selector_file = selector_files[metric_name]
else:
metric_name = 'balanced_accuracy'
selector_file = selector_files[metric_name]
with open(selector_file, 'rb') as fh:
selector = pickle.load(fh)
metafeatures = pd.DataFrame({dataset_name: [X.shape[1], X.shape[0]]}).transpose()
selection = np.argmax(selector.predict(metafeatures))
automl_policy = strategies[selection]
setting = {
'RF_None_holdout_iterative_es_if': {
'resampling_strategy': 'holdout-iterative-fit',
'fidelity': None,
},
'RF_None_3CV_iterative_es_if': {
'resampling_strategy': 'cv-iterative-fit',
'folds': 3,
'fidelity': None,
},
'RF_None_5CV_iterative_es_if': {
'resampling_strategy': 'cv-iterative-fit',
'folds': 5,
'fidelity': None,
},
'RF_None_10CV_iterative_es_if': {
'resampling_strategy': 'cv-iterative-fit',
'folds': 10,
'fidelity': None,
},
'RF_SH-eta4-i_holdout_iterative_es_if': {
'resampling_strategy': 'holdout-iterative-fit',
'fidelity': 'SH',
},
'RF_SH-eta4-i_3CV_iterative_es_if': {
'resampling_strategy': 'cv-iterative-fit',
'folds': 3,
'fidelity': 'SH',
},
'RF_SH-eta4-i_5CV_iterative_es_if': {
'resampling_strategy': 'cv-iterative-fit',
'folds': 5,
'fidelity': 'SH',
},
'RF_SH-eta4-i_10CV_iterative_es_if': {
'resampling_strategy': 'cv-iterative-fit',
'folds': 10,
'fidelity': 'SH',
}
}[automl_policy]
resampling_strategy = setting['resampling_strategy']
if resampling_strategy == 'cv-iterative-fit':
resampling_strategy_kwargs = {'folds': setting['folds']}
else:
resampling_strategy_kwargs = None
portfolio_file = (
this_directory / metric_name / 'askl2_portfolios' / ('%s.json' % automl_policy)
)
with open(portfolio_file) as fh:
portfolio_json = json.load(fh)
portfolio = portfolio_json['portfolio']
if setting['fidelity'] == 'SH':
smac_callback = SHObjectCallback('iterations', 4, 5.0, portfolio)
else:
smac_callback = SmacObjectCallback(portfolio)
self.resampling_strategy = resampling_strategy
self.resampling_strategy_arguments = resampling_strategy_kwargs
self.get_smac_object_callback = smac_callback
return super().fit(
X=X,
y=y,
X_test=X_test,
y_test=y_test,
feat_type=feat_type,
dataset_name=dataset_name,
)
def fit(self,
X,
y,
X_test=None,
y_test=None,
metric=None,
feat_type=None,
dataset_name=None):
with open(selector_file, 'rb') as fh:
selector = pickle.load(fh)
metafeatures = np.array([len(np.unique(y)), X.shape[1], X.shape[0]])
selection = np.argmax(selector.predict(metafeatures))
automl_policy = strategies[selection]
setting = {
'RF_None_holdout_iterative_es_if': {
'resampling_strategy': 'holdout-iterative-fit',
'fidelity': None,
},
'RF_None_3CV_iterative_es_if': {
'resampling_strategy': 'cv-iterative-fit',
'folds': 3,
'fidelity': None,
},
'RF_None_5CV_iterative_es_if': {
'resampling_strategy': 'cv-iterative-fit',
'folds': 5,
'fidelity': None,
},
'RF_None_10CV_iterative_es_if': {
'resampling_strategy': 'cv-iterative-fit',
'folds': 10,
'fidelity': None,
},
'RF_SH-eta4-i_holdout_iterative_es_if': {
'resampling_strategy': 'holdout-iterative-fit',
'fidelity': 'SH',
},
'RF_SH-eta4-i_3CV_iterative_es_if': {
'resampling_strategy': 'cv-iterative-fit',
'folds': 3,
'fidelity': 'SH',
},
'RF_SH-eta4-i_5CV_iterative_es_if': {
'resampling_strategy': 'cv-iterative-fit',
'folds': 5,
'fidelity': 'SH',
},
'RF_SH-eta4-i_10CV_iterative_es_if': {
'resampling_strategy': 'cv-iterative-fit',
'folds': 10,
'fidelity': 'SH',
}
}[automl_policy]
resampling_strategy = setting['resampling_strategy']
if resampling_strategy == 'cv-iterative-fit':
resampling_strategy_kwargs = {'folds': setting['folds']}
else:
resampling_strategy_kwargs = None
portfolio_file = os.path.join(this_directory, 'askl2_portfolios',
'%s.json' % automl_policy)
with open(portfolio_file) as fh:
portfolio_json = json.load(fh)
portfolio = portfolio_json['portfolio']
if setting['fidelity'] == 'SH':
smac_callback = get_sh_object_callback('iterations', 4, 5.0,
portfolio)
else:
smac_callback = get_smac_object_callback(portfolio)
self.resampling_strategy = resampling_strategy
self.resampling_strategy_arguments = resampling_strategy_kwargs
self.get_smac_object_callback = smac_callback
return super().fit(
X=X,
y=y,
X_test=X_test,
y_test=y_test,
feat_type=feat_type,
dataset_name=dataset_name,
)
def fit(self, X, y,
X_test=None,
y_test=None,
metric=None,
feat_type=None,
dataset_name=None):
if self.metric is None:
if len(y.shape) == 1 or y.shape[1] == 1:
self.metric = accuracy
else:
self.metric = log_loss
if self.metric in metrics:
metric_name = self.metric.name
selector_file = selector_files[metric_name]
else:
metric_name = 'balanced_accuracy'
selector_file = selector_files[metric_name]
with open(selector_file, 'rb') as fh:
selector = pickle.load(fh)
metafeatures = pd.DataFrame({dataset_name: [X.shape[1], X.shape[0]]}).transpose()
selection = np.argmax(selector.predict(metafeatures))
automl_policy = strategies[selection]
setting = {
'RF_None_holdout_iterative_es_if': {
'resampling_strategy': 'holdout-iterative-fit',
'fidelity': None,
},
'RF_None_3CV_iterative_es_if': {
'resampling_strategy': 'cv-iterative-fit',
'folds': 3,
'fidelity': None,
},
'RF_None_5CV_iterative_es_if': {
'resampling_strategy': 'cv-iterative-fit',
'folds': 5,
'fidelity': None,
},
'RF_None_10CV_iterative_es_if': {
'resampling_strategy': 'cv-iterative-fit',
'folds': 10,
'fidelity': None,
},
'RF_SH-eta4-i_holdout_iterative_es_if': {
'resampling_strategy': 'holdout-iterative-fit',
'fidelity': 'SH',
},
'RF_SH-eta4-i_3CV_iterative_es_if': {
'resampling_strategy': 'cv-iterative-fit',
'folds': 3,
'fidelity': 'SH',
},
'RF_SH-eta4-i_5CV_iterative_es_if': {
'resampling_strategy': 'cv-iterative-fit',
'folds': 5,
'fidelity': 'SH',
},
'RF_SH-eta4-i_10CV_iterative_es_if': {
'resampling_strategy': 'cv-iterative-fit',
'folds': 10,
'fidelity': 'SH',
}
}[automl_policy]
resampling_strategy = setting['resampling_strategy']
if resampling_strategy == 'cv-iterative-fit':
resampling_strategy_kwargs = {'folds': setting['folds']}
else:
resampling_strategy_kwargs = None
portfolio_file = (
this_directory / metric_name / 'askl2_portfolios' / ('%s.json' % automl_policy)
)
with open(portfolio_file) as fh:
portfolio_json = json.load(fh)
portfolio = portfolio_json['portfolio']
if setting['fidelity'] == 'SH':
smac_callback = SHObjectCallback('iterations', 4, 5.0, portfolio)
else:
smac_callback = SmacObjectCallback(portfolio)
self.resampling_strategy = resampling_strategy
self.resampling_strategy_arguments = resampling_strategy_kwargs
self.get_smac_object_callback = smac_callback
return super().fit(
X=X,
y=y,
X_test=X_test,
y_test=y_test,
feat_type=feat_type,
dataset_name=dataset_name,
)