def _get_model_training_output(
self,
graph: Graph,
) -> Optional[np.ndarray]:
"""Returns training output tuple."""
if self.is_multilabel_prediction_task():
return graph.get_one_hot_encoded_node_types()
if self.is_binary_prediction_task():
return graph.get_boolean_node_type_ids()
return graph.get_single_label_node_type_ids()
def _evaluate(
self,
graph: Graph,
train: Graph,
test: Graph,
support: Optional[Graph] = None,
node_features: Optional[Union[pd.DataFrame, np.ndarray,
List[Union[str, pd.DataFrame,
np.ndarray]]]] = None,
node_type_features: Optional[Union[pd.DataFrame, np.ndarray,
List[Union[str, pd.DataFrame,
np.ndarray]]]] = None,
edge_features: Optional[Union[pd.DataFrame, np.ndarray,
List[Union[str, pd.DataFrame,
np.ndarray]]]] = None,
subgraph_of_interest: Optional[Graph] = None,
random_state: int = 42,
verbose: bool = True,
) -> List[Dict[str, Any]]:
"""Return model evaluation on the provided graphs."""
train_size = train.get_number_of_known_node_types(
) / graph.get_number_of_known_node_types()
if self.is_multilabel_prediction_task():
labels = graph.get_one_hot_encoded_node_types()
elif self.is_binary_prediction_task():
labels = graph.get_boolean_node_type_ids()
else:
labels = graph.get_single_label_node_type_ids()
performance = []
for evaluation_mode, evaluation_graph in (
("train", train),
("test", test),
):
prediction_probabilities = self.predict_proba(
evaluation_graph,
support=support,
node_features=node_features,
node_type_features=node_type_features,
edge_features=edge_features)
if self.is_binary_prediction_task():
predictions = prediction_probabilities
elif self.is_multilabel_prediction_task():
predictions = prediction_probabilities > 0.5
else:
predictions = prediction_probabilities.argmax(axis=-1)
mask = evaluation_graph.get_known_node_types_mask()
prediction_probabilities = prediction_probabilities[mask]
predictions = predictions[mask]
labels_subset = labels[mask]
performance.append({
"evaluation_mode":
evaluation_mode,
"train_size":
train_size,
"known_nodes_number":
evaluation_graph.get_number_of_known_node_types(),
**self.evaluate_predictions(
labels_subset,
predictions,
),
**self.evaluate_prediction_probabilities(
labels_subset,
prediction_probabilities,
),
})
return performance
