pydgc.metrics

pydgc.metrics.utils module

class DGCMetric(ground_truth, predicted_labels, embeddings, edge_index)[source]

Bases: object

DGC metric class.

Parameters:

  • ground_truth (np.array) – Ground truth labels.

  • predicted_labels (np.array) – Predicted labels.

  • embeddings (Tensor) – Node embeddings.

  • edge_index (Tensor) – Edge index.

accuracy(decimal=4)[source]

Calculate clustering accuracy after using the linear_sum_assignment function in SciPy to determine reassignments.

Parameters:

decimal (int, optional) – The number of decimal places that need to be retained. Defaults to 4.

Returns:

Clustering accuracy.

Return type:

float

f1_score(decimal=4)[source]

Calculate F1 score.

Parameters:

decimal (int, optional) – The number of decimal places that need to be retained. Defaults to 4.

Returns:

F1 score.

Return type:

float

nmi_score(decimal=4)[source]

Calculate NMI score.

Parameters:

decimal (int, optional) – The number of decimal places that need to be retained. Defaults to 4.

Returns:

NMI score.

Return type:

float

ari_score(decimal=4)[source]

Calculate ARI score.

Parameters:

decimal (int, optional) – The number of decimal places that need to be retained. Defaults to 4.

Returns:

ARI score.

Return type:

float

hom_score(decimal=4)[source]

Calculate homogeneity score.

Parameters:

decimal (int, optional) – The number of decimal places that need to be retained. Defaults to 4.

Returns:

Homogeneity score.

Return type:

float

com_score(decimal=4)[source]

Calculate completeness score.

Parameters:

decimal (int, optional) – The number of decimal places that need to be retained. Defaults to 4.

Returns:

Completeness score.

Return type:

float

sil_score(decimal=4)[source]

Calculate silhouette score.

Parameters:

decimal (int, optional) – The number of decimal places that need to be retained. Defaults to 4.

Returns:

Silhouette score.

Return type:

float

graph_reconstruction_error(decimal=4)[source]

Calculate graph reconstruction error.

Parameters:

decimal (int, optional) – The number of decimal places that need to be retained. Defaults to 4.

Returns:

Graph reconstruction error.

Return type:

float

purity(decimal=4)[source]

Calculate purity score.

Parameters:

decimal (int, optional) – The number of decimal places that need to be retained. Defaults to 4.

Returns:

Purity score.

Return type:

float

evaluate_one_epoch(logger, cfg=None)[source]

Evaluate one epoch.

Parameters:

  • logger (Logger) – Logger.

  • cfg (CN, optional) – Config. Defaults to None.

Returns:

Results with metric names as keys and metric values as values.

Return type:

dict

build_results_dict(cfg)[source]

Build results dict.

Parameters:

cfg (CN) – Config.

Returns:

Results dict.

Return type:

dict