Garfield.trainer.eval_metrics

Garfield.trainer.eval_metrics(edge_recon_probs: Tensor | ndarray, edge_labels: Tensor | ndarray, omics_recon_pred: Tensor | ndarray | None = None, omics_recon_truth: Tensor | ndarray | None = None) → dict

Get the evaluation metrics for a (balanced) sample of positive and negative edges and a sample of nodes.

Parameters:

• edge_recon_probs – Tensor or array containing reconstructed edge probabilities.

• edge_labels – Tensor or array containing ground truth labels of edges.

Returns:

Dictionary containing the evaluation metrics ´auroc_score´ (area under the receiver operating characteristic curve), ´auprc score´ (area under the precision-recall curve), ´best_acc_score´ (accuracy under optimal classification threshold) and ´best_f1_score´ (F1 score under optimal classification threshold).

Return type:

eval_dict