gigl.src.common.utils.eval_metrics#

Functions#

hit_rate_at_k(pos_scores, neg_scores, ks)

Computes Hit Rate @ K metrics for various Ks, evaluating 1+ positives against 1+ negatives.

mean_reciprocal_rank(pos_scores, neg_scores)

Computes Mean Reciprocal Rank (MRR), evaluating 1+ positives against 1+ negatives.

Module Contents#

gigl.src.common.utils.eval_metrics.hit_rate_at_k(pos_scores, neg_scores, ks)[source]#

Computes Hit Rate @ K metrics for various Ks, evaluating 1+ positives against 1+ negatives.

Parameters:

  • pos_scores (torch.FloatTensor) – Contains 1 or more positive sample scores.

  • neg_scores (torch.FloatTensor) – Contains 1 or more negative sample scores.

  • ks (torch.LongTensor) – k-values for which to compute hits.

Returns:

Hit rates corresponding to the requested ks.

Return type:

torch.FloatTensor

gigl.src.common.utils.eval_metrics.mean_reciprocal_rank(pos_scores, neg_scores)[source]#

Computes Mean Reciprocal Rank (MRR), evaluating 1+ positives against 1+ negatives.

Parameters:

  • pos_scores (torch.FloatTensor) – Contains 1 or more positive sample scores.

  • neg_scores (torch.FloatTensor) – Contains 1 or more negative sample scores.

Returns:

Computed MRR score.

Return type:

torch.FloatTensor