incerto.sp.plot_risk_coverage

incerto.sp.plot_risk_coverage#

incerto.sp.plot_risk_coverage(logits, y, confidence=None, *, ax=None, show_aurc=True, show_bounds=True)[source]#

Plot the risk-coverage curve for selective prediction.

Parameters:

  • logits (Tensor) – Model logits.

  • y (Tensor) – Ground truth labels.

  • confidence (Optional[Tensor]) – Confidence scores (if None, uses max softmax).

  • ax (Optional[Axes]) – Matplotlib axes object.

  • show_aurc (bool) – Whether to show AURC in title.

  • show_bounds (bool) – Whether to show random and ideal (oracle) bounds.

Return type:

Axes

Returns:

Matplotlib axes object.

Contents