fix: Compute correct ROC-AUC score with link_prediction labels in one class

[colinbarry]

Fixes a flaky MAGE tests where occasionally, the graph - which is randomly generated - can have all labels in a single class. This is because ROC-AUC is undefined without both positives and negatives. This change returns a neutral 0.5 AUC in that edge case. Also adds deterministic tests for all-positive, all-negative, and mixed-label scenarios.

---

Note

Low Risk
Small, localized change to metric computation with added tests; behavior only differs in edge cases where prior code errored or produced inconsistent confusion-matrix shapes.

Overview
Fixes flaky link_prediction_util.evaluate() metric computation when a batch contains only one label class.

roc_auc_score is now guarded to return 0.5 for single-class labels, and confusion-matrix values are computed lazily with labels=[0,1] only when TP/FP/TN/FN metrics are requested. Adds pytest coverage for AUC on all-positive, all-negative, and mixed-label cases.

^{Written by Cursor Bugbot for commit ac8ed5f. This will update automatically on new commits. Configure here.}

[colinbarry]

Tracking

• [Link to Epic/Issue]

Standard development

• Update unit/E2E tests
• Compare the benchmarking results between the master branch this branch

CI Testing Labels

• Select the appropriate CI test labels (CI -build=build-name -test=test-suite)

Documentation checklist

• Add the documentation label
• Add the bug / feature label
• Add the milestone for which this feature is intended
  • If not known, set for a later milestone
• Write a release note, including added/changed clauses
  • link_prediction metric evaluation now handles single-class label batches safely, returning a neutral AUC_SCORE of 0.5 when a batch contains only one class. #3845

[sonarqubecloud]

Quality Gate passed

Issues
1 New issue
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[colinbarry]

BugBot run

[cursor]

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR