← Machine Learning Foundations
Free
Loss, Metrics & Evaluation
Pick losses and metrics that match the decision you are making.
Cheatsheet — Metrics
| Situation | Prefer |
|---|---|
| Balanced classes | Accuracy / ROC-AUC |
| Rare positives | PR-AUC, recall@k |
| Outlier-heavy regression | MAE |
| Ranking quality | NDCG / MAP |
Hold out a final test set until the last evaluation.