Theory of Intelligence
MACHINE LEARNING•Supervised Learning•Evaluating Classifiers•2.4 ROC Curve & AUC

Introduction

True and false positive rates, the ROC curve that shows every threshold at once, the AUC summary score, and when to prefer the precision–recall curve instead.

timer1 min read•format_list_bulletedConcept 1 of 5•forumLesson Discussion

Introduction

The precision/recall trade-off depends on the threshold. To compare models (rather than thresholds), it helps to have a picture that shows all thresholds at once, and a single number that summarises it. The most widely used are the ROC curve and the area under it (AUC).

ROC stands for receiver operating characteristic — a name from radar engineers in World War II, who had to decide how sensitive to make radar receivers: too sensitive and every flock of birds triggered an alarm; not sensitive enough and enemy planes slipped through. Same trade-off, same tool.

Discuss theoretical concepts, mathematical formulas, and architectural foundations.

Supports inline code with `backticks`
No comments yet. Be the first to start the discussion!
arrow_backPrevious Lesson
2.3 The Precision/Recall Trade-off
Next Lessonarrow_forward
2.5 Error Analysis