Interactive demo · Foundations of Deep Learning · Chapter 2

Bayes' theorem: medical screening and the bent coin

The sum and product rules give p(T=1) = Σc p(T=1 | C=c) p(C=c), and Bayes' theorem turns the test's error rates into what a patient wants to know, p(C=1 | T=1). When the disease is rare, most positive tests come from the large healthy group, so even an accurate test gives a modest posterior. The second view applies the same rule to a bent coin: each flip updates the probability that its concave side is heads, and today's posterior is tomorrow's prior.

View

Prevalence p(C=1)

Sensitivity p(T=1 | C=1)

False-positive rate p(T=1 | C=0)

Positive tests in a row

Prior p(concave side is heads)

Lands concave side up

Flips

Scoreboard

Try these

In the lecture notes