Skip to content

Probability for ML

Bayes' Rule: Belief Updating

9 min+85 XPStep 1/11

Conditional probability: belief after evidence

Conditional probability — written P(A | B), 'the probability of A given B' — is the formal version of updating your mind on new information.

P(cough) in the general population might be 2%. P(cough | flu) is far higher — conditioning narrows the world to only the cases where flu holds, then asks how common coughs are inside that world.

The trap that catches everyone: P(A|B) and P(B|A) are wildly different. 'Most people with flu cough' does not mean 'most people who cough have flu' — coughs have a hundred other causes.

warning

P(evidence | hypothesis) ≠ P(hypothesis | evidence). Confusing the two is called the prosecutor's fallacy — and it has put innocent people in prison.