λmem.ac
imagesmodepost cover

Bayesian Cats: Updating My Priors

Dinner o’clock?

Bayes’ rule updates a belief in light of evidence:

P(HE)=P(EH)P(H)P(E)P(H \mid E) = \frac{P(E \mid H)\, P(H)}{P(E)}

Our cat’s prior that it’s dinner time is low at 3pm — but the sound of a can opener is strong evidence, so the posterior spikes.

codeExampleexpand_more

Prior: 5%. Likelihood of a can-opener sound given dinner: 90%. The posterior does the rest.

The takeaway

You already reason like this; Bayes just writes it down.