Comment by fishinkiko on Information theory and the symmetry of updating beliefs · 2019-03-29T22:24:49.718Z · LW · GW

Hi - I am just puzzled why inf(A|B) = inf(A) - iev(A, B).

Here is my derivation: inf(A|B) = - log(P(A|B)) = - log(P(A)*pev(A, B)) = -log(P(A)) - log(pev(A, B)) = inf(A) + iev(A, B).

Could you please tell me where is wrong? Thank you very much!