Bayes' rule: Proportional form — LessWrong