Basics of Bayesian learning — LessWrong