P(A,|,B) = dfracP(B,|,A),P(A)P(B), or in continuous form p(θ,|,x) propto p(x,|,θ),p(θ) — posterior is proportional to likelihood times prior. The basis of Bayesian inference (Black-Litterman, Kalman filter, particle filter, MCMC). The denominator P(B) = int p(x,|,θ)p(θ),dθ is the marginal likelihood used in model selection; its intractability motivates variational and Monte Carlo posterior approximations.