Intractable Normalizing Constant
Outside conjugate families the evidence integral has no closed form, and in even modestly high-dimensional parameter spaces numerical quadrature is hopeless — so the posterior is known only up to a constant. This single obstacle is why practical Bayesian computation is built on approximation.
This Concept is waiting for its first lesson!
Outside conjugate families the evidence integral has no closed form, and in even modestly high-dimensional parameter spaces numerical quadrature is hopeless — so the posterior is known only up to a constant. This single obstacle is why practical Bayesian computation is built on approximation.
Are you a teacher? Sign in to start contributing.
Sign In