In short, variational inference is akin to what happens at every presentation you've attended. Someone in the audience asks the presenter a very difficult answer which he/she can't answer. read more
High-Level Explanation of Variational Inference by Jason Eisner (2011) [This was a long email to my reading group in January 2011. See the link below for further reading.] By popular demand, here is a high-level explanation of variational inference, to read before our unit in the NLP reading group. read more
Variational inference (or variational Bayes) is a set of methods that make the computation of certain distributions tractable (as an alternative to MCMC and Gibbs sampling). read more
Variational Bayesian methods are primarily used for two purposes: To provide an analytical approximation to the posterior probability of the unobserved variables, in order to do statistical inference over these variables. read more