|
Hello, Would you say there is a difference between mean field theory and variational inference, or are they considered subsets of one another, or maybe they are just the same thing. Could you please share your insights. |
|
I would say that the mean field approximation is a particular instance of variational inference where the variational family factorizes completely. It seems however that the term is also used for the case when independence is restricted to hold between partitions of the variables. Still, there are other forms of variational inference, e.g., (loopy) belief-propagation, so I think it makes sense to view mean field as a subset of variational inference. |
