Variational message passing


Variational message passing is an approximate inference technique for continuous- or discrete-valued Bayesian networks, with conjugate-exponential parents, developed by John Winn. VMP was developed as a means of generalizing the approximate variational methods used by such techniques as Latent Dirichlet allocation and works by updating an approximate distribution at each node through messages in the node's Markov blanket.

Likelihood Lower Bound

Given some set of hidden variables and observed variables, the goal of approximate inference is to lower-bound the probability that a graphical model is in the configuration. Over some probability distribution ,
So, if we define our lower bound to be
then the likelihood is simply this bound plus the relative entropy between and. Because the relative entropy is non-negative, the function defined above is indeed a lower bound of the log likelihood of our observation. The distribution will have a simpler character than that of because marginalizing over is intractable for all but the simplest of graphical models. In particular, VMP uses a factorized distribution :
where is a disjoint part of the graphical model.

Determining the Update Rule

The likelihood estimate needs to be as large as possible; because it's a lower bound, getting closer improves the approximation of the log likelihood. By substituting in the factorized version of,, parameterized over the hidden nodes as above, is simply the negative relative entropy between and plus other terms independent of if is defined as
where is the expectation over all distributions except. Thus, if we set to be, the bound is maximized.

Messages in Variational Message Passing

Parents send their children the expectation of their sufficient statistic while children send their parents their natural parameter, which also requires messages to be sent from the co-parents of the node.

Relationship to Exponential Families

Because all nodes in VMP come from exponential families and all parents of nodes are conjugate to their children nodes, the expectation of the sufficient statistic can be computed from the normalization factor.

VMP Algorithm

The algorithm begins by computing the expected value of the sufficient statistics for that vector. Then, until the likelihood converges to a stable value, do the following at each node:
  1. Get all messages from parents
  2. Get all messages from children
  3. Compute the expected value of the nodes sufficient statistics

    Constraints

Because every child must be conjugate to its parent, this limits the types of distributions that can be used in the model. For example, the parents of a Gaussian distribution must be a Gaussian distribution and a gamma distribution. Discrete variables can have Dirichlet parents, and Poisson and exponential nodes must have gamma parents. However, if the data can be modeled in this manner, VMP offers a generalized framework for providing inference.