Welcome to Variational-Bayes.org.
This is a repository of papers, software, and links related to the use
of variational methods for approximate Bayesian learning.
What?
Variational Bayesian (VB) methods, also called "ensemble
learning", are a family of techniques for approximating
intractable integrals arising in Bayesian statistics and machine
learning. They are an alternative to other approaches for approximate
Bayesian inference such as Markov chain Monte Carlo, the Laplace
approximation, etc. They can be used to lower bound the marginal
likelihood (i.e. "evidence") of several models with a view to
performing model selection, and often provide an analytical
approximation to the parameter posterior which is useful for
prediction.
NOTE: Although there are a variety of variational methods that can be
used to approximate Bayesian integrals, in this repository we will
focus on methods that provide bounds on these integrals
(e.g. by using Jensen's inequality).
Why?
We hope that this resource will be useful to keep
researchers and practitioners in statistics and machine learning
abreast of this field. Having a single repository for papers and
software in this area will make it easy to keep track of and build on
the work of others.
Please help:
We would kindly ask you to keep us informed of
new papers, software, suggestions, and older papers that were missed
on our first pass.
We plan to include:
all relevant papers that use
variational methods to approximate Bayesian integrals (i.e. integrals
over model parameters and latent variables),
theoretical papers that analyse variational Bayesian methods, and
papers with novel applications of VB methods.
Simply follow the submit link to contact either of us.