I am interested in learning about Variational Bayesian methods. I understand the general idea, explained in Wiki, where the aim is to approximate a posterior using a more tractable distribution, in order to provide an analytical approximation.

However, it is not clear to me how to proceed with new models since the examples typically rely on very simple cases but, for general or more complex models, it is not clear how should one proceed.

Are there any general references where I can learn how to approximate the posterior associated to general regression models (say with random effects, heavy tailed distributions, splines) that are not necessarily linear or normal?

  • $\begingroup$ Machine learning literature has multiple instances of modeling complex non-trivial distributions via variational inference. I would suggest looking at youtube.com/watch?v=ogdv_6dbvVQ for a good lay of the land. $\endgroup$ – randomprime May 18 at 18:00
  • $\begingroup$ @randomprime Many thanks. This looks like a good video, and I know that the speaker is popular in this area. $\endgroup$ – Bari May 18 at 18:07
  • $\begingroup$ The Edward library might be a good start. $\endgroup$ – sdgaw erzswer May 18 at 21:34
  • $\begingroup$ @sdgawerzswer I was thinking of the Library of Alexandria, but that sounds like a good alternative. $\endgroup$ – Bari May 18 at 22:32

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.