# Structured variational inference

**URL:** <https://discourse.edwardlib.org/t/structured-variational-inference/342>\
**Category:** General\
**Created:** [September 6, 2017, 1:40pm UTC](https://discourse.edwardlib.org/t/structured-variational-inference/342 "2017-09-06T13:40:40Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![ajbc](https://yyz1.discourse-cdn.com/flex035/user_avatar/discourse.edwardlib.org/ajbc/32/15_2.png) [@ajbc](https://discourse.edwardlib.org/u/ajbc)\
**Post date:** [September 6, 2017, 1:40pm UTC](https://discourse.edwardlib.org/t/structured-variational-inference/342/1 "2017-09-06T13:40:40Z")

</div>

Is there any way to specify dependencies between variational parameters, or perform structured variational inference ([http://proceedings.mlr.press/v38/hoffman15.pdf](http://proceedings.mlr.press/v38/hoffman15.pdf))?

---

<div class="post-metadata">

**Author:** ![dustin](https://yyz1.discourse-cdn.com/flex035/user_avatar/discourse.edwardlib.org/dustin/32/134_2.png) [@dustin](https://discourse.edwardlib.org/u/dustin)\
**Post date:** [September 9, 2017, 12:14am UTC](https://discourse.edwardlib.org/t/structured-variational-inference/342/2 "2017-09-09T00:14:32Z")

</div>

There’s a structured variational approximation in the [`sigmoid_belief_network.py`](https://github.com/blei-lab/edward/blob/master/examples/sigmoid_belief_network.py) example. I recall fixing a bug to get it to work so you may need Edward’s dev version.

Matt’s SSVI isn’t implemented but an example building and applying it are welcome of course.