# Variational parameters on prior and likelihood variances

**URL:** <https://discourse.edwardlib.org/t/variational-parameters-on-prior-and-likelihood-variances/268>\
**Category:** General\
**Created:** [July 30, 2017, 5:54am UTC](https://discourse.edwardlib.org/t/variational-parameters-on-prior-and-likelihood-variances/268 "2017-07-30T05:54:52Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![MushroomHunting](https://avatars.discourse-cdn.com/v4/letter/m/90ced4/32.png) [@MushroomHunting](https://discourse.edwardlib.org/u/MushroomHunting)\
**Post date:** [July 30, 2017, 5:54am UTC](https://discourse.edwardlib.org/t/variational-parameters-on-prior-and-likelihood-variances/268/1 "2017-07-30T05:54:52Z")

</div>

Hi there,

I’ve been playing around with the Edward and finding it great with the illustrative examples.  
In the regression tutorial [Edward – Supervised Learning (Regression)](http://edwardlib.org/tutorials/supervised-regression) it states

> assume σw2,σb2\sigma\_w^2,\sigma\_b^2σ​w​2​​,σ​b​2​​ are known prior variances and σy2\sigma\_y^2σ​y​2​​ is a known likelihood variance. The mean of the likelihood is given by a linear transformation of the inputs xn\mathbf{x}\_nx​n​​.

If one wanted to learn these, what kind of considerations would be needed?

For instance, for the prior variances, is it sufficient to put another ed.Normal() for the prior variance like such:

> latent\_vars={W: qW,  
> b: qb,  
> Wsig: qWsig,  
> bsig:qbsig }

and so on for additional latent variable definintions

---

<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:** [August 11, 2017, 4:15pm UTC](https://discourse.edwardlib.org/t/variational-parameters-on-prior-and-likelihood-variances/268/2 "2017-08-11T16:15:13Z")

</div>

Not sure if you solved this already, but you need a prior with positive support. A well-behaving one for training is the log Normal prior. Also see [A toy normal model failed (klqp) and why?](https://discourse.edwardlib.org/t/a-toy-normal-model-failed-klqp-and-why/253/2) for some discussion of priors.

---

<div class="post-metadata">

**Author:** ![MushroomHunting](https://avatars.discourse-cdn.com/v4/letter/m/90ced4/32.png) [@MushroomHunting](https://discourse.edwardlib.org/u/MushroomHunting)\
**Post date:** [August 12, 2017, 12:36am UTC](https://discourse.edwardlib.org/t/variational-parameters-on-prior-and-likelihood-variances/268/3 "2017-08-12T00:36:17Z")

</div>

Cheers;

Yep found out that was the problem along with incorrectly specifying the prior inside the likelihood model

it’s working nicely now