# Multi-inputs multi-outputs model support

**URL:** <https://discourse.edwardlib.org/t/multi-inputs-multi-outputs-model-support/233>\
**Category:** General\
**Created:** [July 12, 2017, 4:28pm UTC](https://discourse.edwardlib.org/t/multi-inputs-multi-outputs-model-support/233 "2017-07-12T16:28:20Z")\
**Posts on this page:** 5\
**Page:** 1

<div class="post-metadata">

**Author:** ![ErmiaAzarkhalili](https://yyz1.discourse-cdn.com/flex035/user_avatar/discourse.edwardlib.org/ermiaazarkhalili/32/65_2.png) [@ErmiaAzarkhalili](https://discourse.edwardlib.org/u/ErmiaAzarkhalili)\
**Post date:** [July 12, 2017, 4:28pm UTC](https://discourse.edwardlib.org/t/multi-inputs-multi-outputs-model-support/233/1 "2017-07-12T16:28:20Z")

</div>

Hi Dastin,  
Could you please let me know whether Edward support multi-inputs multi-outputs neural networks?

---

<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:** [July 12, 2017, 7:45pm UTC](https://discourse.edwardlib.org/t/multi-inputs-multi-outputs-model-support/233/2 "2017-07-12T19:45:55Z")

</div>

Could you describe what multi-input/output is? Generally speaking, if the generative process can be described then you can write an Edward program for it. For example, multi-class classification would use a Multinomial distribution for the output.

---

<div class="post-metadata">

**Author:** ![ErmiaAzarkhalili](https://yyz1.discourse-cdn.com/flex035/user_avatar/discourse.edwardlib.org/ermiaazarkhalili/32/65_2.png) [@ErmiaAzarkhalili](https://discourse.edwardlib.org/u/ErmiaAzarkhalili)\
**Post date:** [July 13, 2017, 12:28am UTC](https://discourse.edwardlib.org/t/multi-inputs-multi-outputs-model-support/233/3 "2017-07-13T00:28:29Z")

</div>

So so sorry for my bad explanation, and of course, I certainly know what you explained.

Consider conditional variational autoencoder where we have two inputs as an example of multi-inputs multi-outputs.

Multi-inputs multi-outputs means you have multiple datasets as inputs and want to learn multiple neural networks simultaneously.

Does it clear now? 🙂

---

<div class="post-metadata">

**Author:** ![ErmiaAzarkhalili](https://yyz1.discourse-cdn.com/flex035/user_avatar/discourse.edwardlib.org/ermiaazarkhalili/32/65_2.png) [@ErmiaAzarkhalili](https://discourse.edwardlib.org/u/ErmiaAzarkhalili)\
**Post date:** [August 4, 2017, 11:54pm UTC](https://discourse.edwardlib.org/t/multi-inputs-multi-outputs-model-support/233/5 "2017-08-04T23:54:01Z")

</div>

Thanks man, but I meant something completely different. 😉

let me explain by an example, can we use Edward to solve conditional variational autoencoder where we have two inputs in our model?

---

<div class="post-metadata">

**Author:** ![ErmiaAzarkhalili](https://yyz1.discourse-cdn.com/flex035/user_avatar/discourse.edwardlib.org/ermiaazarkhalili/32/65_2.png) [@ErmiaAzarkhalili](https://discourse.edwardlib.org/u/ErmiaAzarkhalili)\
**Post date:** [September 1, 2017, 5:32pm UTC](https://discourse.edwardlib.org/t/multi-inputs-multi-outputs-model-support/233/6 "2017-09-01T17:32:35Z")

</div>

Hi Dustin,  
Do you have any idea to model conditional variational autoencoder by Edward? I mean to model multi-inputs multi-outputs Keras model by Edward.