wrapper for Edward

[Benjamin Feifke]

Hello!

I've been using Stan for a little bit, and just discovered Edward. (First, I must say, your work is amazing; it's really helping me to prototype graphical bayesian models in an easily testable way, I really appreciate it).
I know this isn't the right place to ask about Edward, but I wasn't sure what was.
My question: Is there anyway to specify model_code using Stan, or save a model using Stan, and then input it to the Edward pipeline to leverage its substantial speed increase etc?

Thanks!
Ben

[Benjamin Feifke]

Answered my own question: http://edwardlib.org/api/model-wrappers

[Michael Betancourt]

To be clear, do not expect anything close to the speedups claimed

in the Edward paper.  They are very misleading and will disintegrate

when you try to build the kinds of nontrivial models for which Stan is

targeted.

--
You received this message because you are subscribed to the Google Groups "Stan users mailing list" group.
To unsubscribe from this group and stop receiving emails from it, send an email to stan-users+...@googlegroups.com.
To post to this group, send email to stan-...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Andrew Gelman]

We love when people answer their own questions!

Also, if you do get this to work for some interesting examples, plz share with the list.

A

On Jan 19, 2017, at 12:53 PM, Benjamin Feifke <benf...@gmail.com> wrote:

[Bob Carpenter]

And if you don't get the speedup you are expecting, please
report that to the list, too! Or to Edward's lists.

- Bob