Alternative Computational Backends


Hey, I am Sharan RY, a sophomore majoring in Mathematics and Computer Science at BITS Pilani, India. I recently came across PyMC3. I was going through the wish-list and found some of the projects really interesting. Especially the one exploring alternative computational back-ends as I am familiar with TensorFlow and have ventured into PyTorch. I am trying familiarize myself with PyMC3 so I can make some meaningful contribution to this project.

Have there been any new developments related to alternative backend? I am asking this because I found discussions regarding PyMC4 backend.

Is this a potential project for GSoC 2018?

Any advice as to how to go about this will be very helpful.



There is currently no official decision regarding new backend. It might be a bit too big for a GSoC 2018 project, but we will certainly plan some backend related projects.


Hey thanks for replying,

Just one small question.
When you say there will be some backend projects planned, do you mean using tensorflow or pytorch OR theano?


I mean project related to choosing a new backends. It could be experiment on either tensorflow, pytorch, mxnet etc, or implementing a few things in all of them to compare.


Also for some context, that wish list @sharan linked was made before we found out Theano development was stopping. You can see for more discussion about the future of this project, and for a summary.